Happy Birthday ChatGPT! 12 months on, what’s changed?

Sascha: [00:00:02] Welcome to the Dive, the podcast that asks who said business news needs to be all business. I'm your host, Sascha Kelly. Happy birthday, ChatGPT. Since arriving in our lives this time last year, it's made news headlines had teachers, industries and governments pretty worried. And on a personal note, it's made many of my menial tasks from writing shopping lists and doing budgets a lot easier. It's Friday, the 1st of December, and today I want to know, well, what exactly has the arrival of ChatGPT changed and how has it influenced the conversations that are happening in businesses around us? To dive into this topic today, I'm joined once again by Dan Van Voom from Capital Brief, who's been writing about this very topic. Dan, welcome back to the dive. 

Dan: [00:00:51] Thank you so much for having me. 

Sascha: [00:00:53] I'm really excited to get into this today. We did do an episode on Sam Altman's Wild Weekend mid-last week, but at that stage, you know, this was an evolving story. We left that episode with two fairly unsubstantiated but reasoned theories from Alec about why Sam was fired. Before we look back at the year that was and sing a happy birthday to ChatGPT, I did want to ask you, do we actually know why Sam Altman was fired yet? 

Dan: [00:01:21] We do not, which I think is a pretty big problem. So there is, as you mentioned, a lot of conjecture as to why he was fired. So it's been pretty extensively reported by a lot of reputable publications that it has to do over a schism within open air and within Okinawa's former boardroom over air safety. So OpenAI was started as a non-profit with the altruistic purpose of developing safe air, essentially as a counterweight to Google. So that was at least that's why Elon Musk invested in it. And so fast forward all these years, now it's on track to make $1 billion a year. It has all these agreements with these big companies. And so there's this tension between we're a Start-Up that needs to make money so that we can get funding to build A.I., because building AI is really expensive. There's tension between that, and our purpose is to build responsible ROI. So there has been heavy reporting that that was the cause of his ouster, that people at the board were concerned that Sam Altman's enthusiasm for commercialising air was kind of about the sacrifice of air safety or safe air development. The problem is no one involved in the situation has confirmed that. Mm hmm. Emmett Scheer, who was the interim CEO for like 36 hours. He claimed that that was not the case, although it was also reported that he didn't know, like he wasn't actually told what the reason was. Microsoft CEO Satya Nadella said in various interviews that somewhat implausibly, but nonetheless, he said that he didn't know the reason Sam was fired. And so it's a really big question mark, but it is an important question mark. It goes beyond the kind of typical intrigues of a boardroom scuffle where there's money at stake. Because Sam Altman has spent, you know, plenty of time over the last year speaking in public about how dangerous A.I. can be. And so if this is actually about safety, then it stands to reason that the public deserves to know if it's at risk. Because Sam, according to reports, was the one less serious about safety in this dispute.

Sascha: [00:03:33] And that is, as you said, like the schism because he was, you know, has been doing the rounds for the last year and talking openly about the challenges in their path and how important it is that they get it right. Do you think we're going to find out why he was fired or what kind of timeline are you suspecting that we'll have before we get answers? 

Dan: [00:03:53] Well, so the new board, which was created upon his return, which includes Brett Taylor, is a former Salesforce CEO. And Larry Summers, I believe it has been reported that the purpose of the new board is to vet a new new board. And so I suspect if we're going to find out about it, it's when the final board kind of take shape. And then I assume they can regroup and get their message in line. But then, you know, it's one of those things where by that time, like it will be very difficult to take whatever is said at face value. Hmm. So whether they speak about it as one thing, but whether they speak earnestly and about it, you know, the truth is another thing. I suspect that will always be something of a question mark around it. But, hey, it's been I feel like after that weekend, it's very difficult to make any predictions with any certainty. So let's just wait and see. 

Sascha: [00:04:44] Yeah, absolutely. One more question before we move on and start looking at the year that was. You wrote a piece that Microsoft played this entire scenario brilliantly, obviously open to their main competitors being Google. And you said Elon Musk. That's why he invested in it. In your opinion, how are they feeling after the debris has settled on what's just happened? 

Dan: [00:05:05] I think now they feel good. I think they were not stoked about it when it happened. I mean, like I said, such a Nadella kind of did some media rounds and he seems like a really happy guy, a real happy guy. But then when he was asked about it, he kind of got a bit more morose and was like, we don't like surprises so much. And he said like, yeah, this we're never going to really let this happen again. So I think, you know, obviously the point is Microsoft owns 49% of open air. Its investment is hard to calculate because a lot of it is in your credits because a lot of openai's development happens this year. That's another thing. But Mark owns nearly half of open air and I believe 75% of Openai's profit at the moment goes to Microsoft until their investment is recouped. So they have a big stake in it and at the time were not super stoked. I suspect that the new board will have representation from Microsoft in one way or another. So I think this is one of those cases where it will actually end up. Kind of to that benefit. There was a bit of a mutiny at OpenAI where about 700 employees, which is about 95% of their workforce, said that they would all go to Microsoft if Sam wasn't reinstated. I'm somewhat suspicious of that because, like, how often in any company can you get 95% of employees to agree on anything? So it could be bad for Microsoft if open, I guess a bit of a black mark against it for whatever caused this schism. But outside of that, you know, the reputational hit of open eye and maybe this maybe this schism isn't over, you know. But outside of that, let's assume this is the end of that story. I feel like it ended up well for Microsoft. 

Sascha: [00:06:51] Hmm. So the reason that we're doing the episode today, Dan, is because it's ChatGPT'ss birthday. Happy birthday, ChatGPT. And I'm aware, as we've been talking about this, the arrival of open air into our lives is kind of brought up more questions than it's answered. And I don't necessarily feel like we're going to get to the bottom of all of this in the podcast today. Having reported on this issue for the past year, what do you think are the big macro themes and challenges that businesses are struggling with now? That chapter is in our lives. 

Dan: [00:07:22] I think privacy is the big one. So obviously ChatGPT came out on November 30th last year and it became a sensation and everyone started using it and finding it very helpful for things. The problem for big businesses is that employees started putting very sensitive information in jeopardy. So that led to a bunch of companies issuing bans. So, for instance, CommBank is one that quite quickly, I believe, banned any of its employees from using chargeback. Although open air offers charge enterprise, which is essentially you pay money and they will actually protect your data or try to protect your data because otherwise your data is just, you know, flowing from its API into the training model. And then, yeah, it's a bit of a free for all there. So certainly data privacy is a really big one and security is a really, really big one that companies are beginning to grapple with and which is kind of becoming part of Openai's business model.

Sascha: [00:08:20] Yeah, absolutely. So looking forward, what do you think are going to be the questions that we might not have answers for straight away, but things that we need to be aware of as consumers, as activity becomes more and more enmeshed in our lives?

Dan: [00:08:36] Well, it's a really interesting time because at the moment it is kind of a one horse race. So ChatGPT is the most famous large language model application, but it's not the only one. People who are disenfranchised with open AI for the exact reasons that we discussed before they left in 2021 and started their own company, Anthropic, which is backed by Amazon and Google and they have a chat bot. But more importantly, Google has yet to release its models. Facebook has an open source model called Llama. We're really early in this in terms of like the industry is likely to look very different in a year. But in terms of like what we should be aware of and what will shape the next year, I think will be regulation. And the EU is about to pass its first major, the world's first major deregulation, the EU AI Act. And there will be very, very significant impacts from that depending on how it goes, particularly as it relates to open source and close source models, which I know is a bit of like a can of worms. But that kind of arcane topic could shape how the industry grows for the next ten years. Close source, which is what open OpenAI now is. It's like a black box. Like no one knows what data GPT four was trained on. No one knows how many parameters it is, which is kind of the size of the model. It's, yeah, a complete black box. And so the argument that was kind of raging before Sam Altman's firing was, is big tech using the bogeyman of an AI apocalypse as a tool to regulate close us out of existence, which again, is an even bigger can of worms. But that will be very this is like an extremely it sounds like a, like I say, an arcane kind of like abstract thing. But this is like an extremely polarising topic within AI circles. And you can make people on both sides who argue very vehemently for, you know, if the wrong side wins, this will ruin AI for the next ten years basically. So watch out for that. 

Sascha: [00:10:40] I'm going to feature more of my conversation with Dan in just a moment. Welcome back to the dive. Quick favour to ask you before we get back into it, could you please give us a five star review? It really makes all the difference in terms of our show getting in front of you is and as a small independent media company, that really helps us. Let's now get back into my conversation with Dan now. Dan, I feel like there's too many threads that I can pick up and ask, but if we're back here in a year celebrating their second birthday, are there some kind of key stories or key themes that you would say to look out for that are developing in the next 12 months? 

Dan: [00:11:24] Yeah, definitely. I think the big question is where does opening I go from here? Not only on an ethics question of like the boardroom situation that we discussed before, but technologically, can it progress at the same rate as it has progressed over the last year? So it's kind of been an insane year in the sense of when they launched it, they launched it running. It was running on GPT three, which was a model they'd developed years ago. And then almost like two months later they released GPT four, which is its latest model. And then it had this huge technical improvement. And since then, they've steadily added features into chat about, you know, can read documents, I can search the web, things of that nature that have been quite significant. But now the question is, can they keep that momentum going? Because there is a big argument to be made that it's kind of a lot closer to the ceiling of what's possible than you would think. You know, I think there's been this idea that I tried ChatGPT is just the beginning and we're going to see a bunch of crazy things like this happen where the reality could actually be like, this is closer to the full picture than we think. Bill Gates recently said, I don't know if GPT five is really going to be that much better than GPT four because the cost of building these models is so enormous. And also these models have already read all the best data on the Internet that the idea is that like this kind of could be it. So in terms of the next year, it's a certainty that OpenAI well, not essentially after that boardroom situation, but is a very high probability that OpenAI will be looking good next year. It's a question of. This time next year, will it just be a little bit faster, a little bit better, or will it build exponentially like it's kind of implied it would? I think that's the big question. And what is Google doing? Google the team in charge. GPT trivia stands for Generative Pre-trained Transformer Transformer was a technology that Google invented and the people that invented it. Google of all since last. But it's a crazy open question of like, what's Google doing with AI and whether they can kind of get out of the gates and to what extent they do. I think we'll have a really big impact on Open AI's next year. So this last year has been an unusual one because I think they have been the only horse in the race. But next year the two questions are what horses will join them and how much faster can OpenAI actually run. 

Sascha: [00:13:50] Yeah, absolutely. I think when you raise that point, I thought most of the improvement that I've seen in chatbots or my experience of it has been in my own ability and my own ability to use the tool and understand what it can do. The actual tool itself hasn't improved as quickly as I have. 

Dan: [00:14:08] Yeah, yeah, exactly. And actually it's worse than that because these models are trained on. They're so complicated. There's a thing called A.I. drift where if you tinker with one part of the algorithm to make it better, it's very difficult to do that without affecting some other part of the model. So every few months you'll see on Twitter or whatever that, oh, it's tragic, it's getting dumber. And it's often not that it's getting dumber or nerfs or what have you. It's just that in tinkering with some part of the model, it actually makes another pot work. So all of which is to say is improving these models is actually really difficult. It's not the same as like making a new iPhone every year where they've got a model and they kind of know how to make it a little bit better. Raya. They still seemingly require breakthroughs such that they claim they had. By the way, the day before Sam Altman got fired, he was actually speaking at a conference and he was like, hey, we we just had one of the four biggest breakthroughs we've ever had in open air. But who knows if that's true and what that means? 

Sascha: [00:15:07] Well I feel like this is an episode where we get to the end with more questions than answers. But the one that seems well, hopefully we get answered in the next couple of weeks is why was Sam Altman fired? And then a whole lot more to be considered in the 12 months to come. Dan, thanks so much for joining me today on the dive.

Dan: [00:15:22] Thanks so much for having me.

Sascha: [00:15:24] And let's leave it there for today. I, for one, am certainly fascinated to see where we end up in 12 months, in two years, another decade with ChatGPT, open air and all this artificial intelligence discovery amongst us. Thanks again for joining me on the dive today. I'm going to be back in your face on Monday when I'm talking with Alec about Kyle and Jackie being in their $200 million contract. We're going to dive into those numbers. Until then.