#387 – George Hotz: Tiny Corp, Twitter, AI Safety, Self-Driving, GPT, AGI & God
Lex Fridman Podcast
Future Language Models with Memory Systems and Source Citing Capabilities
Future language models (LLMs) are expected to have memory systems and retrieval capabilities to mimic human brains. By incorporating a memory system, LLMs can store and access information similar to how the human hippocampus functions. These future LLMs will be smaller in size but will run in a loop and have the ability to retrieve stored information. Furthermore, the inclusion of a retrieval system in LLMs enables explicit source citation, allowing humans to investigate and verify the information provided. This integration of human supervision and the ability to check sources addresses issues such as hallucination and pushes LLMs towards a higher level of performance. Once an LLM capable of citing sources is developed, it could potentially replace search engines, as it fulfills the desired functionality of a search engine.
Remember Everything You Learn from Podcasts
