I’m joined now by Mal Fletcher, a futurist social commentator and chairman of 2030 Plus. We’ve been hearing this week about some research done at the University of East Anglia into the artificial intelligence chatbot, GPT, and their research suggests that it has a significantly systemic left-wing bias. They’re raising concerns about how this bias in artificial intelligence chatbots could impact things like future elections. Are you concerned by this as well?
Yes, I am. It’s been a debate going on for a little while now within the area of AI Tech. I’m not sure of the exact process or parameters in the study, although I did read Sky’s report on it. It is being reported as a bonafide reliable study. If the results are sound, it wouldn’t surprise me in the least. Mainly because of two reasons. AI is trained using data provided firstly by human developers and programmers, and then it trains itself using ongoing exposure to online material provided by other human beings. We’ve all seen from social media just how inaccurate and how fired by heat more than light human interaction can actually be. Within that second category, that’s AI training itself through online data, I suggest that the majority of data related to political news is generated by media and media-related people, whether they’re journalists or researchers and so on. And as a result, most of that material will be at least slightly left-leaning. But one simpler reason journalists and researchers are trained mainly in this country in Liberal Arts and Humanities faculties, and in most universities in the UK, those facilities very much lean towards the left politically speaking.
Now, there have been some suggestions in the past that in your average Newsroom, you’ll have a mix, and maybe the journalists might tend towards left-leaning and the owners might tend towards right-leaning. Of course, that’ll be different in different newsrooms. But when we look at the whole issue of what you’ve been describing, can we see parallels almost with humanity itself? Because we are influenced primarily by our caregivers growing up and then by the people we interact with around us in later life. Is it the same sort of process going on?
It is very much or very similar to that. It depends on where the preponderance of our influence comes from, doesn’t it? As we grow up and develop, and AI is like a child that’s being trained in certain ways of thinking. However, it isn’t human, it’s still a machine, and it is a very good collator of material already provided by human beings. And it’s able to feed that back to us in very human-like fluid ways in various formats. But the potential negative impact of this bias that is said to exist with AI in terms of politics and elections is that it doesn’t just affect politics per se, it affects social policy. And that, let’s face it, shapes most of our lives from the way our children are taught in schools and what they’re taught, to rules that govern behavior in the workspace, to questions surrounding freedom of speech and belief, and a whole lot more.
There have been some articles in the past suggesting that there’s a range of biases within different AI chatbots, and some people suggesting that Meta’s Llama AI is the most right-leaning, and others from the University of East Anglia, as we heard, suggesting ChatGPT could be quite left-leaning. Could it be a case where we end up, as with our newspapers, just choosing the AI that best reflects us? Or will AI not actually work in that way?
Well, I think AI will work according to the channels it’s given to run on. It’s a bit like a train having tracks laid down for it. I don’t think anytime soon, as you seem to be suggesting, we’re going to see any balancing out of political bias either in our teaching institutions that I mentioned, which might have enabled a better balance of views in media professions and therefore in AI. But short of that, I think we’re going to need to focus on the technology of AI itself.
Governments in the U.S and elsewhere are experimenting now with ways of ascertaining where foreign powers are providing misinformation, generating misinformation in local elections, in national elections. And they’re looking at ways of using technology to combat that. And I think we might need to see the same approach when it comes to an overabundance of domestically generated material that papers one-sided politics or the other, if indeed such an overabundance occurs at different times and places, say in an election or in the lead-up to an election.
You think, Mal, just finally, that there’s going to be a financial appetite for this amongst governments across the world because we’re seeing a great emphasis with the war in Ukraine, for instance, on spending money on physical means of defending democracy rather than online AI digital means of defending democracy?
Oh, I think that’s changing, David. I really do. I mean, the Ukrainians have changed that probably more than anybody else. One of the first things they did was set up a digital platform, not for so much propaganda, but just to inform people within Ukraine and beyond about what was actually happening. And you will remember, as many of our listeners will, when President Zelensky made his first brave statements from the underground and also above-ground government buildings in Kiev, saying, ‘I’m not going anywhere, I’m staying here. I’ve been offered exile into other countries, I’m not taking it. We’re fighting this together.’ They were very powerful messages. And since that time, we’ve seen a government body, a government department in Ukraine developed specifically for digital information. And it’s had a powerful impact on, many would say, winning the digital war against the far greater military power of Russia. So, I think we’ve already seen Ukraine prove the benefit to other governments of controlling the digital space or at least shaping the digital narrative. I’m going to see a lot more of it.
