Sequoia Capital, a legendary fund for investing in venture companies in Silicon Valley, USA, has revealed its view on the LLM market.
In the US, it is close to winter for investment, but they say they are open for LLM.
Here's what Sequoia Capital had to say:
Large-Scale Language Models (LLMs) are increasingly being adopted by enterprises across a range of industries.
There are companies that use LLMs in a variety of fields including code development, data science, chatbots, visual arts, marketing, sales, law, accounting, productivity, search, grocery shopping, consumer payments, travel planning, and more.
The new stack for LLM applications includes language model APIs, search mechanisms, and orchestration, with increasing open source use.
65% of enterprises are using LLM applications in production, and 94% are using a basic model API (OpenAIโs GPT is the most preferred).
88% believe that a search mechanism (e.g. a vector database) is essential to the stack.
38% are interested in LLM orchestration and application development frameworks.
Only a few companies are currently looking for tools to monitor LLM outcomes, costs or performance.
Some companies are exploring complementary generative technologies that combine text and speech.
15% of companies build custom language models in addition to using the LLM API.
Custom model learning is on the rise, and a variety of tools and frameworks are being used to do so.
Enterprises aim to customize language models to suit their unique context and data.
Custom methods include training a custom model from scratch, fine-tuning a base model, or using a pre-trained model by discovering relevant context.
The LLM API and the stack for training custom models are expected to be integrated over time.
The stack is becoming more developer-friendly with tools like LangChain for building LLM applications.
A full LLM implementation will require addressing issues of trust, data privacy, security and copyright.
Language model applications will become increasingly multi-modal, combining text, speech/audio, and image/video generation.
AI adoption is still in its early stages, with only 65% of companies surveyed having deployed LLM applications in production.
The infrastructure layer for LLM applications will continue to evolve rapidly over the next few years.
๐ Welcome to Haebom's archives. --- I post articles related to IT ๐ป, economy ๐ฐ, and humanities ๐ญ. If you are curious about my thoughts, perspectives or interests, please subscribe. Would you like to be notified when new articles are posted? ๐ Yes, that means subscribe. haebom@kakao.com
