A Few Things I Believe In

For folks who don't know me, I figured it would be good to lay down a couple of my core beliefs, which shape how I choose and approach problems.

AI systems can and should be understood.

Many have been led to believe that these are black boxes by design; that this is an inherent property. While they may be black boxes right now, that doesn't mean they always will be. In fact, we are making real progress towards understanding various components and behaviors within these systems. Plus, having a working technology before a working theory of it isn't new: the steam engine is a canonical example of this.

AI should respect and compensate IP.

While copyright law and its enforcement, similar to our democratic systems, needs a revamp to meet the moment, the fundamentals still hold. Holders of IP should be able to decide whether and under what conditions their work is used. We've been here before — see the streaming and music industry.

AI-powered attacks, whether cyber, cognitive, or else, can and will be mitigated.

When first introduced, spam and phishing, just like AI slop today, felt like they would upend online communication and shopping forever. And then, we developed spam detectors, additional security protocols, and more, and suddenly they become an annoying, yet contained part of our internet experience. Again, we've solved a problem just like this before.

AI should be regulated and governed using our democratic systems, just like any other technology.

The sheer power imbalance that AI creates demands this. The democratic systems I'm referring to, of course, need an update to reflect the pace and reach of AI. While difficult, this will ultimately be the only way in which AI can be integrated prosperously in a free society, and used against autocratic regimes.

AI is as good as its stewards.

When it comes to AI, it feels like we're being sold either a story of utopia or dystopia. While both are theoretically plausible, I think the truth is much more in the middle. It will take a critical mass of people who genuinely mean well to deliver on the big prosocial promises of AI.

As a lifelong optimist, I do believe we can and will do good on all of these. I wouldn't be working on AI research if I didn't.