I’m Binge Watching Interviews with Sam Altman…
…so you don’t have to… and here are my takeaways. (Article 7)
This is a continuation from a series of articles found HERE.
2023-Lex Fridman Interviews Sam Altman (30:00 to 60:00)
Context:
The interview covered in this article was just released a few days ago. And it’s a doozy — over two hours. So I broke it up into 30 minute chunks, with one article per 30 minute chunk.
As always, I’ll pull out interesting tidbits below, adding in my personal commentary or relevant context sparingly.
Unless otherwise stated, all quotes belong to Sam Altman.
30:00 to 60:00
Sam talks about how amazing a tool ChatGPT is to help people, especially programmers, do their job.
“I think something the A.I. community does is — there’s a little bit of a sleight of hand sometimes when people talk about aligning A.I. to human preferences and values. There’s a hidden asterisk, which is that they’re really calling for: ‘Preferences and values that I approve of*.’”
“These systems can learn a lot, if we can agree on what it is we want them to learn.”
On aligning A.I.: “My dream scenario is that every person on earth will come together… and we will have a conversation on what we want to draw the boundary on this. And we would have something like a US Constitutional Convention.”
Sam goes on to say that ideally, nobody will get exactly what they wanted in the final A.I. ethics discussion. But everyone would get something that they feel good about.
“Most people want a model that has been RLHF’d to the worldview they subscribe to. This issue is really about regulating other people’s speech.”
Sam highlights that many people say that it doesn’t matter what shows up in their own (social media and news) feeds, because they “cannot be radicalized” (in their own mind). But they do worry about what shows up in your feed (suggesting that, unlike them, you can be radicalized).
Lex asks Sam if he and OpenAI ever feel pressure to not be transparent.
Sam answers quickly and decisively: “No.”
“We are happy to admit when we are wrong. We want to get better and better.”
“I don’t like the feeling of being scolded by a computer.”
Sam mentions an obscure story about one of the original Mac computers. He says that supposedly, Steve Jobs decided to include a handle on the back of it, because Steve thought you shouldn’t trust any technology you couldn’t immediately throw out the window.
Editor’s Note: Now that I think about it, it is pretty odd those computers had that big handle!
On defining what a “super-intelligent A.I.” needs to be: “We’re deep into the unknown here. But for me, a system that cannot go significantly add to the sum total of scientific knowledge that we have access to… is not a super-intelligence.”
Sam says that OpenAI is working toward designing a super-intelligent A.I. (by Sam’s definition).
Sam is non-committal to any timeline for when we will achieve an AGI>. But he says that if an oracle traveled back in time to tell us that GPT-10 fits this definition of a super-intelligent AGI, that he could believe that.
“I am excited about a world in which A.I. is an extension of human will.”
Sam says that there are not a lot of people using GPT, especially programmers, that say “get this out of here, turn this off.” He suggests early results from people who actually use it are great.
“I believe chess has never been more popular than it is right now.”
Sam talks about how even though there is a robot that can beat any human in chess, we still play. And in fact, we don’t watch different robots play each other. We usually watch a human play against A.I.
“The improvement in quality of life that A.I. can deliver is extraordinary.”
Sam somewhat gently mocks people who think that nobody is then going to work. He says that people want to feel useful, and that will still be possible in a future “unimaginably-good-standard-of-living” world.
Sam acknowledges that there is some chance that a super-intelligent A.I. cannot be aligned.
“The only way that I know how to solve a problem like this is to iterate our way through it, learning early, and limiting our one-shot-only to get it right type chances.”
“Maybe I’d have thought GPT would be like the tenth fastest growing product in history, but not the number one.”
“Someone said to me over the weekend: ‘You shipped an AGI, and I just went about my weekend as if nothing happened.’”
My Biggest Takeaway from this Interview
This first part of this segment was heavy.
Sam acknowledges much of similar frustrations to what I have voiced to friends and loved ones regarding social media and social justice these past years. A lot of folks spend a lot of time worrying about behaviors in others — but miss that they practice the very behaviors they are accusing others of practicing. Often, it is actually reversed: The ones doing the accusing are creating a false narrative in their own heads, and the story that they have created has led them to adopt these (bad) behaviors that others were not really doing in the first place.
And wow, what a throwback, to think about that neon colored Macs (with the handles). Come to think of it, my middle school computer lab didn’t have windows! So even though the computers have handles, I could not have lived out Steve Jobs’ mission of tossing it out a window if it betrayed me!
More to come!
Articles in Series:
- https://medium.com/@dwolin/im-binge-watching-interviews-with-sam-altman-29a1f9f07ee1
- https://medium.com/@dwolin/im-binge-watching-interviews-with-sam-altman-559bea849356
- https://medium.com/@dwolin/im-binge-watching-interviews-with-sam-altman-44638f1e4eff
- https://medium.com/@dwolin/im-binge-watching-interviews-with-sam-altman-e1d8ac81ca43
- https://medium.com/@dwolin/im-binge-watching-interviews-with-sam-altman-588981e6eb2b
- https://medium.com/@dwolin/im-binge-watching-interviews-with-sam-altman-5ebfe3c79f7e
- ← You are here
- https://medium.com/@dwolin/im-binge-watching-interviews-with-sam-altman-cb504390ab8c
