Beau Cronin

Deep Learning Business Models

The DeepMind acquisition was the starting gun for a deep learning startup goldrush. Most of the new companies are driven by the assumption that there is a big opportunity in bringing deep learning “to the masses” - i.e., all of the companies and startups that could benefit from the technology, but don’t have the in-house expertise that Google, Facebook, Yahoo, and Microsoft do.

I think they key issues are as follows:

  1. Deep learning requires very large training sets, and these shouldn’t be shipped around a lot.

  2. No one knows which applications deep learning will dominate: maybe just the ones it’s already doing well, maybe a few additional ones, maybe a large number. This has a big effect on business model (viability of platform vs. vertical), and there is real uncertainty here that will not be eliminated for some time.

  3. Lots of experience is required in getting deep learning to work well on a given application; the people who can do it well are those who have been at it for a while, and they tend to come from a handful of places. This will change, but the rate of diffusion of expertise is unknown and will also have an impact on business options.

Here is a brief tour of the different models; I know of companies in each category, but I won’t name names since some of them are unannounced.

There’s lots more to say, but this is one of those cases where I’ll never get this out if I don’t post it as-is.

Tell me what you think on twitter.