» » »

Toward models of human cognition that achieve human like abilities

Jay McClelland

In the past, models of human cognition attempted to capture principles of human cognition but they could not actually achieve human like levels of performance in a wide range of domains. For example, they could not recognize objects in images, learn to beat a good Chess player, or translate from one language into another. In contrast, recent advances in AI have created machines that often exceed human abilities, but they do not do so in human like ways. These systems also have some important weaknesses compared to our human cognitive abilities. In this talk, I will review some of the strengths and weaknesses of today’s AI systems and sketch an approach toward building models that might someday reach human-level abilities. There are several challenges and open questions that face this approach. In will raise some of these challenges and open questions, and suggest possible ways to approach them.

Speaker: Jay McClelland, Stanford University

See weblink for building admission information

Monday, 01/27/25

Contact:

Website: Click to Visit

Cost:

Free

Save this Event:

iCalendar
Google Calendar
Yahoo! Calendar
Windows Live Calendar

Stanford Symbolic Systems Forum

Margaret Jacks Hall
460-126
Stanford, CA 94305

Website: Click to Visit