This is the year of AI. We’ve been using Midjourney, DALL-E 2, and Stable Diffusion to generate artwork. ChatGPT wrote most of my blog this past Tuesday. Yesterday, Lensa turned my wife into a warrior princess (she was already fierce, BTW). This is super fun, but you know what’s also fun? Competitive programming. Yes, competitive computer programming is a thing – a big thing, actually – and now DeepMind’s AlphaCode (an AI model that writes computer programs at a competitive level) has achieved “approximately human-level performance” by receiving an average ranking in the top 54.3% in simulated evaluations.
How far can all of this go? AI is experiencing a very visible growth spurt. At the moment, only a handful of companies have the computational power to deal with large models. Is there a plateau coming? And, if so, will it advantage big tech or will AI tools evolve to democratize the industry?
Take a moment to research the size of the data sets required for training the most popular models. Take another moment to research the size of the computational environments required to achieve reliable “high stakes” outputs. (Lensa is low stakes; it’s a parlor trick. Autonomous vehicle AI is high stakes.) Consider which organizations are most likely to be able to field the required amounts of resources. It’s hard to imagine a world where big tech does not get much, much bigger.
One day, the science will change and we’ll all be able to do all of this locally (on our laptops or smartphones). Until then, it looks like AI is going to be as capital intensive as the space race – and of equal economic and military importance.
Author’s note: This is not a sponsored post. I am the author of this article and it expresses my own opinions. I am not, nor is my company, receiving compensation for it.