Reaction to the UK AI Safety Summit 2023

The UK just finished hosting the first AI Safety Summit this week, so this seems to be the right time to re-start my blog. Since my role not so long ago in an AI/ML start-up, I’ve become increasingly interested in the ethics of the implementation of AI.

AI is on everyone’s lips nowadays. It’s no surprise that plenty of businesses are experimenting with AI. The larger ones, with the ability to spend money speculatively and with no need for end product, can afford to spend the time to learn how to harness this next technology revolution (is AI the fifth great revolution after mechanisation, electrification, automation then networked information?).

The problem as I see it is that the technology has democratised faster than our ability to understand it. Small enterprises of single people (and I am one of these) can access ChatGPT for a reasonable monthly subscription, and so have “AI”. Even if they don’t understand the differences between Artificial Intelligence, Artificial General Intelligence, Machine Learning, Deep Learning or any of the other myriad of acronyms that try to convey important nuances. And crucially, they can afford this subscription without knowing if they can use it to make any money, so there is no brake on the democratisation.

Consider the problem through Jurassic Park eyes: everyone with a ChatGPT subscription can now be preoccupied with whether or not they can use AI, rather than whether they should. (With apologies to Dr. Ian Malcolm.)

Every industrial revolution extorted a heavy price that was paid over the course of generations and centuries. Mechanisation led to urbanisation which manifests now in systemic inequalities within and between countries. Electrification and automation led to the insatiable consumption of earth resources and climate change. Networked borderless information is causing globalisation of populations which is causing reactionary isolationism of governments. What will the AI revolution bring?

For one thing, the AI revolution will bring undreamed of opportunities for good. It will increase lifespans and health outcomes, and some believe it will diminish work to be merely a hobby. In order to achieve these benefits and avoid another heavy price, we must consider how we should deploy AI, not just whether we can, for each and every way that we use AI.

This is why I think the UK’s and US’s AI Safety Institutes are vital and welcome their establishment. We need to research tools to evaluate advanced AI systems as vigorously as we develop those advanced AI systems. We must collaborate free of commercial and national interest, with a diversity of experiences, to come to the right answer for us all. We must all cultivate a sense of AI ethics as deeply as our sense of personal freedoms.

On balance, I’m tremendously excited by this future.