4 Optimistic Takes on the Future of AI
Because there's a lot to be optimistic about
This is the most positive article I've ever published on TAB. If you wanted more of that, today's your day.
In September 2022, computer scientist Richard Sutton, author of the well-known essay “The Bitter Lesson” (popular among deep learning advocates and scale maximalists) posted this on Twitter:
I disagree on a few things there (if understanding intelligence is within reach, my bet is on cognitive scientists, not AI researchers, to attain that knowledge). In fact, I feel Sutton’s statement is so disconnected from the current direction of AI research that I see it as wishful thinking more than as a scientific prediction (he may have intended it that way).
However, the motivation to share that Tweet isn't to agree or disagree with Sutton, but to witness—and borrow some of—his unrivaled optimism. I thought: “my readers could use some of that.”
I don't consider myself a pessimist. I don't see AI as something that should’ve never happened or that will, inevitably, bring ruin to us individually and as a species. But I don't stand out for my optimism either—you'll never see me post a Tweet like Sutton's above.
I’m somewhere in the middle of those extremes (you may decide where exactly…): I appreciate the value AI progress brings but can't unsee the partially problematic nature of industrial AI systems and the limitations of dominant research avenues.
If I had to guess I'd say TAB irradiates an aura of “read this to understand the problems of AI” rather than “let's enjoy a light take on AI’s bright future.” Not that I'd change that. We need more of the former given how easily the latter leaves room for unbounded—and unfounded—hype.
But today I'm feeling optimistic. Tapping into the emotions that urge me to be more enthusiastic and excited about what's coming is a nice exercise to counter my pervasive skepticism. And I'm sure you, as a reader, would appreciate that I tone down from time to time (as some of the comments in my latest post suggest).
That's why in this article I'll share four optimistic takes on the future of AI. These are things I'd like to see happen that I think have at least a moderate chance of being true.
This essay is also a reminder that it’s possible to hold apparently conflicting views without exploding (e.g. “AI has deep flaws” and “AI has great virtues”). I even think it’s useful as it provides a richer perspective of the world around us.