Like the steam engine long ago, artificial intelligence represents an industrial revolution in its infancy. Economist Jared Franz discusses the recent tipping point for AI and the long-term implications for investors.
10 years of investment experience
Doug Mehagian: You talk about the concepts for the steam engine dating back to the Greeks and the Romans, which speaks to a long ramp up between the idea and the activation in a kind of a scalable way. So there's a nice parallel there to artificial intelligence. Talk about why 2013? What were the sort of green shoots you were seeing there that this was really an inflection point built on the promises of the past?
Jared Franz: As humans that have been thinking about AI, we have this intellectual baggage of . . . we see the movies. We see Arnold Schwarzenegger. We see, you know, the movies like “AI” — this generalized artificial intelligence. And that's really not anything like what we're actually doing in practice today with artificial intelligence. The practice is much different. It's more along the lines of efficiency.
Efficiency, prediction and pattern recognition — that's really what the artificial intelligence is doing right now. And what really created this environment today — and why it happened post the GFC, the global financial crisis — was these kind of megatrends that came together. The first was just the data, right? So, we are producing more data than we've ever had in history, really.
Doug Mehagian: So, that's kind of the feedstock for artificial intelligence.
Jared Franz: It's the feed stock. And people think, “Oh, it's just the numeric data.” No. We're talking about voice data, video data, audio data; numeric data is obviously part of that. But the amounts of data that we're producing are just exponentially larger than we've ever had in the past. And that creates that kind of rocket ship for AI. AI without data is essentially nothing. It doesn't help you to do anything.
The cost of processing that data is much lower than it ever was in the past. If you took the processing power within an iPad — let's say iPad 2, which was a 2010, kind of old even now — that processing power, which was about $100 in 2010, would have been about a trillion dollars in 1950. That's because of Moore's Law — the numbers have just become much cheaper to process this data. You also have the algorithms; now that they have the data, the algorithms get faster. The hardware is getting faster, and the costs of storing this data in the cloud and whatnot are becoming much cheaper.
And so those developments really came together in the post-crisis period. Even though we were talking about stagnation, we were talking about slow growth, underneath the surface, there were all these innovations that were starting to happen within AI. And they're not just in artificial intelligence. They're also in other areas like blockchain and bitcoin and other technologically interesting areas that will really have an impact, I think, over the next decade or even longer.
Doug Mehagian: So, based on past revolutions — we've spent some time talking about the steam engine — how might we expect this to unfold?
Jared Franz: That's the hard part. So, I think, when we think about revolutions, they're easy to [say,] kind of like, “Oh, historically, in retrospect, that was a big revolution,” whether it's microchips or steam engines or what have you. What we're not good at is predicting how these will evolve over time. And so, I think if I was alive when the steam engine came out, what companies would I have invested in, right?
Doug Mehagian: Yeah.
Jared Franz: And it was not intuitively obvious that you would invest A, B, C and D when the steam engine arrived, because you weren't sure how the technology was going to impact the economy in other areas. And so, caveat number one is that we're not that great at predicting the future and how these will evolve. Caveat number two is that we always think about the labor market disruption. So, we always talk about the bad aspects of technologies and kind of industrial revolutions and what not. But if you look at where we are today, in 1860 two-thirds of the U.S. employment was in agriculture, right? Now it's 9%. And we have more jobs than we've ever had before. And so, I think that when we look at the AI — this revolution that's coming forward — we should think about A, there's going to be jobs that are getting created because of AI and machine learning that are going to be helpful for the labor market. And B, it's going to take a while. It's not going to be something that's going to happen in 10 years.
I think what we're seeing today is just the first innings of what can happen across companies, across industries. And this could happen over two, three, four, five decades. And so, it's going to be a long time. Just like the steam engine, it took a while to even see the productivity improvement in the data; in actuality, it was very slow. With AI, machine learning, I think it's going to be similar.