I think Ireland should have an organisation which does research on AI safety.
Its goals should be:
AI has the potential to be humanity’s most important invention. It’s probably very important that we make sure it goes well.
There also isn’t a lot of time to make sure of this, with the speed of AI’s advancement remaining pretty rapid. On Metaculus, a website where forecasters make predictions (usually fairly accurately), the median guess for when an Artificial General Intelligence (a system that performs at human level for most tasks) may be announced is only January 2033! And many tech CEO’s are regularly revising their timelines downwards as progress moves faster and faster.
It doesn’t take an expert to see that having armies of human-level intelligences appear can be pretty disruptive to society in a whole host of ways, from upending the labour market to changing the nature of war. We’re totally unprepared for this. And that’s assuming a more manageable scenario where progress stops there and the AI systems don’t continue to improve, which they of course will.
So something has to be done, but what?
Research is probably a good place to start. We need to figure out what to focus on and to invent solutions. Thankfully, there are small teams around the world working on this already, but not nearly enough considering the stakes. These researchers focus on everything from how to evaluate AI threats to making them more transparent, to controlling how the computer chips on which AI runs are distributed and used.
Different researchers and different jurisdictions have strengths in terms of what they choose to work on. By playing to our strengths and not duplicating our work, we can make this grand challenge a little more tractable.
So if we’re doing the work of helping ensure our AI-future is a safer one, why should we do any of it in Ireland?
Ireland has a few advantages: