OpenAI’s “breakthrough”

Make Me Smart November 27, 2023 Transcript

Note: Marketplace podcasts are meant to be heard, with emphasis, tone and audio elements a transcript can’t capture. Transcripts are generated using a combination of automated software and human transcribers, and may contain errors. Please check the corresponding audio before quoting it.

Kai Ryssdal 

Charlton Thorp you know what to do. He does. He does. I say something and then he does it. For a long time me and Charlton Thorp, Hey, everybody, I’m Kai Ryssdal. Welcome back to Make Me Smart, where we make today make sense.

Matt Levin 

And I’m Matt Levin filling in for the great Kimberly Adams. Thanks, everyone for joining us on this Monday, November 27. Hope everybody had a wonderful Thanksgiving. How was your Thanksgiving Kai?

Kai Ryssdal 

Mine was good. The house was full. My mom was here. All the kids were back. We had 14 for Thanksgiving. It was nice. It was nice. What about you?

Matt Levin 

It was all right. We have two sets of divorced parents, my my wife and I. So it’s a lot of like driving and then a Thanksgiving meal and then driving and then another Thanksgiving meal. And so I’ve seen enough turkey for quite some time. But it was good.

Kai Ryssdal 

Good, good. Good. All right. Well, we’re gonna do what we usually do some news, some smiles and then we’ll get you on your way. You got to go first. You got like a whole raft of things here.

Matt Levin 

Okay, yeah, I always put way too many notes in this document. Because I’m not a pro like, oh, no, not a pro like you Kai.

Kai Ryssdal 

You know, my rule about being on the radio, have I told you my role about being on the radio?

Matt Levin 

No, what’s your rule?

Kai Ryssdal 

If you think about how you sound while you’re on the air, you’re gonna screw up how you sound on the air. So don’t think.

Matt Levin 

That’s true. That’s true. That’s true.

Kai Ryssdal 

It makes my producers crazy, but don’t think.

Matt Levin 

It’s very good advice. I wish I knew this before I began a career in radio. Let’s start here, Kai in your Thanksgiving, did AI come up at all?

Kai Ryssdal 

Yes, repeatedly, repeatedly.

Matt Levin 

I feel like that was the case set at multiple Thanksgivings across the country. What was the conversation?

Kai Ryssdal 

Well, it was it was my oldest son is very much it’s gonna change the way this entire economy works. And it’s gonna be relatively good and another kid was I don’t know. And then you know, the various in laws chimed in and cousins and then it went from there.

Matt Levin 

Well, I wasn’t really allowed to talk about AI because my wife is so sick of me talking about it, because I report a lot about it, as you know, on Marketplace. So my, my news story is a follow up to the OpenAI melodrama that honestly, I think people were so sick of the melodrama. And then the news I’m about to share broke late Wednesday that I don’t think this has registered to the degree that it possibly should have.

Kai Ryssdal 

I think that’s exactly right. I think that’s exactly right.

Matt Levin 

Okay, I’m going to take you back Kai to that story I did, where I created images of you using AI as an astronaut eating a sandwich. Do you remember that?

Kai Ryssdal 

I do. I do. I don’t think it looked much like me, but whatever.

Matt Levin 

I think it looked decently like you for that stage of AI. This was pre ChatGPT. And the reason I’m bringing this up is I had a conversation with someone who didn’t make the story and who only wanted to talk on background, who said there are people in AI who want to quote unquote, build God. And I went, okay, not quite sure what to make of this, especially pre ChatGPT you know, this kind of science fiction infused AI hype. It’s kind of like go, alright, not sure how to react. Another way of saying that is creating artificial general intelligence, which has varying definitions, but the workable definition is AI that can pretty much do any economically valuable task a human can do but better, right, which is a pretty crazy thing to wrap your head around. Okay. Fast forward to the OpenAI drama. Sam Altman is out. Sam Altman is then back in by the Wednesday before Thanksgiving, and then late on Wednesday, when everyone’s ready to not think about AI anymore and eat turkey, Reuters, and I believe the information also broke that the since departed board of OpenAI had received a letter from multiple staff letter. staff members of OpenAI, warning them about a significant breakthrough that OpenAI had made. And the breakthrough was something known as Q*. The breakthrough was with Q*, it could do math, it could do some elementary school level math. Okay, so why is that a big deal? Anybody can do elementary school level math. ChatGPT famously can’t do math very well. Other AI products built on large language models can’t really do that. And it is a limitation of that form of AI. And because of that there is a limitation to how much AI can actually reason or deduce or you know, kind of to make logical inferences from right, if you can get it to, if you can teach it a mathematical concept, and then get it to apply those concepts to mathematical problems. That is possibly a road to artificial general intelligence, which is Sam Altman’s Holy Grail, right. He wants to use this stuff to make advances in theoretical physics and make advances in you know, fields metal math type stuff, right? I am, you know, there is debate back and forth with very, very smart AI people about how big a deal this is. Some people are like this is so overhyped other people like yeah, this is a big deal. What I find interesting is that open AI has not confirmed this. We have no the only reason we’re learning about this is because of Reuters and the information in the leaks that came because of this Altman controversy. And so the transparency here, that was supposed to be a fundamental value in the launch of OpenAI seems somewhat lacking. And I think there’s an interesting question as to well, like, if they have made some significant breakthrough here on the road to general artificial intelligence, or I’m sorry, artificial general intelligence, does the public deserve to know, do policymakers deserve to know? And I think that the lack of clarity around this is, I don’t know disconcerning? In some, again, vaguely dystopian way.

Kai Ryssdal 

I will, I would submit that it’s more than vaguely dystopian. But let me ask you a couple of questions. Since you are the subject matter expert on this at Marketplace, scale well at Marketplace you are. Scale of 1 through 10, how worried are you?

Matt Levin 

About what exactly?

Kai Ryssdal 

About the, let’s go all the way about the possibility that artificial general intelligence could wipe out humanity?

Matt Levin 

Not, me personally, not that worried.

Kai Ryssdal 

Okay. Yeah. How worried are you? Okay, how worried are you about the new board of directors at OpenAi? And, and their, their sensitivity to the subject matter that they are now entrusted with managing and overseeing?

Matt Levin 

Yeah, well, I heard you and Kimberly talking about this. And your, your boy, Larry Summers. Yeah, that’s right. So I mean, that. The truth is here is that the originally the OpenAI board was a nonprofit board, I guess, technically, it is still a nonprofit board. And what the Altman controversy very clearly showed was that Microsoft and these other investors really have the power. And that commercialization will take precedent on some level over safety concerns. Right. I don’t, you know, I do think Altman like legitimately values safety. Like, I want to make that clear. I don’t think they’re just they have a whole host of headaches in front of them if they don’t take safety seriously. But in terms of priorities, like yes, I think like the new constitution of the board makes it very, or whatever that will end up being Microsoft will, I’m sure have a hand in that too. It will. It’ll be like the board of any commercial entity. Right. Except they’re developing possibly a technology that could wipe us off the face of the earth, apparently.

Kai Ryssdal 

So let’s get the promo in here. Right. I think it’s running tomorrow. Right? Your piece on governance.

Matt Levin 

Yeah. Yeah, it is. Yeah. On what a board of directors actually does,

Kai Ryssdal 

What it does, and we’ll you know, sort of frame it in this context about, you know, some very serious challenges that this board now has, I think, yeah. It’s a good story. And, and Matt’s been covering the hell out of it for us. So you should definitely keep following him. That’s all I got to say. Okay, here’s mine. It’s a quickie, you ready?

Matt Levin

Yep.

Kai Ryssdal

All right. Elon Musk went to Israel today, and took a tour of some of the places that were attacked by Hamas on October the seventh. And I’m not sure whether this is genuine repentance, or somehow greenwashing his anti-semitic tendencies, I just can’t really decide. And I just, I’ve just wanted to make sure that people knew that this had happened. This is the most powerful and richest man in the world, a co-founder by the way of OpenAI, right. Also, he’s talking about Starlink. And using that to aid people in Gaza. Apparently there’s an agreement with the Israeli government now that he’s not going to turn Starlink on for Israel and Gaza. It’s just a mess. Anyway, I don’t know what I think about this other than to say, huh, this is very interesting. And I’m just I’m deeply troubled, well as I think listeners of this podcast are aware of Musk’s continued involvement in geopolitics, because I’m not sure he’s the right guy for that. It’s all I got.

Matt Levin 

Just kind of flipping the perspective of this. I find it interesting that especially recently, we’ve seen these tech magnates, Musk, Altman did a world tour here. Like this is like foreign dignitary type stuff. They’re like, why he’s with Netanyahu. It’s, it’s crazy.

Kai Ryssdal 

Yes, I totally agree. I totally agree. Look, look, it’s not like geopolitics. And grand strategy are, like hard, right? I mean, these are difficult problems. And you require people with lots of experience to get them done. And it can be very, very delicate. Elon Musk is not an elegant person. And the and I just yeah, and it’s not great. It’s not great that he’s got all this influence on on geopolitics. It’s just not great. Yeah.

Matt Levin 

Also somewhat ironic that I mean, he got in trouble on his own platform. Oh, and that that platform is not good at civil discourse around very sensitive issues, right. Like it’s even Twitter before X was not a good place for this. So it’s kind of hoisted on your own petard type thing?

Kai Ryssdal 

Absolutely. Absolutely. What do you got man?

Matt Levin 

Merriam Webster. This is make us smile. Merriam Webster came out with its word of the year this year. And the word of the year. I liked it. It was “authentic.” I’m not sure exactly,

Kai Ryssdal 

Really? I did not know that. That’s something I would know in the course of my day, I would have seen that anyway.

Matt Levin 

I was yeah, this actually didn’t make as many headlines as I thought it might. But anyway, their little blog spiel about it, a high volume lookup most years, I think they’re referring to how many people look it up on their website, “authentic” saw a substantial increase in 2023 driven by stories and conversations about AI celebrity culture, identity, and social media. It seems very very zeitgeist at that word.

Kai Ryssdal 

Very, very. Hmm.

Matt Levin 

Oh, yeah. Literally searching for authenticity.

Kai Ryssdal 

Oh, snap. That’s very good. That’s very good. All right. Mine is is very less much less meta. Much more frivolous. But I think it is a sign of of late-stage capitalism going completely your I saw this in The Washington Post today that Doritos the chip people have made, I can’t believe I’m actually saying this. They have made a crunch-cancellation software that removes the sound of chewing from voice chat Zoom, or any call that uses headphones designed specifically for gamers who apparently eat chips a lot when they are gaming. It’s basically noise cancellation on the input end, and Okay. Okay. I mean, if that’s what we need to spend our time and money on. Okay, there we go.

Matt Levin 

This seems like a great use of technology. I’m all for this. I’m glad they were working on this. That’s great.

Kai Ryssdal 

Oh man, I’m not I’m not a Doritos guy or a chip guy. But yeah, I could see that. Oh, man. Oh, man. Alright, we’re going out on that. We’re going on the search for authenticity and crunch cancellation software. We are coming up that tomorrow is Giving Tuesday. By the way I’m sure you haven’t heard that yet. Because that’s not anywhere. We have a huge, ginormous challenge on the table. Because that’s what our corporate bosses said we’re going to do $100,000. We want to raise $100,000 on Giving Tuesday, $100,000 in 24 hours. If that happens well like another $100,000 from the Investors Challenge Fund if you’re listening Monday today, give early, it counts somehow in  bookkeeping. Help us get a head start unlocking this challenge funds marketplace.org/givesmart or follow the link in the show notes. $100,000 in 24 hours or longer actually, if you start today, because it counts. It counts. Make Me Smart is produced by Courtney Bergsieker. It’s a lot. Today’s program was engineered by Charlton Thorp. Ellen Rolfes writes our newsletter. Our intern is Niloufar Shahbandi.

Matt Levin 

Marisa Cabrera is our senior producer. Bridget Bodnar is the director of podcasts and Francesca Levy is the executive director of digital. Still got leftovers Kai or are you out?

Kai Ryssdal 

No, we got leftovers. I think I have to make some turkey chili actually tonight.

Matt Levin 

Ooh nice. Yeah. Underrated leftover option.

Kai Ryssdal 

Yeah.