250 words of advice on AI policy to the UK

My September testimony to the APPG came with a requirement to summarise my talk in 250 words, which I've just met and decided to share.

Firstly, AI is not mathematics but computation. Intelligence is not an abstraction, but rather a physical process subject to natural laws and principles of computer science. Computation requires time, space, and, energy. Even the number of possible chess games of 35 moves or less is greater than the number of atoms in the universe. Thus singularity-oriented concerns about AI are misplaced. Intelligent systems design is a long-term arms race for advantages in insight, comprehension, and planning.

The UK is well-positioned for AI expertise, but AI as ICT implies business potential to reach billions of customers. For this, the UK requires transnational cooperation. The EU presently leads globally in AI and data policy, and also embodies a leading model of transnational economic and legal cooperation -- a position gained with British ingenuity.

AI is definitionally an artefact, meaning it is built deliberately, and therefore is from inception a human responsibility. We should maintain human and corporate responsibility for all AI products, because our justice system rewards and dissuades humans, not machines. Auditing of AI does not require knowing the `meaning' of neural network weights any more than auditing accountants involves individual human neurons. By maintaining standard product liabilities for AI, we encourage not only responsible manufacturing and operation of intelligent systems, but clean, maintainable code benefiting also industry. We should not reward companies for poor systems engineering practices by reducing liability for systems they cannot predict or maintain.

Cybersecurity is essential to reliable AI, and AI to cybersecurity.

For more, see also: