Happy Sunday and Welcome to InsideAI! I'm Rob May, Founder and CTO at Dianthus. You can read about what we do here. I'm also a very active angel investor, in over 100 AI related companies. If you have an early stage AI company, let me know. I've been the very first check into dozens of deals.
Lots of tech companies have had layoffs. If you were affected, I have a large portfolio plus Dianthus is still growing rapidly and still hiring so please reach out.
This week I want to talk about an interesting book - Why Greatness Can't Be Planned: The Myth of The Objective. The authors, Ken Stanley and Joel Lehman, work on "open endedness" which may be a key technology needed to move AI forward. The book is tough to swallow because they do a great job of demonstrating that objectives often work against us, stifle creativity, and make us miss our goals. But once you accept that idea, you are stuck with the realization that measuring progress is difficult. And there is nothing investors or managers hate more than lack of clarity around exactly what you are trying to do. But you should read the book and make your own judgment.
The key idea I took from it is the idea that sometimes you need to value "novel" or "interesting" ideas over direct progress towards an objective. Building algorithms to do this is difficult, and it's the core of the authors' research. What I like about it is that I've written before about this idea of how data sets could collapse into smaller, less diverse data sets because of AI. A perfect example is Gmail autocomplete. As it uses more and more algorithmically generated responses, will those increasingly drown out the nuanced diversity of human written responses? And if so, will the new data set be less diverse because it was mostly machine written? If that does happen in certain domains, one way out is algorithms that generate novel responses. These are currently difficult to do because, how do you define "novelty" in a way that still makes sure it meets some criteria of correctness.
Creativity is a core part of being human, and AI is a long way from doing more than mix-and-match style creativity. If you are also working on open-endedness and have some interesting ideas, I'd love to hear about them.
Thanks for reading.
@robmay
Comments
Post a Comment