martes, 1 de agosto de 2017

The hard part is getting it (the data) organized and figuring out what’s relevant to your process.

Robo-Advisors Aren't A Difficult Tech Problem And Are Already On Their Way To Being Commoditized

Robin: Well, the buzzwords, I’ve noticed, are very much in the financial industry… Is machine learning. But, basically, using more modern artificial intelligence techniques in everything from investing to solving problems that they have on the consumer side, [inaudible 01:01:00]
How much of this is, to return to the cool kids trying to sound with the times, [inaudible 01:01:05] machine learning is very much [inaudible 01:01:07], but do you think, is this a future we’re gonna see more things move that way?
Joe: Yeah. That one, I’m more skeptical of a lot of the efforts I’ve seen in Wall Street. So it might be a little bit more of the cool kids syndrome. I’ve had several CIOs of big institutions ping me saying they have a big machine learning project, and can I connect them to the right people for what they are doing recently.
I mean, machine learning is a tool that all the great technology companies are using now to get done certain things. When you have a bunch of data and you need to figure out what the best answer is, I think you apply it.
I mean, it’s kind of funny, right? It’s artificial intelligence, and once it works we call it machine learning, and there’s all this research on the AI front. I’m definitely bullish on what AI is making possible, but I don’t think it’s this panache or this thing that solves everything. I think the harder problem, frankly, is getting all of your data organized in an infrastructure where you could use it for the problems.
So I think the hard thing to say is, what data should we be using to solve this problem? What is that process? Then once you have all of the data organized, you can hire a single Ph.D. from Stanford. He will do your machine learning thing for you. That’s not that hard. The hard part is getting it organized and figuring out that it’s relevant to this process.
So that’s much more of a strategy and process question to me.
Robin: But you think that it’s gonna be more, I mean, on artificial intelligence. I like that definition, that artificial intelligence is everything we haven’t been able to do that. Everything else is [inaudible 01:02:21]
But how much will we be able to do in the future? I mean, looking forward a little bit.
Joe: Sure.
So I definitely have a few investments in artificial intelligence companies that are trying to push the boundary and change things. I mean, I was actually really surprised that Demis and the team at DeepMind were able to solve the Go problem.
I think that was pretty impressive. But even looking into it, it’s not really capturing, in my view, human creativity, or anything even close to that yet. So I think we’re still decades and decades away from needing to worry about being replaced.
I think the much more interesting problem that we’re focusing on kind of the next 10 or 20 years is, how do we organize the infrastructure of our companies to get the data structured and available to use it for decisions?
I think it’s not gonna be the people who are best at AI and machine learning who win. It’s gonna be the people who are best at that data infrastructure and process problem. I mean, most of the top companies in Silicon Valley, like linear regressions and things, get you 90% of the way. I’m not seeing a lot of really, really hard AI that makes the big difference, for at least the things I’m seeing.

No hay comentarios: