toomuchtodo's comment is on the money, it is a bit of a fad right now with the recent acquisitions. But only by a margin.
AI is a bit of a counter-intuitive market: it's a high growth segment, but the structure of the market top heavy, and several things make it an unattractive investment:
1. Democratization: much of the technology is the public domain, meaning anyone with a CS degree can muddle together their own (and many have). That means if you invest, you investing in pure AI, which means you're investing in cutting-edge technology - i.e. Research, which very difficult to assess.
2. Dependencies & Go-To-Market: intelligence (i.e. pattern predictions) that's actually useful on a market wide basis require massive amounts of training data. There are a couple of resources, like ImageNet & Yahoo's data dump, but in the interim - AI needs to be domain specific. As such, the dependency is a widely used application (i.e. Facebook, Instagram, Google). You see the dilemma. Therefore, what seems the likely outcome is acquisition.
3. Unit of Economics: For this sector, it's the accuracy of its predictions. This is almost too one dimensional to assess as the basis for investment. Acquisitions decisions are made almost exclusively by the R&D department. Therefore, if you invest, you need to know what makes one AI different/better than the next.
IMO, it's too early to put money in. The structure of the market is still nascent, heavily favoring incumbents (Google & Amazon offering AI predictions as a service). It's difficult to see how an AI startup can break in, and how it can be profitable.
Take the above with a grain of salt, as it's tailored to my risk appetite.
Comments
toomuchtodo's comment is on the money, it is a bit of a fad right now with the recent acquisitions. But only by a margin.
AI is a bit of a counter-intuitive market: it's a high growth segment, but the structure of the market top heavy, and several things make it an unattractive investment:
1. Democratization: much of the technology is the public domain, meaning anyone with a CS degree can muddle together their own (and many have). That means if you invest, you investing in pure AI, which means you're investing in cutting-edge technology - i.e. Research, which very difficult to assess.
2. Dependencies & Go-To-Market: intelligence (i.e. pattern predictions) that's actually useful on a market wide basis require massive amounts of training data. There are a couple of resources, like ImageNet & Yahoo's data dump, but in the interim - AI needs to be domain specific. As such, the dependency is a widely used application (i.e. Facebook, Instagram, Google). You see the dilemma. Therefore, what seems the likely outcome is acquisition.
3. Unit of Economics: For this sector, it's the accuracy of its predictions. This is almost too one dimensional to assess as the basis for investment. Acquisitions decisions are made almost exclusively by the R&D department. Therefore, if you invest, you need to know what makes one AI different/better than the next.
IMO, it's too early to put money in. The structure of the market is still nascent, heavily favoring incumbents (Google & Amazon offering AI predictions as a service). It's difficult to see how an AI startup can break in, and how it can be profitable.
Take the above with a grain of salt, as it's tailored to my risk appetite.