Societal Impact of AI: Government Policy
I had better press on, because we have a number of colleagues to get through. Regarding generative AI and its effect on information integrity, which my right hon. Friend the Member for South Holland and The Deepings referred to earlier, we should bear in mind that so far the massive growth in fraud, misinformation and disinformation has all been done with cheapfakes. People talk about deepfakes, but it is done with rubbish, which can still take in large numbers of people. Just imagine what happens when such fakes become convincing. For example, there is the email that says, “I am stranded in Bermuda”, or whatever it is, “and I need you to send me £100 overnight.” Imagine when that scam actually involves a video of someone’s child saying it to them. However, we soon realise that those things seem relatively benign when people start talking about agentic AI and super-intelligence, whereby machines take control of a situation, which some people have described as being ultimately an existential threat to humanity. I do not know what all the answers are, but I think that as politicians—as the House of Commons—we need to ensure that our organisational architecture is equipped to consider all these questions as they evolve. However, I do not think that our Select Committee structure today is equipped in that way. For example, I do not think that having a Science, Intelligence and Technology Committee, which obviously covers technology, is a suitable way to address all these matters, and I also do not think that it is appropriate to expect every other Select Committee to develop such expertise themselves. Somehow, we need to meld the two structures. We must have some form of super-Select Committee—a support team, or something like that—to work with all the other teams or Committees across the piece. I will finish by mentioning two specific things that we can do right now, because there is a ticking bomb with AI, specifically regarding children. In education now, if someone asks the typical year 9 pupil, “How many of your friends use AI to help them do their homework?”, that child would just laugh and say, “All of them.” The idea that we have some sort of issue where we have to encourage children to use AI is just nonsense. It is also crazy to suppose that we can just say to them, “Don’t do it,” and they will comply. Every child in the country learns computing in primary school. We need to make sure that the curriculum evolves so that children get to understand how the thing works, how it generates its content, why it might hallucinate, why it tries to flatter the questioner and why we cannot fully rely on it. There is a real danger. Cognitive offloading is a fancy term, but it basically means relying on a machine to do the work for us. One way we can mitigate that effect with schoolchildren is ensure that when they do their exams at the end of the year, they do them with a piece of paper and a pen in exam conditions. I am afraid there is a move across exam boards to make more exams digital. Of course, they will say that the computers are not attached to the internet and children cannot do this, that and the other, but pen and paper is the safest thing. Of course, there should be exceptions for children with a special educational need or disability that means they need to use a keyboard, and exceptions for computer science, but in general, children at school doing their public exams should do them on pen and paper. That will make sure it happens lower down the school as well. It will also ensure that children continue to write. This sounds like a stupid thing to say, but people already ask, “Why is my child learning to write with a pen, because by the time they grow up nobody will be doing that?” Finally, we are about to make the same mistake with chatbots as we did with social media. I am talking about anthropomorphism, where a computer program develops a personality of its own and develops relationships with people. I welcome the development of an age-appropriate design code for chatbots, but it is not enough. The regulatory regime we have in this country, with Ofcom and the Information Commissioner’s Office, relies on there being evidence of harm. The evidence will not come before the harm. We need to change that system and adopt a precautionary principle when it comes to protecting children.