Why AI Is Incredibly Smart — and Shockingly Stupid Yejin Choi TED

Interesting. speech Yejin Choi talks about OpenAI’s LLM and GPT models on what they already know and where to dig next. The main message of the talk is that the size of the models gave an impetus to the fact that large language models began to denounce AGI outbreaks. (artificial general intelligence)They lack a common sense understanding. But this thing is like the dark matter of cosmology, where most of the matter is not visible, but it affects what we can see and perceive directly. In fact, the common sense has a similar position in the field of AI, and in the end the author says that language models!= knowledge models and her group are engaged in creating mechanisms for symbolic knowledge distillation, when they pass through which LLM models they generate “human-inspectable, symbolic, common-sense knowledge representation”, which can be studied, corrected and used to teach other models.

#AI #ML