3709 Trousdale Parkway, Los Angeles, CA 90089

View map

Title: How much are LLMs narrowing the gap to human intelligence?

 

Abstract: GPT is paradigmatic of the “transformer" class of Large Language Models (LLMs). ChatGPT's simple interface allows users to type in queries and receive multi-paragraph responses. Its capacities have lead to predictions that jobs in education and elsewhere will soon be replaced by AI, and to even stronger claims that “AI has been solved” and that human-level intelligence or even superhuman-level intelligence is just around the corner. However, LLMs fail in interesting ways that expose real gaps between them and the kind of self-monitoring intelligence we aim to foster in our students. I will draw out some lessons from these failings, and use these lessons to rebut overly-optimistic takes on the capacity of LLMs to think, such as those promoted by David Chalmers in his 2023 APA Presidential Address.

Event Details

See Who Is Interested

0 people are interested in this event

User Activity

No recent activity