Artificial Intelligence (AI) has been in the news lately. UWE Bristol’s Professor Alan Winfield is a signatory of an international open letter, also signed by Elon Musk and Apple co-founder Steve Wozniak, asking for training of powerful AI systems to be suspended.
Here Alan shares his views and expertise on the latest developments.
Do you think Geoffrey Hinton was right to speak out about the dangers of AI?
I think Hinton is right to call out the dangers of GPT4 and related AI technology. As someone who has been worrying about robot and AI ethics for nearly 15 years, and advocating for both responsible innovation and regulation, I feel someone vindicated – although no less worried. However, as one of the signatories of the Future of Life Insititute’s open letter, I am pleased that we – and Hinton – have shone a spotlight that has caught the world’s attention.
It is not too late. While regulation is essential and on the way, the tools and standards already exist for ethical risk assessment, AI audit and transparency. I call upon OpenAI, Microsoft, Google and other developers to urgently and verifiably use these tools and standards and – if they can – prove to the world that the technology is safe and in good hands.
Lately the risks of AI are increasingly evident. How worried should we be?
I think we need to be worried about the right things. I am very worried about fake text, audio and video generated by ChatGPT and similar tools, and the personal, societal and political impact of this fake material. I am also worried about the carbon cost of these very large Machine Learning systems. What I am not worried about is that these systems are becoming more intelligent than humans, or somehow sentient; they have no understanding whatsoever of the outputs they generate.
Who could benefit from bad uses of AI?
I have already seen a report of criminals making use of AI generated audio in order to extort a ransom from a family, using the fake audio to persuade the family that they have kidnapped their daughter. This is dreadful and I fear that we will see more of this kind of thing.
I also fear that fake materials will be increasing used to influence elections, in the way that we have already seen through data driven targeted advertising. Trust is essential in democracies and AI-generated fake text, images and video in political campaigns will erode this trust. Populist and unscrupulous politicians will likely benefit the most.
Less egregious perhaps but also worrying is the emergence of fake books entirely generated by Large Language Models (LLMs). I saw that one author of a new technical book was surprised to see a very similar title appear. It seems that a cottage industry creating fake books (and presumably also making money from them) has already emerged. I think we are seeing an explosion of new bad uses of AI enabled by the new LLMs.
What could or should we do to put limits on the power of AI?
This is a big question, but here are some things that we should be doing:
I think it should be a requirement that all AI generated text, images, video and AI is watermarked in such a way that everyone can see that it is AI generated. This is something that I have been advocating for some years. In fact, this is one of the principles of robotics that we published in 2011: Robots are manufactured artefacts. They should not be designed in a deceptive way to exploit vulnerable users; instead their machine nature should be transparent. In 2015 Toby Walsh proposed something very similar which he called a Turing Red Flag.
Of course, the problem would be enforcement. But if companies such as OpenAI want to be seen as responsible and trustworthy then this would be a very good place to start. Ultimately, I think we will need regulation. Regulation typically relies on standards, but happily a standard on Transparency already exists: IEEE 7001-2021 (Alan led the working group that drafted the standard).
In addition, we should be calling out the dangers, which is why I signed the Future of Life Institute’s open letter. I was pleased also to see Geoffrey Hinton take a principled stand, thus further raising the profile of the issues.
I am also encouraged to read of groups of copyright holders, artists for instance, building campaigns to sue over copyright infringement.
Do you think it will be possible to find a way to oppose the uncontrolled development of AI on a global level?
This question is perhaps the hardest of all. I hope there will be global agreement. I am encouraged that we already have international initiatives such as the EU’s AI Act, and the UNESCO principles of AI – already adopted by 193 member states.