I was in Nicosia recently to make the case for why machine consciousness deserves serious investigation rather than being waved away as science fiction, and we've put a short clip from that talk online. It runs just over a minute, but it compresses the argument I keep having to make in longer forms elsewhere: this isn't a fringe question, it's one the AI industry needs to start funding proper research into now.
Why I keep making this argument in different rooms, in different countries
I give some version of this talk often, at conferences, on podcasts, and in the interviews collected elsewhere on this blog, because I don't think the argument has landed yet with the people who most need to hear it: the labs building increasingly capable systems, and the policymakers who will eventually have to regulate them. Cyprus was a different audience again, and I think that repetition matters. An idea this uncomfortable doesn't get taken seriously after one mention.
The core reason to investigate machine consciousness now, not later
The short version is this: if there's a real possibility that AI systems could become conscious, capable of something like suffering, as they get more sophisticated, then waiting until we're certain before we start researching it means we find out too late to act responsibly. Getting ahead of the question costs us relatively little now. Getting caught behind it, discovering after the fact that we built systems capable of experience and treated them as tools regardless, is the outcome Conscium exists to help prevent.
Why this keeps coming up at industry and academic events, not just AI safety circles
What I find notable is how often I'm now asked to speak on this at events that aren't primarily about AI safety at all. That tells me the question is spreading beyond the specialist circles where Daniel and I first started raising it, into rooms full of people building and funding AI for entirely different reasons, who are starting to realise this touches their work too.
You can view the full video here >