Today I went to the Perplexity user meetup. I didn't go there because it was special, but because I'm a pro user. When I got there and tried to sign up for the registration list, I found out that I was the second pro user to sign up. Today's user meetup was hosted by a magazine company called The Milk. To sum up today's meetup in one sentence, "Perplexity compensated for the time I was annoyed by the organizers?" When Sam Altman from OpenAI visited Korea last time, he attended an event hosted by the Ministry of SMEs and Startups, and there was a Q&A session. Those who were given the opportunity to ask questions about something prepared a questionnaire and asked questions in English as if they knew they would be asked questions. Today, I don't know what happened, but the representative who participated as a moderator gave the opportunity to ask questions during the Q&A time. It was really coincidental. You are good at asking long questions in English, and you are in a related industry, or by chance, you are a team leader that the CEO knows^^ I thought about it from a different perspective. What would it have been like if I were there? I think it would have been nice to see a familiar face among the many audience members. I also thought, what if it had been a private event where only their subscribers were invited? After some unpleasant times, it was time for Perplexity users only. But at this time, many people left. A full-fledged user meetup led by the lead in charge of Asia Pacific! Before starting, he said that many of his enthusiastic fans asked him questions, so he prepared a gift. And then suddenly my name was called. I received a hat as a gift and took a commemorative photo with the CEO of Perplexity. The overall event today is It was conducted using Eventcat, which uses simultaneous interpretation technology, and it seems to have replaced about 95% of the work of simultaneous interpretation. At first, I thought it was XR AI, but it was called XL8. I had misheard it as ExcelPal. However, XR AI Glass also has very good simultaneous interpretation capabilities. Yesterday, Perplexity announced that it will provide SKT customers with the Pro version for free for one year. I started using Perplxity shortly after it came out, and on August 5, 2023, when there was a 20% discount, I made my first payment and made another payment a while ago. I have subscriptions to ChatGPT, Claude, and Perplexity, but the one I am most satisfied with and recommend the most to people is Perplexity. There are 9 people who have signed up for a paid membership through my referral link. The reason I am satisfied is that it is fast. It's not that the speed is fast, but the response of the Perpelxity team is really fast.

Based on Q2 2024 US startup investment data released by Carta, we take a look at the key trends in the current investment market. 1. From Seed to Series C: The Numbers of Startup Growth Seed : Median enterprise value $15.0M, investment amount $3.5M Series A : Enterprise value $40.0M, investment amount $10.1M Series B : Enterprise value $116.3M, investment amount raised $22.7M Series C : Enterprise value $214.8M, Investment amount raised $31.0M 2. 15x increase in corporate value: The gap between Seed and Series C Comparing the median Seed stage value ($15.0M) to the median Series C stage value ($214.8M), there is a difference of approximately 14.3x, showing the rapid increase in valuation of successful startups. 3. Evolution of investment size: From $3.5M to $31M From the average investment amount at the Seed stage ($3.5M) to the Series C stage ($31.0M), the investment size increases by approximately 8.9 times. You can see that startups’ demand for funding increases significantly at each stage. 4. The Secret of Dilution: Why is it the Lowest in Series C? Seed: 20.5% Series A: 19.8% Series B: 14.0% Series C: 11.1% As the investment stage increases, the dilution rate tends to decrease, suggesting that later-stage startups can attract investment on more favorable terms. 5. Step-by-step polarization: The reality of the top 25% and the bottom 25% There is a significant gap when comparing the 25th percentile to the 75th percentile at each stage. For example, at the Series C stage: 25th Percentile Enterprise Value: $115.2M 75th Percentile Enterprise Value: $438.8M This shows that there are significant differences in performance and evaluation between startups even within the same stage.

The Chrome browser can greatly enhance your user experience through various plugins. This guide introduces 35 useful Chrome plugins in various fields such as productivity, AI support, content management, and more. Writing and grammar support Grammarly Real-time grammar and spell checking Suggestions for improving sentence structure Helps to improve clarity and tone of writing Hyperwrite AI-based writing support Sentence structure improvement and idea suggestions Wordtune AI-based writing improvement suggestions Provides options to change synonyms and sentence structure Texta.ai AI-based writing tool Support for various writing styles AI Chat and Content Creation ChatSonic AI Chatbot Features Includes image generation capabilities NoteGPT AI-based note-taking tool

https://youtu.be/qZT57PZXG3o?si=GR_nqQBeXrGp-PHa We will explain in detail the amazing capabilities of Claude AI and the differences from ChatGPT and Gemini. We will learn how to use Claude in everyday life and business, what to watch out for, and suggest ways to prepare for the AI era. The emergence and growth of Claude AI Developed by Anthropic in 2022, Claude has emerged as a versatile AI assistant that can be used for everything from personal questions to professional discussions. It currently has around 50 million users worldwide, and is particularly popular in Japan. This is a small number compared to ChatGPT’s 160 million and Gemini’s 180 million, but it is growing rapidly. Claude's secret to growth lies in its excellent language processing ability and user-friendly interface. It is particularly well-received among Japanese users due to its excellent Japanese processing ability, which is an important factor in Claude's competitiveness in the global market. ChatGPT vs Gemini Claude is comparable to ChatGPT and Gemini in natural language processing, multilingual support, and coding assistance. Claude is particularly outstanding in the area of programming assistance. In complex code generation and debugging, and algorithm explanation, Claude often outperforms ChatGPT. On the other hand, Claude is relatively weak in the image recognition and speech recognition fields. Gemini is strong in these fields, especially in multimodal functions. However, the Claude team is continuously updating to reduce this gap, so it is expected to be competitive in this field in the future. In terms of user experience, Claude is well-received for its intuitive interface and fast response speed. While ChatGPT sometimes experiences connection delays due to overload, Claude provides stable service. Claude's Key Features and Usage Sentence generation and editing: Claude helps you write a variety of documents, including blog posts, reports, and emails. For example, if you ask, “Write a 1,000-word essay on the ethical aspects of AI technology,” Claude will generate a structured essay that you can edit and supplement as needed. Coding assistance: Answer programming language-related questions and generate code. For requests like "Create a web scraping program in Python," Claude provides detailed code and explanations. Data Analysis: Assists with data interpretation and visualization. For example, for a request like “Analyze and graph global temperature change data over the past 5 years,” Claude analyzes the data and provides visualization code. Language Learning: Assists with foreign language learning and translation. In response to a request to "write an email to schedule a meeting in business English," Claude writes an email using appropriate expressions and structure. Creative work: Supporting idea generation, storytelling, etc. In response to the request “Come up with a plot for a short story on the topic of environmental protection,” Claude comes up with a creative storyline. Precautions when using Claude Lack of Up-to-Date Information: Claude’s knowledge is limited to August 2023, so information after that date may be inaccurate. Be mindful of this when asking about the latest news or events. Fact-checking is required: Always cross-check AI answers with other sources. Especially in critical decision-making or academic work, don’t blindly trust Claude’s answers, but do additional verification. Privacy: Avoid entering sensitive personal information. Your conversations with Claude may be stored on Anthropic’s servers, so we recommend that you do not enter sensitive data such as financial or medical information. Ethical Use: You may not use Claude for any purpose that infringes copyright or is illegal. For example, you may not use Claude to copy someone else's work without permission or to create harmful content. Attitude toward preparing for the AI era

https://youtu.be/fv1rkctrEPk?si=8cKHOvAgzzk5PuZK We will introduce the process of developing a keyword search app in YouTube videos using Claude 3.5 and Cursor IDE. Efficient coding and fast problem solving are possible with the help of AI. Experience the practicality of AI tools firsthand with this project. Start developing apps with AI The goal of this project was to create an app that would transcribe the content of a YouTube video URL and allow users to search for keywords and jump directly to that part. During the development process, we used AI tools called Claude 3.5 and Cursor IDE. Building the Frontend: React Meets AI First, we built the front-end using React. With the help of Claude 3.5, we were able to quickly design the basic component structure and styling. We also easily improved the UI, such as applying a dark theme, using Cursor IDE. Backend Development: Experience the Power of AI AI’s true value is further demonstrated in backend development. AI’s advice has been a great help in implementing complex functions such as downloading YouTube videos, transcription, and text processing. In particular, AI’s help has been crucial in implementing the transcription function using OpenAI’s Whisper API. A new paradigm for problem solving AI played a big role in resolving many bugs and errors that occurred during the development process. Cursor IDE’s AI-based debugging function allowed us to quickly identify and fix problems in our code, which was much more efficient than traditional development methods. Improving User Experience: The Fine Touch of AI In order to improve the completeness of the app, we added detailed functions such as progress bars and improved search result display methods. AI suggestions were also very useful in this process. In particular, AI provided UI/UX improvement ideas that took user feedback into account, which helped us take the quality of the app to the next level. Conclusion: AI, the new companion for development Through this project, I have experienced first-hand how much AI tools can help throughout the development process. From writing code to debugging to improving UI/UX, AI’s advice and support have greatly improved the efficiency and quality of the project. In the future, AI will become an essential tool and companion for developers. This project was a great opportunity to see the future of development with AI. This innovative app that finds the exact moment you want in a YouTube video is a great example of the power of AI. I expect to see more innovative projects using AI in the future.

Key takeaways: • 🤝 Perplexity Announces Revolutionary Publisher Revenue Sharing Program • 🔍 Leading the paradigm shift in search with AI-based ‘answer engine’ • 💡 Increasing user curiosity is a key driver of business growth • 🏃 Compete with giants at high speed and focus • 📊 Rapid growth with over 250 million questions asked per month • 🌐 Providing publishers with customized AI search solutions through API sharing Innovative Combination of AI and Search: Perplexity's Challenge In an interview with Perplexity’s Chief Business Officer Dmitry Shevelenko, we get a glimpse into the future of AI search. More than just a search engine, Perplexity is positioning itself as an “answer engine,” with the goal of providing direct and accurate answers to users’ questions. Win-Win with Publishers: Innovative Business Models Perplexity recently announced a revenue sharing program with publishers. This is a different approach from traditional search engines, and it creates a structure that recognizes and appropriately compensates the value of publishers’ content. Through this program, Perplexity shares a portion of its advertising revenue with publishers, and this is designed to be a sustainable model through multi-year contracts. “Our success is impossible without the success of the journalism ecosystem,” Shevelenko stressed, making it clear that the program is not simply a defensive strategy, but a long-term, win-win strategy. User-Centered Innovation: Fueling Curiosity Perplexity’s secret to success is its focus on user curiosity. The way it accepts natural language questions rather than standard keyword searches has dramatically changed the way users ask questions. According to Shevelenko, the average question length on Perplexity is 10 words, compared to 2-3 words on traditional search engines. This approach encourages users to ask more complex and in-depth questions, resulting in richer information exchanges. Rapid Growth and Future Strategy: Speed and Focus In the face of fierce competition from giants, Perplexity is armed with speed and focus. Under the philosophy of “improving 1% every day,” they focus on continuous product improvement. This strategy is producing noticeable results. In the past month, they have asked more than 250 million questions, which is already more than the number of questions asked in the entire past year. Shevelenko expressed confidence that competitors' similar product launches "have the effect of increasing our traffic," and he defined Perplexity's core competencies as "speed and focus," which he said are helping to close the gap with the big boys. Technological Innovation: API Sharing and Customized Solutions Perplexity is sharing its API with publishers, giving them the opportunity to build their own AI search solutions, making it easy for publishers to implement customized AI search services based on their content. A particularly notable feature is the ‘Generate Related Questions’ feature. Perplexity reports that 40% of questions lead to follow-up questions, enabling publishers to increase reader retention and drive deeper content exploration. Ethical Considerations and Transparency Perplexity also considers ethical aspects of AI technology use, particularly when it comes to copyright and attribution, Shevelenko said, acknowledging the importance of proper citation of original sources, although he noted that “there is no monopoly on facts in a democratic society.” The company has recently improved its user interface to better cite sources and, in some cases, even mention individual reporters by name, showing a greater respect for journalistic practices. Conclusion: The Future of AI-Based Information Ecosystems

Did you know that Edge AI is on the rise? We've summarized the key points about how it differs from existing AI and what its advantages are! • What is Edge AI? AI processing close to the data source • Key benefits: real-time response, privacy protection, independent operation • Application areas: autonomous driving, financial transactions, healthcare monitoring, etc. Edge AI, what’s different? Edge AI is a technology that goes one step further than the existing cloud-centric AI paradigm. It is a method of performing AI computations at the edge of the network, i.e., the 'edge' where data is generated. Why is this important? Traditional cloud-based AI suffers from delays, bandwidth issues, and privacy concerns when transmitting data to a central server. Edge AI solves these problems while providing real-time decision-making, optimized data processing, and reduced dependency on network connections. Edge AI’s strengths are particularly evident in areas such as autonomous vehicles, real-time financial transactions, and remote medical monitoring. For applications that require fast response speeds in milliseconds and where data privacy is important, Edge AI can be the optimal solution. The core technology of Edge AI lies in the convergence of IoT (Internet of Things), edge computing, and AI. These three technologies meet to enable real-time data collection, processing, analysis, and decision-making. For example, in a smart home system, Edge AI enables immediate environmental control without communication with a cloud server. However, Edge AI implementation faces several challenges. Representative hardware constraints include limited computing power, memory, and energy. In addition, issues such as data quality management, model training, security and privacy assurance, scalability, and interoperability must be addressed. To overcome these challenges, various innovations are being made. Examples include the development of low-power, high-performance chips, lightweight AI model design, and distributed learning techniques. In particular, federated learning is attracting attention as a core technology of Edge AI. This technology can improve the performance of the overall AI model without sharing data from individual devices. The development of Edge AI will bring about great changes to our daily lives. In smart cities, traffic flow can be optimized in real time and energy use can be managed efficiently. In industrial sites, equipment failures can be predicted in advance and production lines can be automatically adjusted. In the medical field, wearable devices will enable continuous monitoring of patient conditions and immediate response. Edge AI is also important from an ethical perspective. Compared to centralized AI systems, it is advantageous for securing privacy and data sovereignty because data can be easily localized and decentralized. It also helps increase the transparency and explainability of AI decision-making processes. In the future, Edge AI technology is expected to become more advanced along with the development of 5G and 6G networks. Based on ultra-high-speed, ultra-low-latency networks, the performance and application range of Edge AI will greatly expand. In addition, it is expected that even more powerful edge computing capabilities will be secured through convergence with quantum computing technology. Edge AI will be a revolutionary paradigm that will change the future of AI beyond a simple technology trend. As centralized cloud AI and distributed edge AI develop mutually complementarily, a smarter, more efficient, and safer AI ecosystem will be created. We are now witnessing a new era of AI technology. Aren’t you excited about how the innovative changes brought about by Edge AI will change our lives and society? If you are interested in AI technology, pay attention to the development trends of Edge AI. It is definitely a key technology that will lead the future! https://media.licdn.com/dms/document/media/D4D1FAQGIUGpDxE2Cmg/feedshare-document-pdf-analyzed/0/1722053072186?e=1723075200&v=beta&t=B3Vmbf9787W50NaSQkr-p_wMfRaJY41L3cMUZjNDhP0

• Target audience: Developers worldwide, especially those interested in AI and integration • Topic: Developing AI-based apps and integrations for the Canva platform • Period: June 17, 2024 - August 12, 2024 (registration and submission) • Total Prize Pool: $50,000+ (Cash + Bonuses) • Core tools: Canva Apps SDK, Connect API • Key Opportunity: Exposure of your app to Canva’s 185 million monthly active users • Special offer: Invitation to Canva Extend 2024, direct meetings with the Canva team Hackathon Detailed Overview The Canva AI & Integrations Hackathon 2024 is a unique opportunity for creative developers to build innovative solutions using Canva’s powerful platform. The competition focuses on developing apps and tools that enhance the Canva user experience through AI technology and various integrations. Key benefits for participants: Gain skills: Get hands-on experience using Canva’s latest APIs and SDKs Networking: Connect with developers and Canva experts around the world Visibility: Potential opportunity to showcase your app to millions of Canva users Career Advancement: Outstanding achievements that can be added to your resume when you receive an award Business Opportunity: Potential monetization of the app you develop. Detailed schedule Registration Period: June 17, 2024 9:00 PM - August 12, 2024 5:00 PM (PDT) Submission Period: June 25, 2024 9:00 PM - August 12, 2024 5:00 PM (PDT) Review Period: August 14, 2024 2:00 PM - August 26, 2024 5:00 PM (PDT) Winner Announcement: August 30, 2024 at 2:00 PM (PDT) Eligibility and Requirements Be of legal age of majority in the participant's country of residence Participation is available to individuals, teams, or organizations Excluded countries: China, Hong Kong, Brazil, Quebec, Russia, Crimea, Cuba, Iran, North Korea, Syria, etc. Existing projects are also possible, but require significant updates during the hackathon. Submission Requirements Details Submitting your app/integration through the Canva Developers Portal

Current situation Model Collapse Risk : AI models trained with AI-generated content may degrade over time Loss of information about the actual data distribution As a result, it produces less diverse output that is biased and erroneous. Insufficient data : The Internet Is Overflowing with AI-Generated Content Lack of new human creation or natural data Solution for synthetic data : Mimic the statistical properties of real data Provide sufficient amount of data required for AI training Ensures inclusion of multiple data points Applications of synthetic data Healthcare: Analyzing patient trends, developing diagnostic tools Finance: Market trend forecasting, risk management Customer Service: AI-Based Support System Various Industries: Solving Model Collapse, Improving Data Privacy Challenges and Risks Data Quality : Ensuring accurate reflection of actual data characteristics

As AI technology fundamentally transforms the film industry, there is an exciting opportunity to be at the forefront of this revolution. The AI filmmaking competition called Project Odyssey is here to provide a platform for AI enthusiasts and filmmakers to explore the infinite possibilities of future filmmaking. AI and Creativity Competition Project Odyssey is an international AI filmmaking competition that will run from June 17 to July 15, 2024. The competition presents a new paradigm for filmmaking using AI technology, providing creators with the opportunity to create works that go beyond existing limitations. Five Challenge Areas The competition consists of five categories that demonstrate the various applications of AI technology: 3D Animation : Innovative character and background animation using cutting-edge AI video workflows Narrative : A place where AI-generated visual effects and human storytelling abilities meet Behind the Scenes : A Making-of Documentary Revealing the Hidden Secrets of AI Filmmaking Music Video : A Work Showcasing the Synergy of AI Music Generation Tools and Visual Technology Open Format : Free-form works that experiment with the infinite creativity of AI These categories provide the perfect stage for anyone interested in AI technologies to bring their ideas to life. Breakthrough prize money and industry recognition Project Odyssey has prepared over $28,000 in cash and credit prizes to encourage innovative attempts by participants. In addition to the winners in each category, there will also be a 'Community Choice Award' directly selected by the community and a 'Beginner's Award' for new creators. What is especially noteworthy is that leading companies in the AI industry, such as Civitai, ElevenLabs, and ThinkDiffusion, are providing additional prize money and subscriptions. This will be a valuable opportunity for the winners to receive industry attention and lead to future collaborations. Reviews from AI and movie experts The entries will be judged by a panel of leading AI technology and film industry experts, including Eric Solorio, Matt Wolfe, and Sebastian Kamph. This provides participants with more than just a cash prize. This opportunity to receive feedback from industry-leading experts will be an invaluable learning experience for anyone interested in AI filmmaking. Competition schedule to watch out for June 17: Competition begins and detailed rules are released. July 15: Deadline for submissions (11:59 p.m. PT) July 15-21: Heated community voting and finalist selection

Since ChatGPT surprised the world, AI technology has been rapidly changing our daily lives. In this wave of change, there is a new platform that is attracting attention. It is Poe. Poe, which has emerged as a next-generation AI platform along with Claude and Perplexity, what makes it special and what possibilities does it have? Poe is a platform that makes AI technology easy for anyone to use. It brings together various AI models and provides unique functions that allow users to create their own AI. This service developed by Quora is showing new possibilities for AI by making AI technology easier and more fun to use. Now, let’s take a closer look at what Poe is, what features it has, and what changes it can bring to our daily lives. Evolution of AI conversation platforms First, let’s take a quick look at some of the AI platforms that are currently in the spotlight. This will help us better understand what sets Poe apart. Claude: AI that values ethics Claude is an AI chatbot created by Anthropic. It aims to be a safe and correct AI, and values accuracy and honesty. It can understand various information such as text, images, and sounds, and handles long content well. It is characterized by being honest about what it does not know. Claude's key features include: Providing ethical and safe responses Capable of handling input in various formats Large context window of 200,000 tokens High accuracy and transparency Perplexity: A new look at smart search Perplexity is an AI-powered conversational search engine. It combines the strengths of a general search engine and an AI chatbot. It provides simple and accurate answers to questions and searches the Internet in real time to provide the latest information. Its main feature is that it uses multiple AI models to provide better results. The main features of Perplexity are: Real-time web search function Utilize various AI models (GPT-4, Claude, Mistral Large, etc.) Provide concise and relevant answers Available in free and pro versions Poe: A meeting place for AI models Unlike these, Poe is a platform that allows you to use multiple AI models in one place. You can meet various AIs such as GPT-3.5, GPT-4, Claude, and DALL-E in one place. In addition, there is a function that allows you to create your own AI bot. This is what makes Poe special. Quora, Why Did You Create Poe?

• Launch of ‘Previews’ feature that allows you to create and interact with web applications directly through AI chatbots in Poe • Compatible with coding-specific LLMs such as Claude 3.5 Sonnet, GPT-4, Gemini 1.5 Pro, etc. • Ability to create various web apps including games, interactive animations, and data visualizations • Enables all PoE users to create custom interactive experiences regardless of programming ability • Supports HTML, CSS, JavaScript, with additional formats planned for future support Poe has introduced a revolutionary new feature called ‘Previews’, which allows users to create and interact with web applications directly while talking to Poe’s AI chatbot. Previews works particularly well with large-scale language models (LLMs) that excel at coding, such as Claude 3.5 Sonnet, GPT-4, and Gemini 1.5 Pro. This new feature allows all Poe users, regardless of their programming skills, to create custom conversational experiences. Users can use Previews to create a wide range of web applications, including games, interactive animations, drag-and-drop interfaces, and data visualizations. Poe has provided some interesting sample projects to demonstrate the potential of this feature: Interactive presentation based on financial reports: https://poe.com/s/NlX2WRElDUvtuuMSFrZq Dynamic flashcards on specific topics: https://poe.com/s/HDAXQjBX9qmsInjjgV4X Interactive Drum Machine: https://poe.com/s/G4HshkCQCaXVfg0ndPZK Creating new shades through color clash: https://poe.com/s/neKbyqEJYUtiRjeC6b5f Web apps created with Previews can be shared with anyone via a dedicated link, and the results can be viewed in a new tab outside of the chat. Users can also take advantage of other Poe features, such as multi-bot chat, file upload, and video input, to help create custom web applications. Previews are currently available to all users of the web version, supporting CSS and JavaScript features along with HTML output. The Poe team says they plan to introduce support for additional formats in the coming weeks. We look forward to seeing what creative projects Poe users will create with this innovative feature. Poe hopes that users can explore the infinite possibilities and bring new ideas to life with Previews. Experience the features of Previews firsthand through the provided sample projects and create your own unique web application!

Anthropic has made significant progress in deciphering the inner workings of the large-scale language model Claude Sonnet, the first detailed look inside a modern, real-world large-scale language model. Research Methods and Key Findings The Anthropic research team used a technique called 'dictionary learning' to extract millions of 'features' from within the model, which show how the model represents various concepts. Broad concept representation: The extracted features represent a wide range of concepts, including cities, people, scientific fields, and programming constructs. For example, features have been discovered for San Francisco, Rosalind Franklin, immunology, and function calls. Multilingual and multimodal support: These features respond to text in multiple languages as well as images. For example, the Golden Gate Bridge feature responded to mentions and images in multiple languages, including English, Japanese, and Chinese. Representing abstract concepts: Features have also been found for more abstract concepts, such as bugs in computer code, gender bias in the workplace, and conversations about keeping secrets. Identifying relationships between concepts: By measuring the 'distance' between features, we were able to identify conceptual similarities. For example, features related to 'internal conflict' were found near features related to disconnection, conflicting loyalties, and logical contradictions. Feature manipulation: The research team found that they could manipulate these features artificially to change the model's responses. For example, when they amplified the "Golden Gate Bridge" feature, the model began to perceive itself as a bridge. Significance of the study Improving AI Safety: This discovery could contribute to making AI models safer. For example, it could be used to monitor risky behavior, guide them to desirable outcomes, or eliminate certain risky subjects. Discovered traits associated with bias and problematic behavior: The research team also discovered traits associated with sexism, racist claims, AI power seeking, manipulation, and secrecy, which could help address these issues in the future. Improved understanding of model behavior: By manipulating features, we were able to observe changes in the model's behavior, which helps us understand how the model's internal representations influence its actual behavior. Future tasks Discover more features: The features discovered so far are only a small part of all the concepts learned by the model. More features need to be discovered and analyzed. Solving the computational cost problem: With current technologies, the computational cost required to find all features greatly exceeds the model training cost. An efficient method is needed to solve this problem. Understand how features are used: Now that we have confirmed the existence of features, more research is needed to fully understand how the model uses them. Application to safety improvement: We need to develop ways to actually use the safety-relevant features discovered to improve AI safety. Anthropic expects this research to be an important milestone in further understanding AI models and improving their safety. The company plans to continue investing in interpretability research to contribute to the advancement of AI technology and ensuring safety.

Tips for getting clear and useful responses from ChatGPT: Write clear instructions (including long and short explanations). Provide other text for AI to reference. Request a “flow of thought”, which gives ChatGPT time to come up with the right answer. Providing ChatGPT with output from other tools such as RAG or code execution engines. Sample prompts for developers Find errors in code [Paste code] Complete in [language] Architecture and code writing for [project] Write a [language] command that can solve [problem] Writing documentation for [Project] that includes [Section] Sample prompts for writers My story is a plot with a new twist on [topic]. [result] Write a comprehensive online article on [topic] in [number of words] Generate conversation between two characters in [situation] Writing the beginning of the story about [Scenario] Create creative motivation for [character description] Sample prompts for marketers Provide blog post ideas on [topic] Write a product description for [product or service]

• Developed by the Paris Research Institute, France • Open source model supporting voice input-output • 70 expressions of emotions/styles • 8-person team, implemented in 6 months • Experience it first-hand with a web demo • Planned to be fully open source in the future Moshi is gaining attention as the first open-source model that can compete with GPT-4's speech function. Developed by a research institute in Paris, France, this AI supports both speech input and output and responds at an incredible speed. Key Features: Variety of expressions: 70 emotions and styles can be realized Rapid development: 8 researchers completed prototype in 6 months Low latency: Fast response times suitable for real-time conversations Scalability: Supports a variety of environments with 8-bit and 4-bit quantization Experience the Mosi: Mosi is currently available as a web demo at __T56349_____ . Key features include: Experimental conversational AI, does not guarantee accuracy of responses Conversation time is limited to 5 minutes Implementing the ability to think and speak at the same time Provides natural conversation flow with continuous voice input and output You can chat about a variety of topics, including pirate role-playing, how to make lasagna, and the movies you’ve seen recently. Provides optimized performance on Chrome browser Depending on your location, EU demo servers may provide better response speeds The development team will soon release the full codebase, including the 7B parameter model, audio codec, and optimized stack, along with a technical report. The emergence of this model heralds the democratization of AI voice technology. In the future, when local installation becomes possible, it is expected that high-performance AI voice assistants will be implemented on personal computers as well. Moshi's open source strategy follows the success story of Mistral AI. It is expected to accelerate the development of AI technology and encourage participation from a wider developer community, breaking away from the monopoly of large corporations. Moshi's development is expected to bring about a new wind of innovation in the field of AI voice technology.

• 86% of CEOs see AI as a key tool for revenue growth • Improving customer experience is the top priority for using AI • Productivity gains and cost reductions are the main motivations for AI adoption. • Data analytics and automation are key applications of AI • CEOs’ optimistic outlook on AI adoption is clear A new paradigm for business innovation Today, we are in the midst of a technological revolution. AI is no longer a possibility of the future, but a reality in progress. The results of Gartner’s 2024 CEO Survey clearly demonstrate this reality. It is noteworthy that 86% of CEOs recognize AI as a revenue driver. This is not just a simple introduction of technology, but a fundamental change in business models. The rise of customer-centric AI strategies It is very meaningful that the top priority of AI adoption is ‘improving customer experience’. This shows that companies recognize AI as a core element of creating customer value, not just a simple internal efficiency tool. This result proves that the ‘customer-centric’ philosophy that Steve Jobs always emphasized is still valid in the AI era. Balance between efficiency and innovation The fact that productivity gains and cost reductions are the main motivations for AI adoption shows that CEOs are acutely aware of the practical value of AI. However, the high ranking of ‘opening new products and markets’ reflects the expectation that AI will become a driving force for innovation beyond a simple efficiency tool. The era of data-driven decision making The fact that 'enhanced intelligence and analytics' is cited as a key area for AI utilization shows that companies are recognizing the importance of data-driven decision-making. AI will become a powerful tool for extracting meaningful insights from massive data. Optimistic outlook for AI adoption The fact that 86% of CEOs plan to use AI reflects strong optimism about AI. But it also means that AI adoption is becoming a necessity rather than an option. Companies that fail to effectively use AI risk falling behind the competition. Conclusion: Leadership in the AI Era AI is not just a technological innovation, it is redefining the nature of business. CEOs recognize the potential of AI, but the next challenge is to turn it into real business value. True leadership begins with thinking beyond the technical possibilities of AI and what value it can provide to customers and society. Success in the AI era will depend not on the technology itself, but on the innovative value we can create through it.

The beginning of innovation In 2017, eight researchers at Google published a paper titled “Attention Is All You Need.” This paper revolutionized the field of artificial intelligence and natural language processing. The core of the paper was a new mechanism called “self-attention,” which allowed AI systems to get closer to how humans understand language. The seed of an idea The project began with an idea from Jakob Uszkoreit, who came up with the concept of self-attention while working on improving a question answering system at Google. Uszkoreit and his colleagues wrote a design paper called "Transformers: Iterative Self-Attention and Processing for Various Tasks," which later became the basis for the Transformer model. The power of diversity The project team gradually expanded, with researchers from various nationalities and backgrounds joining. This diversity brought new perspectives and ideas to the project. In particular, the participation of Noam Shazeer was crucial. His outstanding implementation skills took the project to the next level, to the point where the team called it “magic.” Running towards the deadline The research team worked day and night in anticipation of the submission deadline for the major AI conference Neural Information Processing Systems (NeurIPS). They conducted experiments, analyzed results, and completed the paper until the last minute. In the process, they achieved results that greatly surpassed the previous machine translation performance records. Revolutionary achievements The transformer model has shown performance that surpasses the existing recurrent neural network (RNN) and long short-term memory (LSTM) models. In particular, it has achieved remarkable results in the field of machine translation, and the research team predicted that this technology can be applied to various fields such as images, audio, and video in addition to text. Industry Reaction The Transformer model has created quite a stir in the AI community. Interestingly, however, it has not led to any immediate change within Google. On the other hand, startups like OpenAI have quickly adopted the technology and developed innovative language models such as the Generative Pre-trained Transformer (GPT) series. Researchers' New Journey All eight authors of the paper eventually left Google to start their own AI startups or join other AI companies, a testament to the dynamism and pace of innovation in the AI field, as well as their confidence in the potential of Transformer technology. The Legacy of Transformers The paper "Attention Is All You Need" marked a major turning point in the history of AI. The transformer model became the basis for modern language models, leading to the birth of innovative AI systems such as ChatGPT, BERT, and GPT-3. This technology has greatly improved the performance of various natural language processing tasks such as machine translation, text generation, and question answering. Outlook for the future Transformer technology continues to evolve and expands beyond language processing to include computer vision, speech recognition, robotics, and more. Many experts predict that Transformers will continue to be a key driving force in AI development. This case study illustrates the birth of innovative ideas, the importance of diversity and collaboration, and the impact of technological advancement on industry and society. Through the process of how a small idea that started at a giant company like Google changed the landscape of AI technology around the world, we can get a glimpse into the essence of technological innovation.

• Key Summary: ChatGPT takes overwhelming first place in AI chatbot sector Wand is the most popular automation tool Midjourney is a leader in image creation New tools are making a splash in video, coding, and meetings In July 2024, the advancement of AI technology continues unabated. AI tools are driving innovation in various fields and are becoming the choice of users. Looking at this month's 'Top AI Tools' ranking, interesting trends are emerging in each category. AI Chatbot: ChatGPT’s Unique Position In the AI chatbot category, ChatGPT took an overwhelming first place with 134 votes. Google's Gemini (8 votes) and Llama3 (6 votes) followed, but the significant difference in votes confirms ChatGPT's unrivaled position. This shows that ChatGPT's excellent performance and user-friendly interface are receiving great responses from users. Automation Tools: Wand vs. Microsoft Copilot In the automation tool category, Wand took first place with 29 votes, and Microsoft Copilot came in second with 19 votes. This suggests that users are showing great interest in automation tools to handle tasks efficiently. Wand and Microsoft Copilot each have their own strengths, and seem to meet the diverse needs of users. Image Generation: The Dominance of Midjourney In the image generation category, Midjourney took first place with 16 votes. It was followed by Freeijk AI Image Generator (2 votes) and DIKA (1 vote) in second and third place respectively, but Midjourney had a clear lead. Midjourney seems to have won the hearts of users with its high-quality image generation capabilities and user-friendly interface. Video Tools: Runway's Propaganda In terms of video tools, Runway came in first with 9 votes , and CapCut came in second with 4 votes. Runway is highly regarded among professionals for its advanced video editing and creation features. CapCut seems to be popular among general users for its easy-to-use interface. Coding Tools: Replit and GitHub in Action In the coding tools category, Replit took first place with 3 votes , and GitHub took second place with 2 votes. Replit has the advantage of being accessible, offering a browser-based integrated development environment (IDE). GitHub continues to be popular among developers for its version control and collaboration features. Meeting Tools: Where Sembly AI Stands Out In meeting tools, Sembly AI took the undisputed first place with 16 votes . Sembly AI is gaining popularity among business users for its ability to automatically record and summarize meeting content. Spinach.io, which came in second, was a distant second with 1 vote. The Present and Future of AI Tools This ranking shows that AI technology is actively being used in various fields and is continuously developing to meet the needs of users. In particular, tools such as ChatGPT, Wand, and Midjourney are leading in each field, allowing us to gauge the future direction of AI technology development.

Research Summary Jobs Replaceable by AI Decrease by 21% After ChatGPT Launch Image-generating AI jobs to drop 17% after launch The higher the interest in AI, the greater the job loss The jobs that remain tend to be more complex and more expensive. New Jobs Asking for “ChatGPT Skills” The ability to collaborate with AI is expected to become important in the future job market. Research Background The online freelance market is undergoing a major transformation with the recent emergence of generative AI such as ChatGPT. Researchers from Harvard Business School and Imperial College London analyzed data from global freelance platforms to investigate the impact of AI on the labor market. Job Changes After AI Introduction According to our research, the number of jobs that are susceptible to AI replacement decreased by 21% in the eight months following ChatGPT’s launch . By occupation, the changes were as follows: Writing-related jobs: 30.37% decrease Software/App/Web Development: Down 20.62% Engineering: 10.42% decrease On the other hand, jobs that require a lot of manual work, such as data entry or video editing, have not changed much. The impact of image generation AI The impact of image generation AI was also significant. Graphic design and 3D modeling jobs decreased by 17.01% in the year following the launch of image generation AI such as DALL-E 2 and Midjourney . The details are as follows: Graphic Design: 18.47% decrease 3D Modeling: 15.52% reduction The correlation between AI interest and job losses

• Disclosure of analysis results on the use of generative AI • Technical support and troubleshooting were the most used at 23% • Content creation and editing followed at 22% • Used in various fields such as personal and work life support, learning, and creation • Acts as an important indicator of the future development direction of AI technology With the rapid development of artificial intelligence technology, generative AI is deeply penetrating our daily lives. Recently published analysis results provide interesting insights into how people are actually using generative AI. This data is expected to play an important role in understanding the current status of AI technology and predicting its future development direction. According to the analysis results, the areas of use of generative AI can be largely divided into six topics. The highest proportion was 'technical support and troubleshooting', which recorded 23%. This suggests that many users are using AI to solve everyday technical problems or receive help. The second highest proportion was 'content creation and editing', at 22%. This shows that AI is becoming an important tool in the creative process. AI is being used in various creative activities such as writing, editing, and idea generation. 'Support for personal and work life' ranked third at 17%, showing that AI is helping in various aspects of daily life beyond just being a technical tool. It was found that AI is being used for things like writing emails, preparing for meetings, and managing personal schedules. The 'Learning and Education' category took up 15%, showing that AI is being used for personalized learning, explanation of complex concepts, and language learning. The 'Creativity and Leisure' category took up 13%, showing that AI is also being used for artistic creation and leisure activities. Finally, the field of ‘Research, Analysis and Decision Making’ took up 10%, indicating that AI is being widely used in data analysis, decision support, and research activities. It gets even more interesting when you look at the specific use cases. AI is being used in a variety of fields, from practical uses such as generating ideas, editing text, writing emails, writing Excel formulas, and sampling data, to creative writing, coding, translation, and music composition. This shows that AI technology is affecting almost every area of our lives. These analysis results provide important implications for AI technology developers and companies. It will be possible to identify users’ real needs and develop functions and new services accordingly. It also raises the need to pay closer attention to AI ethics and safety issues. In conclusion, this analysis clearly shows how deeply generative AI has penetrated our lives. AI is now moving beyond being a simple technical tool and becoming a companion in our daily lives. As AI technology develops and becomes more widespread, it will be interesting to see how our lives, work, and creative activities will change. At the same time, social discussions on the responsible use and development of AI technology should continue. https://hbr.org/2024/03/how-people-are-really-using-genai

Claude AI systems have always been about working with people and meaningfully improving their workflows. As a step in this direction, Claude.ai Pro and Team users can now organize their conversations into projects. This allows them to collect curated knowledge and conversation activities in one place, and share the most useful conversations with their team. This new feature enables Claude to generate ideas, make more strategic decisions, and drive great results. Introducing project features Available to all Pro and Team customers on Claude.ai Powered by Claude 3.5 Sonnet, it has proven to perform well in various benchmarks. Each project provides a 200K context window (equivalent to about a 500-page book) Improve Claude's efficiency by adding relevant documentation, code, and insights Solving initial adaptation problems The Projects feature allows users to contribute their internal knowledge to Claude. This can include style guides, code bases, interview transcripts, past work, etc., so that they can receive expert help on a variety of tasks. Writing emails in a marketing team style Writing SQL queries at the data analyst level You can also define custom instructions for each project to further refine Claude’s responses, such as using a more formal tone or asking him to answer questions from a specific role or industry perspective. Improve collaboration with Claude The Artifacts feature allows you to work more efficiently with Claude. When you ask him to create code snippets, text documents, graphics, diagrams, website designs, etc., the artifacts appear in a dedicated window next to the conversation. Enhanced coding features, especially for developers, with a larger code window and live preview for the frontend. You can opt into the Artifact feature preview from the Account menu in Claude.ai. Providing inspiration through idea sharing Claude Team users can share their most useful conversations with Claude in the team’s shared project activity feed. The activity feed provides inspiration for each team member to see how they are working with Claude in different ways, and helps the entire team improve its AI capabilities. Sharing the work you create with Claude can help drive innovation in areas like product development and research. By pooling organizational knowledge across the company, you can create higher-quality results. Customer Case: North Highland At North Highland, a leading change and innovation consulting firm, hundreds of people across consulting, business development, and marketing teams use Claude to improve their work efficiency. From proposal writing to analyzing complex documents like 10-K reports, teams use Claude to improve the quality and scale of their professional services.

OpenAI has announced that it is delaying the launch of ChatGPT's advanced voice mode, a feature that was introduced in an update last spring and is expected to enable natural, real-time conversations with AI. Reason for change in release schedule The alpha version was originally scheduled to be released to some ChatGPT Plus users starting in late June. However, OpenAI decided to extend the development period by another month to improve quality. Major improvements currently in progress The development team is focusing on three key areas: Content Moderation : We are improving our models’ ability to detect and reject inappropriate content. User Experience : We are making the overall user experience more convenient and natural. System Scalability : We are preparing the infrastructure to enable real-time response even when a large number of users are using it simultaneously. Phased rollout plan OpenAI took a step-by-step approach: We'll start with an alpha test with a small number of users to gather initial feedback. Based on this, we plan to gradually expand the service. The ultimate goal is to have voice mode available to all Plus users by the fall. However, the exact schedule may be adjusted depending on whether OpenAI's rigorous safety and reliability standards are met. Other features in development In addition to voice mode, OpenAI is also working on new video and screen sharing features, which have also been previously demonstrated, and will provide separate information on their release schedules at a later date. Features of Advanced Voice Mode ChatGPT’s advanced voice mode is designed to understand and respond to emotions and nonverbal expressions, a breakthrough that will make conversations with AI more natural and human. OpenAI is introducing these new features cautiously, with the safety of users and the reliability of the system as top priorities. As development progresses, users can expect to experience more intuitive and interactive AI conversational experiences in the near future.

The impact of AI technology on the job market is becoming increasingly visible. With 30% of global workers concerned about their jobs being replaced by AI, the winds of change are already blowing in some occupations. The current impact of AI can be summarized as follows: Full-scale job replacement has not yet occurred AI is not yet ready to completely replace most jobs AI excels in certain task areas (e.g. customer support) In the short term, AI will replace specific tasks, not entire jobs. However, in some industries, the impact of AI is even more pronounced. For example, Klarna replaced 700 customer support staff with AI bots. In the freelance market, there are reports that repetitive writing and coding tasks have decreased by 30.4% and 20.6%, respectively, since the launch of ChatGPT. Changes are also being felt in terms of compensation. Routine, low-value work is being paid less, and some creative jobs are also being affected by AI. The case of film concept artist Reid Southen shows the reduction in workload due to AI tools. OpenAI CTO Mira Murati's remarks — "Some creative jobs may disappear, and maybe they should never have existed in the first place" — foreshadow the scale of the changes AI will bring and the resulting social controversy. In the era of AI, we must develop new job skills and prepare appropriate social countermeasures. Coexistence with AI is becoming an inevitable reality, and preparation and adaptation to it are more important than ever.

Watching the ChatGPT Edu customer demo, I was able to catch a glimpse of the future of education through the reactions of the participants. The sparkle of curiosity and anticipation in the eyes of the faculty, staff, and students made me realize the magnitude of the change that AI will bring to the field of education. A variety of creative ideas poured out during the demo. From the question, “Can we have this customized GPT act as a professor and write the lecture announcement for week 1?” to the request, “Can you summarize the lecture content for week 7 and highlight the key topics?” to the suggestion, “Can you create a study guide for this lecture from the student’s perspective?”, the various uses of ChatGPT were vividly revealed. These ideas show that AI has the potential to go beyond being a simple assistant tool and change the paradigm of education. A fundamental change in the way knowledge is delivered and acquired is unfolding before our eyes. For example, personalized learning experiences powered by AI can tailor education to each student’s learning pace and style. AI can help teachers reduce administrative tasks and focus more on teaching students, and can assess students’ understanding in real time and provide immediate feedback. Furthermore, AI can be a new tool to stimulate students’ creativity and critical thinking. Through conversations with AI, students can look at problems from different perspectives and explore their ideas in depth. This will greatly help them develop the ability to solve complex problems in the real world, beyond simple memorization. AI also has the potential to significantly improve access to education. It can provide new opportunities for learners who have been excluded from traditional education systems due to language barriers, learning disabilities, and geographic constraints. Multilingual classes with real-time translation capabilities, personalized learning content, and more are innovative ways to help every student reach their full potential. However, for these AI-based educational innovations to be successful, several important challenges must be addressed. First, educators need to be properly trained and supported to effectively utilize AI tools. Second, clear guidelines must be established for the ethical use of AI and data privacy protection. Third, we must always keep in mind that AI should be developed to complement, not replace, the role of human educators. Through the ChatGPT Edu demonstration, we could clearly feel that we are at the beginning of an educational revolution. The future of education will change greatly depending on how we utilize this technology. We must continue to seek ways to maximize the potential of AI while preserving the essential value of education through collaboration between educators, students, and technology developers. The changes in education brought about by AI are only just beginning. It will be exciting to see how tools like ChatGPT Edu will create a better, more equitable, and more effective education system. We are witnessing a pivotal moment in the opening of a new chapter in education. Source: https://www.linkedin.com/posts/amartyosen_i-just-got-off-a-demo-for-a-chatgpt-edu-customer-activity-7208899000593174528-JbOn?utm_source=share&utm_medium=member_desktop

Introduction This section explains the purpose and structure of the guide. It provides basic information on how to use AI tools to “supercharge” your business. Each section includes specific applications, best practices, and examples to help you quickly adopt the right AI strategy. The guide presents ways to use AI to achieve a variety of business goals, including improving productivity, developing new product ideas, and gaining a competitive edge. ChatGPT Business Prompt Ideas This section presents five business scenarios leveraging ChatGPT: A) Business idea: Role: Business Consultant Objective: Provide a concise yet fully feasible business idea. Method: Develop ideas tailored to client interests and market opportunities Focus: Briefly explain each concept and consider key factors such as cost, profit potential, competition, and risk. B) Business Strategy: Role: Business Strategist Objective: Assess the client's situation and develop a 3-5 year strategic priority plan. How to: Set actionable goals and suggest initiatives Focus: Strengthening competitiveness and ensuring long-term growth C) Marketing consulting: Role: Marketing Consultant Objective: Develop a targeted campaign approach Methodology: Evaluate client portfolios and provide actionable recommendation initiatives Focus: Providing strategic guidance within constraints D) Report Analysis: Role: Business Analyst Objective: Identify key opportunities for clients and suggest improvements

Data Science is an academic discipline that collects, processes, and analyzes data to derive meaningful information and insights. It is used to discover hidden patterns and relationships from massive amounts of data and predict the future. It can be said to be an interdisciplinary field of study that combines various fields such as statistics, computer science, machine learning, and domain knowledge. The goal of data science is to provide information to decision makers so that they can make data-driven decisions. To do this, data scientists process and analyze vast amounts of structured and unstructured data using various technologies and tools. Advanced analysis techniques such as data mining, machine learning, natural language processing, and text mining are used. In order to extract useful information from data, you need to understand the characteristics of the data well. Therefore, exploratory data analysis (EDA) that identifies the distribution, variability, and outliers of the data must be conducted first. Various visualization techniques such as histograms, box plots, and scatter plots are utilized. Before analyzing data, a data preprocessing process is required. This is the process of processing data into a form suitable for analysis by handling missing values, removing outliers, transforming variables, and creating derived variables. The preprocessing process requires a lot of domain knowledge. This is because you need to determine which variables are important and how to transform them. Statistics and machine learning are mainly used in data analysis. Depending on the type of data, various techniques such as regression analysis, classification analysis, cluster analysis, and association analysis are applied. Recently, advanced machine learning techniques such as deep learning are also widely used. Regression analysis is a method to identify linear relationships between independent and dependent variables. There are simple regression and multiple regression. Classification analysis is a method to predict categorical dependent variables, and representative examples include logistic regression, decision trees, and SVM. Cluster analysis is a technique to group objects with similar characteristics, and association analysis is an analysis that finds simultaneous purchase patterns between products in transaction data. To handle text data, natural language processing technology is required. There are methods such as Bag-of-Words and TF-IDF that quantify documents based on the frequency of word appearance, and Word2Vec and GloVe that convert words into embedding vectors. Recently, powerful deep learning language models such as BERT and GPT-3 have been in the spotlight. Computer vision technology is used for image/video data analysis. It is used in various fields such as image classification, object detection, face recognition, and autonomous driving. Deep learning algorithms such as CNN and R-CNN are showing excellent performance. Image generation using GAN is also receiving much attention. In data science, big data technology that handles large amounts of data is very important. Distributed processing using Hadoop and Spark is widely used, and recently, cloud-based big data platforms are also in the spotlight. Various technologies such as NoSQL, stream processing, and real-time analysis are being utilized. Python and R are representative programming languages used for data analysis. Python, which provides a vast library including Pandas, NumPy, and Matplotlib, is the most popular language in the field of data science. R, which specializes in statistics and visualization, is also widely used. Data scientists need to have not only programming skills, but also domain knowledge and communication skills. They need to be able to interpret data analysis results from a business perspective and clearly communicate them to decision makers. Another important skill is to be able to express complex information in an easily understandable way through visualization. Data science has become a key field that drives business innovation in companies. Data science is being used in various fields such as market prediction, customer segmentation, personalized marketing, anomaly detection, and recommendation systems. Data science has enabled us to capture new business opportunities and improve operational efficiency. Data science has now become a must-have capability in every industry. Data-driven decision-making is essential for companies to gain a competitive edge. Data scientists will need to evolve into talents with business insight and problem-solving skills, not just statistical or coding skills.

As artificial intelligence (AI) technology rapidly develops, the role and influence of AI in our society are growing. In particular, children are the generation that will grow up with AI technology, so it is important to understand how children perceive AI. In this study, we investigated children aged 5 to 12 years old on their perception of text-based AI models (ChatGPT) and visual AI models (DALL-E). The study found that most children perceived AI positively and expected AI to help them in their daily lives. Children viewed AI as friendly and associated AI with positive rather than negative characteristics. However, they did not think that AI had human-like emotions or physical senses. Interestingly, when children used AI models, they preferred visual AI models. In addition, while text-based AI models asked more questions about things that exist in the real world, visual AI models asked more questions about imaginary beings or objects. This suggests that visual AI models can stimulate children’s creativity and imagination. The research team hopes that the results of this study will help design AI tools for children, as visual AI models can be powerful tools that stimulate children’s curiosity and creativity. As AI technology advances in the future, it will be important to ensure that children can use AI in a positive and creative way.

Summation: Advances in artificial intelligence technology are leading to a surge in harmful content for children. The main sources are Meta's platforms (Facebook, Instagram, WhatsApp). The surge in generative AI-driven child abuse content (AIG-CSAM) is further exacerbating the problem. Preventing misuse of AI tools requires a joint effort by AI developers, platforms, governments, non-profits, law enforcement, and parents. The problem of harmful content related to children is not a new one, but the seriousness of the problem has increased recently due to the advancement of artificial intelligence technology. According to the National Center for Missing and Exploited Children (NCMEC), 36 million suspected cases and 100 million files were received in 2023 alone. In particular, 85% of these cases occurred on meta platforms such as Facebook, Instagram, and WhatsApp. This is the result of exploiting the high accessibility and wide user base of meta platforms. What is even more concerning is the rapid increase in child abuse content (AIG-CSAM) created using AI tools that anyone can easily use. Criminals are creating deepfakes by manipulating everyday photos of children or existing harmful content found on the Internet. In June of last year, the FBI warned that AI-generated sexual exposure blackmail cases were on the rise. The proliferation of AI-generated child abuse content makes it harder to distinguish between genuine abuse that harms real children. NCMEC has added a “generative AI” section to its reporting form, but many reports do not include this information, as it can be difficult to distinguish between AI-generated and genuine content. AI-generated abuse content remains illegal, and possession of it is also a crime. There are several steps that need to be taken to address this issue. First, AI developers should adopt more rigorous design practices to prevent their tools from being abused to create child-harming content. For example, they should remove harmful data from AI models’ training data and restrict AI models from creating child-harming content. In addition, developers should understand how AI models can be abused and conduct stress tests to prevent this. Second, platforms should invest more in digital fingerprinting hashing, machine learning algorithms, and AI artifact detection models. These are important for identifying existing child-harming content and detecting new AI artifacts. Platforms like Meta have introduced systems to detect and flag AI-generated content, but there are still many areas where improvement is needed. For example, Meta’s system is mainly focused on detecting benign content, making it difficult to find real child-harming content. Third, while the government is making efforts such as enacting the REPORT Act, it must provide sufficient funding to organizations such as NCMEC to handle the surge in reports. The REPORT Act requires reporting of all types of child-harming content, but the surge in AI creations is increasing the burden on NCMEC. To address this, the government must provide sufficient resources to organizations such as NCMEC so that they can respond effectively. Finally, parents should also communicate with their children about online risks and be cautious about sharing family photos. Parents should educate their children about online risks and take steps such as setting their children’s social media profiles to private. Parents should also set their own social media accounts to private and be cautious about posting photos of their children online. The problem of child harmful content is an urgent issue that we all need to work together to address. As AI technology rapidly advances, our response efforts must evolve accordingly. All stakeholders, including AI developers, platforms, governments, non-profits, law enforcement, and parents, must work together to address the problem of child harmful content. It is time for everyone to work together to maximize the benefits that AI technology can bring to our society while minimizing its negative consequences. #Child protection #Artificial intelligence #Generative AI #Child harmful content #Cybercrime Original text: https://www.fastcompany.com/91136311/were-unprepared-for-the-threat-genai-on-instagram-facebook-and-whatsapp-poses-to-kids

With the recent advancement of AI technology, ethical concerns continue to be raised. In particular, criticism of the behavior of leading AI companies is growing. At the center of it all is OpenAI. In June, 11 AI experts, including researchers from OpenAI, published an open letter called “Right to Warn.” They asked AI companies like OpenAI to voluntarily adhere to four principles. Prohibition of punishment and retaliation for risk-related criticism Establish procedures for raising risk-related concerns anonymously Encourage a culture of open criticism among employees and allow them to raise concerns while protecting trade secrets. No retaliation for disclosing public information when other reporting procedures fail They expressed concerns about the risks that AI technology could pose, such as exacerbating existing inequalities, spreading manipulation and misinformation, and losing human control. They also pointed out that AI companies are trying to avoid proper oversight due to profit motives. They further raised the issue that these companies are not transparent about the capabilities and limitations of their technology, the adequacy of safeguards, and the level of risk. This letter illustrates the controversy over transparency and ethics among AI companies, especially OpenAI. OpenAI has also been embroiled in several problems due to this incident. There have been a series of incidents, including the dismissal of the former CEO, a conflict with actress Scarlett Johansson, and the departure of the head of AI safety research. As AI technology advances, ethics and transparency issues are becoming more prominent. The “Right to Warn” principle requires AI companies to listen to these concerns. It seems that more efforts will be needed to establish AI ethics in the future. Original text: https://venturebeat.com/ai/more-openai-researchers-slam-company-on-safety-call-for-right-to-warn-to-avert-human-extinction/

With the recent popularity of conversational AI models such as ChatGPT, it has become important to write effective prompts to get the results users want. OpenAI suggests seven prompting strategies to get the most out of these conversational AI models. Write clear instructions It should be concise but contain the necessary details so that the model can accurately understand the user's requirements. It is helpful to use delimiters to clearly separate each part of the input, to describe specifically the steps the model should perform, and to specify the desired output length. Example: "Please write a summary in English of approximately 200 characters based on the following content: [original content]" Provide reference text Language models generate answers based on given text, so providing reference text can help the model provide more accurate answers, especially when asking about difficult topics or URLs. Example: "Read the article below and explain how climate change affects agriculture. [article URL]" Divide complex tasks into subtasks If you ask the model to process a complex task at once, it may show a high error rate. Therefore, it is better to divide the complex task into several simpler subtasks and synthesize the results of each subtask. Example: "Please write a recipe using the following steps: 1) List the ingredients needed, 2) Describe the cooking process in order, 3) Suggest ways to plate the finished dish." Give your model “time to think” AI models like ChatGPT also need time to think. If you give them about 17 to 28 seconds, the model can infer more accurate answers on its own. Example: "Please think deeply about the problem presented and suggest a possible solution after 30 seconds." Request a description of the model's inference process You can understand the reasoning process by asking the model to explain for itself how it reached its conclusions. You can use the internal dialogue technique, or you can start by asking for a summary of the user's query and then gradually build up to a longer conversation. Example: "Please analyze the given data and explain the analysis process and results in detail step by step." Using external tools To complement the model’s performance, you can leverage text search tools (RAG, search alignment generation, etc.). OpenAI’s Code Interpreter can also help your model perform math/code-related tasks more accurately. Example: "Run the Python code below and explain how it works and what the results are. [Python code]" Systematically testing changes When modifying prompts to improve model performance, you should define a comprehensive test suite and establish performance evaluation criteria. Keep in mind that while performance may improve for some examples, it may degrade for others. Example: "Please compare and analyze the performance of the old prompt and the new prompt using various examples." If you use ChatGPT with reference to the above 7 strategies and examples, you will be able to get more effective and accurate results. The key is to clearly convey the user's intention, maximize the model's strengths, and find ways to complement its limitations. I hope this article will be helpful to those who are having difficulty using ChatGPT.

ChatGPT is an artificial intelligence language model that can help you with a variety of SEO tasks. In this article, we will look at how ChatGPT can be used in six key SEO areas: keyword research, competitive analysis, on-page SEO, content optimization, technical SEO, and content creation. Keyword Research ChatGPT makes it easy to get keyword ideas related to a specific topic. For example, if you want to write about healthy lifestyles, you can ask ChatGPT: "Using 'healthy lifestyle' as a central keyword, please give me 15 long-form keywords that people can use when looking for healthy lifestyle habits." ChatGPT will then suggest relevant long-form keywords to help you get content ideas and understand what your users are actually searching for. Competitive Analysis Using GPT-4 with the Web Requests plugin can help you analyze competitor websites and find content gaps. For example, if you’re writing about social media marketing, you can ask ChatGPT: "My blog focuses on social media marketing. Please analyze [competitor website link] to find topics and subtopics they cover that I don't." This way, you can find out what topics your competitors are covering and fill in the content gaps by finding things they don't cover on their own blogs. On-page SEO ChatGPT is also useful for suggesting SEO-friendly URL structures and optimizing header tags. For example, if you are writing about the future of healthy living, you can ask ChatGPT: "I'm going to write an article about 'The Future of Healthy Living'. What would be the most SEO-friendly URL structure?" ChatGPT suggests SEO-friendly URL structures. You can also get advice on optimizing your header tags. Content Optimization ChatGPT can help you optimize your title tags and meta descriptions. For example, if you wrote an article about sustainable living, you can ask ChatGPT: "I wrote a blog post on 'How to Live a Sustainable Life in 2024'. My current title tag is 70 characters. Please suggest a SEO-friendly title tag that captures the essence of the post while maintaining the optimal length." ChatGPT generates SEO-optimized title tags and meta descriptions based on the information provided, allowing you to create content that is more visible to search engines. Technical SEO You can also get help with technical SEO tasks like generating an XML sitemap or configuring your robots.txt file. For example, you can request it like this: "I run a website with a blog, an online store, and a members-only section. How should I structure my robots.txt file to optimize indexing of these different types of content?" ChatGPT provides a step-by-step guide to configuring your robots.txt file to effectively manage crawling and indexing of your website. Content Creation ChatGPT can help you create SEO-friendly and engaging content. For example, if you’re writing about sustainable living, you can ask ChatGPT: "I'm writing a blog post about 'Sustainable Living 2024'. Please write an interesting, SEO-friendly introduction that includes the keywords 'Sustainable Living 2024'." ChatGPT suggests introductions that capture the attention of your readers while including the given keywords, so you can create high-quality, SEO-friendly content. ChatGPT and AI are very useful, but sometimes they can provide inaccurate information. Therefore, it is recommended to use the content generated by ChatGPT as reference material rather than using it as is, and always check the facts. In addition, if human review and editing are added, more complete content can be created.

Perplexity AI recently announced a group of industry-leading investors and advisors to accelerate the development of innovative AI solutions. All of these individuals have distinguished careers in technology and business and are playing a key role in driving Perplexity AI’s growth and development. Introducing Key Investors and Advisors Rich Miner , co-founder of Android, is a strategic advisor to Perplexity AI. Tobias Lütke , CEO of Shopify, brings his experience in driving innovative developments in e-commerce platforms. Susan Wojcicki , former CEO of YouTube, brings valuable insights from her experience running a video platform. Brad Gerstner , Founder and CEO of Altimeter Capital, brings his venture capital and investment experience to Perplexity AI’s financial growth. Soleio Soleio , former design lead at Facebook Messenger , serves as an advisor at Figma, providing key advice on design and user experience. Experts from academia and practice Pieter Abbeel , a professor at UC Berkeley and co-founder of Covariant, serves as a bridge between academia and AI research. Paul Buchheit , the founder of Gmail, is a technical advisor and a pioneer in email innovation. Oriol Vinyals , a research scientist at Google DeepMind, is at the forefront of AI research. Naval Ravikant , co-founder of AngelList , is an expert in the startup ecosystem and is supporting Perplexity AI's growth alongside Nat Friedman , co-founder of AI Grant. Leaders in technological innovation Former Uber CBO Emil Michael serves as strategic advisor, while Y Combinator CEO Garry Tan and Vercel CEO Guillermo Rauch share their extensive experience in startups and developer communities. Google AI SVP Jeff Dean and Meta’s Chief AI Scientist Yann LeCun are at the forefront of AI research and technological innovation. Contributions in various fields Jeff Bezos , founder and CEO of Amazon , is a leader in global e-commerce and technology innovation, and is illuminating the future of Perplexity AI. Mikhail Parakhin , former CTO of Yandex, and Amjad Masad , founder and CEO of Replit , are strengthening Perplexity AI’s technical capabilities with their diverse technical backgrounds. Andrej Karpathy , former AI director at Tesla, and Ashish Vaswani , co-founder of Isomorphic AI, are making significant contributions to the AI space. Balaji Srinivasan , former CTO at Coinbase, and Bob Muglia , former CEO at Snowflake , are contributing their data and cryptocurrency experience. Conclusion In this way, Perplexity AI is leading the innovation of AI technology with the support of investors and advisory boards with diverse backgrounds and experiences. With their rich experience and knowledge, Perplexity AI is expected to provide even more powerful technology and solutions.

You’ve used Perplexity to find answers, explore new topics, and expand your knowledge. Now it’s time to share what you’ve learned. Introducing Perplexity Pages, a new tool that lets you easily transform your research into visually stunning, comprehensive content. Whether you’re writing an in-depth article, a detailed report, or an informative guide, Pages streamlines the process so you can focus on what matters most: sharing your knowledge with the world. Smooth production Pages makes it easy to create, organize, and share information. Search for any topic and instantly get well-organized, beautifully formatted articles. Publish your work to a growing library of user-generated content and share it directly with your audience with one click. What's special about Perplexity Pages Customizable : Whether you're writing for a general audience or a subject matter expert, you can tailor the tone of your Page to resonate with your target audience. Adaptability : You can easily modify the structure of your article. Add, rearrange, or remove sections to best fit your material and engage your readers. Visuals : Take your articles to the next level with visuals you create in Pages, upload from your personal collection, or source online. Tools for everyone Pages is designed to help creators from all walks of life share their knowledge. Educators : Develop comprehensive learning guides for your students by breaking down complex topics into easily digestible content. Researchers : Make your research accessible to a wider audience by writing a detailed report of your findings. Hobbyists : Share your passion by creating engaging guides that inspire people to explore new interests. Original text: https://www.perplexity.ai/hub/blog/perplexity-pages?utm_medium=social&utm_campaign=pages-launch

An affordable product that allows universities to responsibly introduce AI to their campuses. Announcing ChatGPT Edu, a version of ChatGPT built for universities. ChatGPT Edu is designed to enable universities to responsibly deploy AI for their students, faculty, researchers, and campus operations. Powered by GPT-4o, ChatGPT Edu can reason from text and visuals, and use advanced tools like data analytics. This new offering includes enterprise-grade security and control features, and is affordable for educational institutions. We developed ChatGPT Edu after seeing success with ChatGPT Enterprise at universities like Oxford University, the Wharton School of the University of Pennsylvania, the University of Texas at Austin, Arizona State University, and Columbia University. ChatGPT can help with a variety of tasks across campus, including providing personalized tutoring to students, reviewing resumes, helping researchers write grant applications, and supporting faculty with grading and feedback. Our university partners have found innovative ways to make AI accessible to students, faculty, researchers, and campus operations. Some examples include: Columbia University professor Nabila Elbasel leads an initiative to integrate AI into community-based strategies to reduce overdose deaths. Her team has built GPT to analyze and synthesize large data sets to inform interventions, reducing weeks of research work to seconds. Undergraduate and MBA students taking Professor Eden Mollick’s class at Wharton School completed a final reflection assignment through discussions with a trained GPT using course materials, and reported that ChatGPT made them think more deeply about what they had learned. Assistant Professor Kristian Reves of Arizona State University is developing a personalized language buddy GPT that will allow students to engage in German conversations appropriate to their language level while receiving personalized feedback. This GPT will help students build their communication skills and save faculty time on assessments. Based on these use cases, we designed ChatGPT Edu as an accessible option for universities to introduce AI to their campuses at scale. ChatGPT Edu includes the following features: Access GPT-4o, the top-of-the-line model that excels in text interpretation, coding, and mathematics. Advanced features like data analysis, web browsing, and document summarization. Ability to build GPT (a customized version of ChatGPT) and share it within the university workspace Much higher message limits than the free version of ChatGPT Improved language features in terms of quality and speed, supporting over 50 languages Robust security, data privacy, and administrative control including group permissions, SSO, SCIM 1, and GPT management. Conversations and data are not used to train OpenAI models “Integrating OpenAI’s technology into our educational and operational frameworks is accelerating ASU’s transformation. We are collaborating across the community to leverage these tools and extending our learnings as a scalable model for other institutions.” —Kyle Bowen, Arizona State University Vice President Original text: https://openai.com/index/introducing-chatgpt-edu/

RTF Description: RTF is a framework that performs a task as a specific role and displays it in a specific format. Why use: Useful for expressing in a concrete form the tasks that need to be performed in a particular role. component: Role: Role to be performed Task: A task to be performed Format: The format in which the work will be displayed Example: Facebook Ads Marketer: Design a compelling Facebook ad campaign to promote a new fitness apparel line for a sports brand. Create a storyboard that outlines the sequence of creative for the ad, including the ad copy, visuals, and targeting strategy. SOLVE Description: SOLVE is a framework that defines Situation, Objective, Limitations, Vision, and Execution. Why Use: Useful for systematically analyzing situations and making action plans when managing projects or solving complex problems. component: Situation: Define the situation Objective: State the goal Limitations: Define limitations Vision: Define your vision Execution: Execution Plan Example: Manage a project to deliver a new software feature within a tight deadline with limited resources. Explain these constraints, provide a vision for the feature, and propose a step-by-step implementation plan. TAG Description: TAG is a framework for defining Tasks, Actions, and Goals. Why use: Useful for clarifying tasks, actions, and goals, such as when evaluating team member performance or setting goals. component: Task: Define a task Action: Specify the action Goal: Clarify your goals Example: This is a task to evaluate the performance of your team members. As a direct manager, evaluate the strengths and weaknesses of your team members. The goal is to increase the average user satisfaction score from 6.5 to 7.5 by the next quarter. DREAM Description: DREAM is a framework that defines the following steps: Define, Research, Execute, Analyze, and Measure. Why Use: Useful for systematic planning of research, implementation, analysis, and measurement processes for product development or problem solving. component: Define: Define the problem Research: Research Execute: Execute Analyse: Analysis Measure: Measurement Example: Define a product development problem, research potential solutions, run pilot projects, analyze the results, and measure the impact on product quality. BAB Description: BAB is a framework for defining Tasks, Actions, and Bridges. Why use it: Useful for describing a problem, specifying required actions, and providing a bridge to reach a goal. component: Task: Description of the task Action: Specify the resulting action Bridge: Elements that play the role of a bridge Example: We are not visible at all in SEO rankings. We want to be in the top 10 in our niche within 90 days. Create a detailed plan that lists all the steps you need to take and also include a list of your top 20 main keywords. PACT Description: PACT is a framework that defines Problem, Approach, Compromise, and Test. Why Use It: Useful for systematically planning your approach to solving customer engagement problems on digital platforms. component: Problem: Problem definition Approach: Defining the Approach Compromise: State a compromise Test: How to test Example: You have a problem with low customer engagement on your digital platform. Suggest an approach such as introducing interactive features, recognize the tradeoffs such as potential increased costs, and explain in detail how you will test the effectiveness of these features. CARE Description: CARE is a framework that defines Context, Action, Result, and Example. Why you might use it: Useful when launching a new product line or planning a marketing campaign. component: Context: Description of the situation Action: Description of action Result: Clarification of Results Example: Presenting an example Example: We are launching a sustainable clothing line. Can you help us create an eco-friendly campaign that highlights our environmental commitment? Our goal is to increase product awareness and sales. A similar successful example is Patagonia’s “Don’t Buy This Jacket” campaign.