Professor Richard Sutton is the one who academically established reinforcement learning, which is the basis of modern AI.
In 2019, John Kemeny and Richard Sutton were researching AGI. At that time, there was no ChatGPT like today,
and they researched in a different way.
This time, Richard Sutton founded the OAK research institute after resigning.
July 14th article
https://officechai.com/ai/father-of-reinforcement-learning-richard-sutton-starts-oak-lab-to-work-on-new-approaches-to-ai/
Sutton's Long-Standing Criticism of LLMs
Sutton has spent a lot of time over the past year articulating the core arguments of OAK lab.
In podcasts, he argued that large language models (LLMs) cannot set their own goals without receiving feedback from the real world,
and therefore, LLMs and all models built on top of them cannot truly be considered intelligent.
Introducing the OAK Architecture Beyond LLMs
Our algorithms learn in real time without storing or replaying data.
Combining these algorithms with event-based neural networks allows us to implement learning systems that reduce computing resources and
energy consumption by several orders of magnitude compared to conventional methods.
High computational efficiency enables continuous real-time learning.
Our goal is to discover and implement algorithms that enable agents to achieve goals in large environments.
Our focus is on enabling learning through direct experience rather than relying on curated datasets.
It's still in the research stage, but will smarter intelligence find its way to us even with today's data centers?
LLMs use a lot of memory and GPUs...
