Foundation Capital reposted this
Enterprise VC-engineer-company builder. Early investor in @databricks, @tubi and 6 other unicorns - @cohesity, @eightfold, @turing, @anyscale, @alation, @amperity, | GP@Foundation Capital
What’s next after LLMs? Every CEO, whether they make computer chips or potato chips, has announced an AI strategy. But deployment of LLM-based applications remains nascent, and LLMs have plenty of shortcomings. At Foundation Capital, we think 3 innovations will be huge for building in AI’s next era. 1. Multimodal models are moving beyond text. Inputs / outputs as images, audio, and video opens up a world of new use cases. Customer service chatbots get faster + more accurate voice agents. Multimodal unlocks healthcare, where there’s a mixture of data types. GPTs get deployed for cancer treatment 2. Multi-agent systems transform automating complex tasks. Hands down, autonomous agents calling multiple interacting systems is one of the biggest breakthroughs since ChatGPT. Multi-agents transform AI from passive tool to active player 3. New model architectures will address some limitations of transformers. State-space models (Cartesia), large graphical models (Ikigai), RWKV. Post-transformer architectures have the potential to be equally (or more) performant than LLMs, especially for specialized tasks, while also being less computationally intensive, exhibiting lower latency, being easier to control Founders may feel they’re building on quicksand, with every layer of the AI stack moving fast. But the flywheel also presents a once-in-a-generation chance to build magic. ✨ Latest blog in the comments 👇
I'm convinced that Jim's Flywheel was advanced from the "abcd decision making model", which was borrowed from the Krebs Cycle (which likely also had many predecessors). In speaking with a fellow from Daniel Kahneman and Amos Tversky's Cohort. He hinted at the true intention/purpose of "Thinking Fast & Slow" wasn't to get folks to know more about their system 1 and system 2 thinking processes. It was a love letter from psychologists to economists. Penned to bridge the gap between the sciences by acknowledging the faults of firm biases. Asimov predicted this in Foundations with "Psychohistory". LLMs paired with your insights will continue on their path to bridge that gap. As first principles/mvp/assumptions of market are as useless as "judging a book by it's cover"/first impressions. Meaning that through data and sociology/psychology. We will be able to find truer context of what actually is. Rather than what we exclaim that it might be. Excellent thoughts and blog post.
Existing state of the art also has issues around foundational embeddings ( responsible for understanding our data ), this is where companies like subtl.ai come in
Very succinctly stated. Multi-modal, multi-agent, new architectures are definitely the next in many steps required to build successful system.
Successful architectures are generational (von neumann - fetch execute-Relational DB) . Enhancements may be but not new architectures in a short span
All this magic requires the right data which is often proprietary.
👍 for point #3. Post-transformer architectures/models. So why did Google allow microsoft to make GPT, even though it invented the Transformer?
Enterprise VC-engineer-company builder. Early investor in @databricks, @tubi and 6 other unicorns - @cohesity, @eightfold, @turing, @anyscale, @alation, @amperity, | GP@Foundation Capital
1mohttps://foundationcapital.com/beyond-llms-building-magic/