The revolution brought about by artificial intelligence: Coatue’s perspective
Haebom
Three line summary
AI is not a bubble like crypto or metaverse, and can create real economic value. Right now, everyone is attracted to generative AI, so other things seem small, but remarkable technological progress is taking place.
Ultimately, it will be a battle of computing power and model methods. Ironically, this has a different grammar from the formula by which existing startups beat large capital.
Nevertheless, there is no disagreement that artificial intelligence is a major axis that will change the existing market order. Coatue positively views the technological acceleration in the AI field and will make aggressive investments.
The conclusion is the investment company's industry trend report...
Below is a 115-page report by Coatue, a US venture capital firm that operates approximately $70B in AUM. It is called a venture capital firm, but now boasts the credit and scale of a hedge fund.
In our time, we expect AI to become more accessible, scalable, and useful, allowing everyone to harness its power. As AI makes it easier for more people to program, debug, and deploy software, we expect the best coding languages to be natural languages (e.g., English, Korean, etc.). The power of AI is moving from the data center to mobile, potentially making anyone an AI user. Training AI with personal data sets could unlock new capabilities in everything from healthcare to retail. Finally, we expect research to continue to innovate AI models, making them more intelligent and capable.
The AI revolution is not without its challenges and drawbacks. In the short term, founders, developers, and investors will need to focus on several issues, including solving the illusion of AI model output, enabling multimodal use cases, reducing the cost of AI, and making AI deployment easier and more secure. In the long term, AI development is expected to become an increasingly engineering (rather than research) task, as AI scales to billions of users.
1.
AI has the potential to meaningfully improve the world, not just a bubble.
Despite more than $20 billion invested in private AI companies, widespread AI adoption remains in its infancy, but there’s no doubt about the enormous opportunity AI holds for improving the world.
Data on AI use cases shows promising signs: developers using GitHub Copilot complete tasks 50% faster, AI has reduced one company’s customer support costs by up to 95% (both through increased user satisfaction and faster response times), and a BCG study found that AI-enabled knowledge work improves the quality of knowledge by 40%.
A survey of about 600 corporate executives on their perception of AI capabilities also showed that 60% of executives plan to introduce AI to both new and existing products, proving that the opportunity for large-scale AI adoption is just around the corner.
The speed at which AI autonomy and usability are improving is particularly encouraging, especially compared to previous technological revolutions. AI is making groundbreaking progress and I believe it deserves attention and investment, and I anticipate that it will have an impact on the modern economy and our daily lives.
2.
Open source is key to AI, but not all open source is public. (Sounds like a pun, but it’s true...)
Open source research, data, and community are at the heart of the AI revolution. From the first Transformer paper to the latest LLaMA-2 release, open source development has enabled technologists to iterate, improve, and adapt AI to drive adoption and uptake.
This open source is not truly open, and is changing its approach from being all open to some. In the past 12 months, as companies have realized the value of content, they have increasingly limited the research they publish, kept their models private, and even refused to release their training data repositories.
Despite this, the open source AI community is vibrant and growing. As of November 2023, over 200,000 developers are contributing to AI projects on GitHub, over 300,000 models have been published on Hugging Face, and the AI-related Discord channel now has over 18 million members.
Big tech companies are committed to supporting an open ecosystem for AI, and see open source models taking hold alongside proprietary ones.
3.
Founders and investors need to understand the new AI-centric technology stack.
AI is changing the technology stack. To create value in the new wave of AI, it is important to understand the fundamental dynamics of each technology stack layer, how they interact, and how they differ from previous technology stacks.
While the AI model layer has received the most attention and investment to date, there are opportunities for innovation at every layer, from the underlying data centers that power AI to the end-user apps that make AI accessible to everyone.
AI model
We believe that the era of 'Intelligence as a Service' or IQaaS is dawning in the model, and the winner will be determined by the competition of talent, data, and computing power. Scaling model performance also means scaling high-quality data sets. We are seeing better performance when models are trained on 30% larger, well-curated datasets, which opens up opportunities for innovative engineering teams. Synthetic data and user feedback are also potential areas to explore.
Computing power and cloud platform
Hardware accelerators—primarily graphics processing units (GPUs)—are the foundation of AI performance. The insatiable demand for GPUs has had ripple effects across the economy, from stresses on the power grid to accelerated growth in the semiconductor supply chain to a reacceleration in demand for cloud services. As AI goes mainstream, inference workloads are expected to become significantly more compute-intensive. Serving 100 million GPT-4 users could require four times more compute per day than what it takes to train the model. As AI begins to be deployed in production environments, AI startups have the opportunity to innovate across the training and inference stack and drive dramatic cost savings.
Developer Tools
One of the biggest changes since the explosion of AI in 2022 has been the democratization of AI and the rise of AI engineers. The emergence of new AI operational tools has made it much easier to not only train, fine-tune, and deploy models, but also build useful applications on top of those models. The emergence of a vibrant AI operational tool layer has made AI accessible to approximately 30 million developers worldwide, creating the potential for explosive growth within the application layer.
AI Applications
The adoption of AI applications is only just beginning, and more use cases are expected to emerge. Consumer and enterprise interest in AI is already gaining momentum, from creative areas using tools like Runway and Tome to professional use cases like Reply and Aurora Solar. However, startups building AI applications will need to move quickly and create new behavioral changes to beat incumbents.
📚 Welcome to Haebom's archives. --- I post articles related to IT 💻, economy 💰, and humanities 🎭. If you are curious about my thoughts, perspectives or interests, please subscribe. Would you like to be notified when new articles are posted? 🔔 Yes, that means subscribe. haebom@kakao.com
