-- Commentary --
I read a really interesting article this week about The Artificial Intelligentsia. It chronicles the author's experience at Predata, a social media predictive startup. I want to address a couple issues in the article related to startups, and AI startups in particular, and then look at a key question in the article - why can AI startups get funded in cases where there isn't direct demand for their product?
I want to start by cautioning you to read too much into an article with this tone. It could be highly factual, and it could be one pissed off employee's jaded view. My experience as an entrepreneur and investor is that most tech stories about startups, and most gossip you hear about them is incomplete at best and inaccurate at worst. In fact, the whole startup scene is often like this. I remember when I first moved to Boston and met some of the key entrepreneurial figures here, I would get mixed reviews on them from others. "Oh, you are meeting that person? He's brilliant." Then later the same week "Oh, you are meeting that person? He got lucky. Total incompetent asshole." Startups are very difficult, and entrepreneurs tend to have personalities that are simultaneously charming and abrasive which is why you hear different sides of a story out of the same company. You never know what experiences someone had, and why, and companies in this situation rarely jump in to tell their side of the story because it just draws more attention to it.
That said, startups are definitely filled with many pretentious engineers so wrapped up in their brilliance that they can't be challenged. I don't hire them, but many of my CEO friends have, and have paid the price. There is a class of people who come out of college and just go where the money is, and while they used to go to Wall Street, then consulting, now they go into startups. This isn't bad in and of itself, but, it kills the vibe of the people that go into those same fields because they really care and enjoy it. Part of the problem with Silicon Valley today is that what was built by nerds who really loved technology is now filled with people who want to play startup because it's fun and you can still get acquihired for $1M if you fail. The nerds were the soul of the Valley and they have been supplanted by the money chasers.
So, yes, the article does highlight things that are real problems at startups but, be careful about reading too much into the personality assessments and criticism of company strategies. Doing new things is much much harder than writing about doing new things, and criticizing those who do new things. I don't know the Predata team, so I won't weigh in, and I'm not saying the expose is wrong. I'm just saying be careful what you read into these things. They are one point of view.
But the author has one line I think is worth discussing. He points out "No one, as far as I could tell, was clamoring for a social media-derived signals-processing tool to predict world events." This misses the point of startups. No one was clamoring for the iPhone either, or Facebook, or many new tools. There is nothing wrong with showing people new ways of working that they haven't embraced yet. Someone has to have vision and lead the market to a new place. Most of the clear high demand market needs are owned by big companies. Startups have to find the ones where the market need is unclear or missed, or too small too matter to a big company.
When I started my last company in 2009, which was cloud computing focused, almost every VC I met told me enterprises would never move their data to the cloud. I took a beating from VCs, eventually being rejected by over 60 firms, who all turned out to be wrong. There was very little demand for the product we were building, except that now it's an entire industry with multiple players competing and the 4 main startups all getting acquired. So I have lived through the phase where Predata is, and that is why I want to address this in the broader context of the question "why is there so much AI funding despite real results?"
Here is the world as I see it. First of all, AI has potential. We see this in few applications that work really well. Secondly, AI is still more art than science, which means it is difficult to predict, without actually going to build something, which things will work well and which won't. Third, the economics of venture capital, which is how most AI startups are funded, are such that all that matters is a few really big winners.
This last point is crucial to understand. If you are a VC, and you get in the one company that becomes the enterprise AI behemoth worth more than $10B, the rest of your fund doesn't matter. Therefore, the way you invest isn't that you look for highly logical ideas that are clear to everyone. You look for the one idea no one else has noticed yet that could be transformative. Most of the time, that idea turns out not to work, or not to be transformative, but in the 10% of cases where it does, it makes up for the losses from the other 90%. This is why, as a class, venture capital funds a lot of ideas that turn out to be dumb ideas but it is still entirely rational behavior.
What is happening in AI markets then, is this - you have a much greater difficulty of predicting eventual success than you did with SaaS companies, which were/are easier to build. There is more early stage capital than ever before. VCs realize it is all about getting into possible big big winners, so the competition for any deal that looks decent drives up the price. The higher price allows entrepreneurs to raise more capital for the same amount of dilution. The net outcome of this is that companies can get more funding, earlier, with less traction, if there is a reason to believe they could be on to something. The market configuration supports the rational funding of good stories because we don't really know where AI will be successful until we try, and we have tons of early stage capital chasing anything that could be a really big win.
While this looks bad in some contexts, it is good for society overall in terms of the things we are learning about AI, how to build it, sell it, and deploy it. For all the criticism that we are far off (which is true), the way we solve these things is by funding thousands of small practical experiments, which are AI startups. And while you may hear a lot about how AI is on the wrong track, the truth is that people are exploring all kinds of approaches, even many outside the mainstream and currently popular deep learning approaches, that are pushing the industry forward.
This is why AI funding continues to climb even though as an industry, the results we have delivered so far are modest. But they are coming. I see it, and I am a believer. A big part of the problem is that the buyers of AI are still figuring out what they want, and learning what the tech can do, and how to integrate it into their workflows.
I don't know if Predata will make it. Predicting the future for startups is really hard. I don't know if their tech works, or is needed, or is a good idea. It's possible that in all the chaos they are actually the first ones to find something no one else has discovered. Either way I am glad they exist, and I hope something good for the industry comes from their existence, regardless of where they actually end up as a company.
