Published Nov 16, 2023

“There’s a high chance we’ll be surprised again by AI” | Unconfuse Me with Bill Gates

Bill Gates and computer science professor Yejin Choi delve into the uncertain future of AI, exploring how its complex knowledge encoding, unpredictability, and limitations in reasoning shape the ongoing debate about its scalability and potential need for new architectures.

Episode Highlights

Topics covered

Popular Clips

AI's Limitations Revealed
Play Clip

Episode Highlights

Related Episodes

“We’re naive about what it takes to match human capacity” | Unconfuse Me with Bill Gates
Answers 383 questions
“I didn’t expect ChatGPT to get so good” | Unconfuse Me with Bill Gates
Answers 383 questions
“Nobody learns very well by reading random web data” | Unconfuse Me with Bill Gates
Answers 383 questions
“This might be the last hard thing I ever do” | Unconfuse Me with Bill Gates
Answers 383 questions
“This is the stupidest these models will ever be” | Unconfuse Me with Bill Gates
Answers 383 questions
“The very first invention is never the optimal solution” | Unconfuse Me with Bill Gates
Answers 383 questions
“It’s easy to think that language is a basket of words” | Unconfuse Me with Bill Gates
Answers 383 questions
“We have to find a way to let universities push things ” | Unconfuse Me with Bill Gates
Answers 383 questions
“The ultimate challenge might be a steak” | Unconfuse Me with Bill Gates
Answers 383 questions
“We need the equivalent of international weapons inspectors” | Unconfuse Me with Bill Gates
Answers 383 questions
“You start to realize your language doesn’t make sense” | Unconfuse Me with Bill Gates
Answers 383 questions
“We fail to recognize the amount of human progress made” | Unconfuse Me with Bill Gates
Answers 383 questions
“To be a good scientist, you have to leave a little room for chance” | Unconfuse Me with Bill Gates
Answers 383 questions
Episode 5: Yejin Choi
Answers 383 questions
Episode 6: Sam Altman
Answers 383 questions

Dexa/Unconfuse Me with Bill Gates

“There’s a high chance we’ll be surprised again by AI” | Unconfuse Me with Bill Gates

Topics covered

Popular Clips

AI's Limitations Revealed

Episode Highlights

AI Model Understanding

AI Limitations and Failures

Related Episodes

“We’re naive about what it takes to match human capacity” | Unconfuse Me with Bill Gates

“I didn’t expect ChatGPT to get so good” | Unconfuse Me with Bill Gates

“Nobody learns very well by reading random web data” | Unconfuse Me with Bill Gates

“This might be the last hard thing I ever do” | Unconfuse Me with Bill Gates

“This is the stupidest these models will ever be” | Unconfuse Me with Bill Gates

“The very first invention is never the optimal solution” | Unconfuse Me with Bill Gates

“It’s easy to think that language is a basket of words” | Unconfuse Me with Bill Gates

“We have to find a way to let universities push things ” | Unconfuse Me with Bill Gates

“The ultimate challenge might be a steak” | Unconfuse Me with Bill Gates

“We need the equivalent of international weapons inspectors” | Unconfuse Me with Bill Gates

“You start to realize your language doesn’t make sense” | Unconfuse Me with Bill Gates

“We fail to recognize the amount of human progress made” | Unconfuse Me with Bill Gates

“To be a good scientist, you have to leave a little room for chance” | Unconfuse Me with Bill Gates

Episode 5: Yejin Choi

Episode 6: Sam Altman

“There’s a high chance we’ll be surprised again by AI” | Unconfuse Me with Bill Gates

Topics covered

Popular Clips

Episode Highlights

AI Model UnderstandingYejin Choi, a leading computer science professor, joins Bill Gates to unravel the complexities of AI. They discuss the opaque nature of knowledge encoding, the art of prompt engineering, and the unpredictable future of scaling AI models.

AI Model Understanding

AI Limitations and FailuresYejin Choi and Bill Gates explore AI's surprising limitations in mathematical reasoning and its broader implications. They discuss the scientific community's divided perspectives on overcoming these challenges and the potential need for new AI architectures.

AI Limitations and Failures

Related Episodes