What happens when AI gets ‘human’ cognition?

Sam Hawley: Hi, I’m Sam Hawley, coming to you from Gadigal land. This is ABC News Daily. A year ago, ChatGPT was just taking off. The artificial intelligence bot amazed people with its ability to do anything from writing job applications to travel itineraries and essays. Since then, the technology’s been evolving. So are you an AI Boomer or Doomer? Do you think artificial intelligence will make the world a better place? Or are you worried it could destroy our way of life? Today, Professor Toby Walsh, the chief scientist at UNSW’s AI Institute, on the recent fight over AI in Silicon Valley and the latest innovations we need to know about.

Toby, I want to start with the chaos last month at OpenAI, the company which of course created chatbot ChatGPT. ChatGPT, it took us all by surprise really, didn’t it? Just remind me how advanced this is now. What can it do, Toby?

Toby Walsh: ChatGPT, it definitely caught us all, I think, rather unawares. It can answer questions from astronomy to zoology. It can write poems in the style of Shakespeare. It can write your business letter. Indeed. I’ve written my last business letter – ChatGPT writes them all for me.

Sam Hawley: Oh, no.

Toby Walsh: So it’s a very capable chatbot. But equally, there’s still some things where it’s remarkably dumb. So we’ve still got quite a distance to go, but it could do useful stuff. It can do a lot of the routine, repetitive stuff that that we humans used to do.

Sam Hawley: Right. Sounds like it’s cutting your work hours down, Toby, which sounds pretty good. But the company that came up with it, of course, OpenAI, it landed itself in a spot of bother last month, didn’t it? Tell me about that.

Toby Walsh: It did. I’m not sure whether it was, um, uh, Succession or Game of Thrones. There was a real drama. The CEO of the company, Sam Altman, a very charismatic, youthful figure, was fired out of the blue on a Friday afternoon.

ABC News clip: The artificial intelligence company issued a statement accusing Altman of lying to the board, and that it no longer had confidence in his ability to lead.

Toby Walsh: And then there was a big uproar. Microsoft, which had put over $10 billion invested in the company, was told a minute before that he was going to be fired that the board had decided to fire him.

ABC News clip: The start up that delivered the world ChatGPT is in turmoil after a majority of its staff threatened to follow sacked co-founder Sam Altman out the door to Microsoft.

Toby Walsh: There was a campaign from within the company. Many of the employers threatened to resign and move to Microsoft unless he was reinstated. And at the beginning of the following week, he was reinstated.

ABC News clip: The former CEO of OpenAI is returning to his old company just days after he was controversially sacked by the board.

Toby Walsh: And the board who had fired him was changed, and only one member of the board still stays. And there’s a completely new board in charge of the company. The board is supposed to be ensuring that OpenAI, as a company, goes about its mission, which is to develop AGI, artificial general intelligence, AI that’s as smart as a human and do that in an open way and for the benefit of all of humanity.

Sam Hawley: AGI – this is the next stage. So let’s come to that in a minute. Toby, this all sounds like a fantastic TV series, by the way. But, uh, according to reports by Reuters, before the removal of Sam Altman, for that short period of time, there had been some really big advances at OpenAI, and some of the staff were really worried about them, weren’t they?

Toby Walsh: Yes. The chief scientist, Ilya Sutskever, has become he’s part of the safety team, and suddenly he and others within the organisation are, as are, you know, many other people in the in the wider world, concerned about, you know, where this technology is going to take us. And the concerns range from the very media about the use of this technology, the way that it might displace people’s jobs, the way that it might be used to create misinformation and disrupt our elections onto the existential threat that it might pose, that if we’re no longer the smartest thing on the planet, what happens when these machines have ideas of their own?

Sam Hawley: Mmm alright. But there was certainly a split in the company between those who were worried that OpenAI is moving too quickly, and those who think the technology is the solution to all of the world’s problems. And as you mentioned, that’s a split across the industry as a whole between those who’ve been dubbed the Boomers and those who’ve been dubbed the Doomers. So tell me, how fast is the technology progressing?

Toby Walsh: It definitely is moving very fast. ChatGPT is the fastest growing app ever. It’s now just celebrated its first birthday. It’s available to over a billion people, and we’ve never had technologies that could impact people’s lives so quickly in the past. Not only was ChatGPT the fastest growing app ever, but OpenAI, the company behind it, is the fastest growing company by many measures ever. It’s going to take $1 billion of revenue this year, so it’s gone from almost zero revenue last year to $1 billion this year. We’ve never seen such wealth being generated.

Sam Hawley: Hmm and writing business letters is one thing. But with this artificial general intelligence that we mentioned before, that’s AI with human cognitive abilities. Tell me just a bit more about that.

Toby Walsh: It’s worth pointing out that we’re probably not at AGI yet, that we’re not- We haven’t got a computer, an AI, that can match human reasoning, human cognitive abilities, and all of the the breadth of capabilities that humans have. I mean, I’ve been studying the field for more than 40 years now, and I still have huge respect for the human brain and all of our, our capabilities. But the the goal, the long tum goal of the field is to try and match and if we can match, perhaps even exceed the cognitive capabilities of humans. And there are some parts of that where we’re still very far away from. It’s, it’s very easy to point out the ways in which something like ChatGPT can be remarkably stupid at times. I mean, my favourite example, if you ask, I just checked. It’s still the case. If you ask ChatGPT how many Ss in the word bananas, it will tell you confidently there are two Ss in the word bananas.

Sam Hawley: There’s a trap for all the school children. Watch out..

Toby Walsh: It is, but nevertheless, they can still do some quite remarkable things. And given the rapid advances that we’ve been making in the last couple of years in AI, it’s natural, I think, to consider, well, maybe we’re going to get to AGI in the not too distant future. It’s hard to know how long it’s going to take. Is it going to take 5 or 10 years? Or is it going to take 20 or 30 years? But I think most of my colleagues think it’s not going to take much more than that in the worst case. And that’s that’s going to be quite profound because it’s going to have a huge, great impact. It’s going to change the nature of work. Maybe lots, quite a few number of jobs are going to be displaced, but also it’s going to be very disruptive on other aspects of our society. On the course of our democracy. We’ve all seen, you know, deepfake videos now where generated by AI, people are starting to use that in elections. We saw that in the recent election in in Argentina, there was an election in Slovakia a few weeks ago where there was a deepfake audio released right at the last moment that many commentators suspect actually had a material impact on the outcome of the election. And next year, we’re going to see India is going to the polls, Europe’s going to the polls, the United States is going to the polls, the United Kingdom is going to the polls. Many people are fearful that these tools are going to be used to try and persuade us to do things that perhaps aren’t in our best interests, and at the same time, we’re seeing many of the tech companies cut back on the amount of content moderation and the and the teams that they’ve got in place to ensure that that people aren’t getting up to any mischief.

Sam Hawley: So, Toby, what about regulation? We haven’t been really good with regulating social media, of course, but there has been movement in the EU in terms of AI. What have they done in the past week?

Toby Walsh: The EU has now agreed to the terms of what’s going to be the world’s first AI Act. It will leave many case uses of AI aside, and it will focus on the high risk applications and the large platforms. It certainly is a welcome development to see that there is on the policy books soon regulation that is going to actually, perhaps potentially ensure that AI is used in a responsible way.

Sam Hawley: All right. And what about Australia? Toby? The Minister for Industry and Science, Ed Husic, he says the government is watching very closely what the EU is doing…

Ed Husic, Minister for Industry and Science: Harmonising our approach where we can is important, but acting locally, where we have to is always been in our government’s thinking.

Sam Hawley: …But are we actually progressing on this matter here?

Toby Walsh: Unfortunately, we are laggards in this race to regulate. Um, there was earlier this year a call for submissions to the Department of Industry, the body responsible for, for this area. Um, there were over 500 submissions from various parts of civil society, academia and the tech companies themselves. So there’s definitely an appetite to do something. But the government has still yet to make any announcements as to how they’re going to respond to all of the ideas in these submissions.

Sam Hawley: The question is, if these risks are so serious, why are we doing this at all? We have the power to stop this, right? Should we really keep going down this path?

Toby Walsh: Well there was an open letter, a call for a moratorium, a six month pause to try and catch up on the on the safety and the regulation. There are a number of reasons why I’m doubtful that that will come to anything. One is we’ve already talked about is the huge great financial benefit at stake here. We’re seeing trillions of dollars of value. It’s been estimated that AI will add 15 per cent to the world’s GDP in the next decade or so, but equally, we face some pretty wicked problems today, whether it be the climate emergency, whether it be increasing inequality. We see within and between societies the sorts of technologies, the AI technologies that we’re talking about do offer promise to actually help us tackle some of these wicked problems, to help us use the planet in a more sustainable way, to to help improve health care and education. So there’s there’s great benefits the technology is going to bring. But equally, because it’s a powerful technology with with great applicability across so many different aspects of our lives, there’s also potential risks as well. It’s about, you know, moderating those risks, ensuring that ultimately we get a net benefit out of these technologies.

Sam Hawley: In the end, who do you think is more likely to be right, the Doomers or the Boomers?

Toby Walsh: The you know, the truth is always somewhere in between. There is going to be huge, great benefits this technology brings, but equally, they’re going to be huge, great harms. And the technology also brings. And so we’ve really got to carefully navigate between these two extremes and try and ensure that we end up with a better planet.

Sam Hawley: Professor Toby Walsh is the chief scientist at the University of New South Wales AI Institute. This episode was produced by Bridget Fitzgerald, Nell Whitehead, Laura Corrigan, Sam Dunn and Anna John, who also did the mix. Daniel Golding composed our theme song. Our supervising producer is David Coady. I’m Sam Hawley. The ABC News Daily team will be taking a break until January the 29th. Thank you for your company throughout the year. We’re really looking forward to being back with you in 2024. Thanks also to all the guests who’ve joined me this year, helping us to understand the news and the big issues shaping the world. Next week, we’ll bring you some of the episodes that covered the biggest issues of 2023. Thanks for listening.