A lot has been going on at the intersection of AI and the art field recently. There's a lot to say about how AI will change our lives, and I feel that the current debates are missing the point.
I've laid down my thoughts on the subject here:
- I don't believe that MidJourney / DALL-E "steal from artists". The training process is eerily close to ours, only on a much bigger scale.
- I find the claim that ChatGPT helps cybercriminals dubious, at best. For them, ChatGPT only solves non-problems. The only practical cybersecurity application of ChatGPT I'm aware of is my IDA plugin Gepetto, and it helps defenders.
- Overall, I see a lot of negative reactions that can be boiled down to "[AI product] is bad because it threatens my job". But all jobs are going to change, this is the discussion we should be having.
- People don't understand how these models work, which prevents them from properly thinking about them.
All this noise gets in the way of the actual discussions we need to have. What happens when most occupations are about supervising AIs? Is it okay to have Microsoft control what could very well be mankind's gateway to knowledge in 10 years? What power structures are embedded in these tools?
ChatGPT is amazing, but it's also a general-purpose chatbot. Odds are, narrower AIs will be even better at specialized tasks, these are the ones we should be worried about.
I've been thinking about this with some degree of panic since last December. Every time, I'm hitting a wall, I can't for the life of me imagine where this is all going, what the ceiling is on this thing. If anyone has a clearer picture of what's coming next, be sure to ping me.