It is an exciting time for AI. There is now an extremely wide array of both customer and enterprise-focused products, with major advancements having taken place over the last few years. And it is only accelerating.
AI is, therefore, clearly an extremely attractive investment. As of 2017, over $70 billion has been invested in autonomous vehicles alone. Once you include cybersecurity, healthcare, and augmented reality, that number more than doubles to $150 billion. Speaking at the Machine Learning Innovation Summit in New York, John Frankel, Founding Partner at ff Venture Capital, explained why this is likely merely the beginning, and discussed what investors who are considering moving into the space need to be on the lookout for.
The first thing Frankel notes is that not all AI is equal. 'There are certain areas within AI which are just not investible and other areas worth chasing and spending time on' explains Frankel. 'AI is a fascinating tech stack and I think one of the things to understand what happens is, when you get these architectural shifts, is each one sits on top of the other and the other ones just sort of blends into the background.'
The issue is that machine learning is progressing so quickly that innovations that would have been hailed as the mind-blowing just a few years ago are being met with apathy today. Frankel uses the example of the Babel fish to explain this point. The Babel fish is a creature from the book 'Hitchhikers Guide To The Galaxy' which, once inserted in an ear, 'would allow you to instantly understand anything said to you in any form of language.' 'When I was a kid,' Frankel says, 'this was the most magical thing ever! Could you imagine a world where you put this in your ear and it translates things instantly, could travel the world, wouldn't have to learn languages...and now, its a couple hundred bucks and no one cares.'
This is good in some ways, but also makes it more complicated to understand the tech that will be investible. Exciting innovations may thrill, but could have little commercial future. We are becoming increasingly accustomed to AI in our daily lives and you are looking for something that will slot into the routine. For example, with AI in customer service, 'when you call up a utility and it's an AI that answers, you actually go 'oh, well that's a relief.' Frankel says, '...people are willing to tell Wendy [an AI recruitment bot] more about their careers than they are willing to tell a human. They feel Wendy is less judgmental, and they are having 30-40 minute long conversations.'
Carl Sagan once said 'any advanced enough technology will be indistinguishable from magic'. And as with magic, there is an element of faith involved in the evolution of AI. While most of us are consciously aware that investment in the technology is a 'faith-based' strategy, though, it still affects our decisions because they are clouded by our excitement at the new. As Frankel explains, you have 'to be careful because it's easy to overstate the speed of development'.
The problem is, the bar has been raised far too high. Like the Babel fish/Google Pixel Buds example, consumer AI has advanced so far, it has overtaken human capabilities many times over. Microsoft DeepCoder is now autonomously writing code, Google AutoML is AI writing AI and, DeepMind can accurately read lips better than any human being. Even previous projects are being dwarfed by new innovations. For example, AlphaGo Zero's domination of AlphaGo. 'AlphaGo Zero didn't learn how to play Go from humans' Frankel explains, 'they just gave it the rule-set and it played against itself. And it not only conquered [Alpha]Go in, I think, in 3 days, using deep learning and reinforcement learning, but it also conquered chess as well...The team that developed it said its moves seemed completely alien to them, it would sacrifice high-value pieces in ways that a human simply would not do.'
Acceptance of a technology breeds apathy amongst the masses, though. We are becoming jaded by it. People can only be impressed by so much and it has made it even harder to predict what will impress even a few years from now. So knowing what to invest based on what will be seen as groundbreaking in a few years can be near impossible.
So it helps to categorize the spaces AI is currently being developed within:
2. Enterprise (internal process)
3. Enterprise (customer facing)
4. Horizontal stacks
Of the 4, Frankel believes the most logical focus for those interested in investing in AI is enterprise. '...Consumer, you are challenged by large tech companies... but if you can leverage deep learning layers of data within vertical solutions, that's where we still think investment opportunities still exist.' He further writes 'significant opportunities exist in enterprise solutions that are vertical, that have deep pools of domain-specific data, and where the AI can compound this advantage over time.'
Either way, there are still a multitude of questions and actions that need to be considered in order to successfully enter the market. Never before has an architectural technology shift been so consequential and so rapid. Eventually, the hype will subside and deflation of the market will begin, but this is when we will likely move away from the novelties that are creating a lot of the hype right now and move on to more consequential innovations. So I think Frankel phrases the most important piece of advice I could give perfectly as thus:
'Do or die. And do very quickly'
To learn more about the future of AI and machine learning, join us at the Machine Learning Innovation Summit in San Francisco on 9-10 May
BONUS CONTENT: Why AI is Investible—and Why Money Will Be Lost