At C10's Demo Day, I was asked to share thoughts on "The Future of AI." It made me think back to my first days at MIT three and a half years ago, when professors confidently listed things AI couldn't do, such as common sense understanding and reasoning in physical space. Those limitations didn't last long.
These professors are much smarter than I am. They didn’t know the future, and I won’t pretend I can either. That being said, there are places we can look for clues.
First, the future is often hiding in plain sight. While everyone asks how to keep up with AI's rapid changes, the underlying progress is actually more gradual. Slow dow. Read relevant papers in the vertical you’re interested in. Understand the fundamentals and where ideas originate. The current under the surface is a lot slower and easier to follow than the constant barrage of new consumer tools.
Previous jumps offer clues too. Open source datasets like LAION and Objaverse, plus foundation models like CLIP, unlock hundreds of thousands of downstream innovations, eventually leading to new products in the market.
While I was preparing for this talk, my friend Leroy noted that he’s “Constantly amazed by the propensity of humans to see something extraordinary and place it in the commonplace.” This is the typical path in technology hype cycles - truly useful breakthroughs get absorbed into existing systems and become background features. Depth-first search was once cutting-edge AI; now it's just another algorithm. The key is spotting what matters amid the noise.
I want to conclude with a message of optimism. There’s no ceiling to human innovation. Recently, I attended a presentation about a robot learning to perform tasks underwater. The speaker paused to call out the beauty of the experiment. The robot was safe there, protected from falls and overheating. He said, there's something poetic about new forms of intelligence emerging from water, just as biological life did. In our rush toward progress, remember to pause and look for these moments of beauty.
Image credit: https://aquabot.cs.columbia.edu/
