Bull Session
Human Intelligence Vs. AI
October 5, 2017
Episode Summary
On the podcast this week, we’ll discuss some of the areas where human intelligence may outperform AI and vice versa. AI is good at processing and discovering certain kinds of information — from data mining to predictive forecasting to optimization — while humans can add a layer of expertise, judgment, and insight. Join us as we talk about about how humans might work separately from as well as in conjunction with machines in an AI-driven future.
Resources:
Where human intelligence outperforms AI
I know I set this up as a human versus machine thing at the very beginning, but there’s also the possibility for machines to accentuate and enhance human thinking. One interesting way that’s being approached right now is by using crowd source solutions to discover areas of information that might not be readily available. So if you don’t have that information nicely formatted in data in your database, it’s going to be awfully difficult to search that data and to optimize that if you don’t have it in there at the moment. For items like research into particularly difficult areas, areas that require a lot of expertise, sometimes crowd sourcing can yield great results and information that may not be accessible to an AI, or a Google, for instance.
Dirk, as we see more and more of the wave of Artificial Intelligence in forming the way software is getting built, and the way that we are analyzing information, how do you see humans both working in conjunction with AI, and then also perhaps other areas that we think that human beings can sort of exclusively do better than Artificial Intelligence?
Where context in a broader sense can be brought to bear is in General Artificial Intelligence, which is also the strong AI. And General Artificial Intelligence right now is science fiction, so we are, you know, Chris Nelson was telling me last week that there was somebody did online an aggregation basically of all of these top scientists, engineers, people who would know, and said how far are we from General AI. The average, or the mean, or the, basically it’s 50 years. 50 years, some say faster, some say slower. So, 50 years is a long freaking way out, and so that still lives in the space of science fiction. And then of course the ultimate in Artificial Intelligence will be Super Artificial intelligence, that’s an Artificial Intelligence that is basically creating a being or maybe keeping it less charged I won’t call it a being, but creating a thing that is sort of demonstrably more powerful, more intelligent, is an improvement over humanity. And that’s total science fiction, right? I mean, that’s centuries not decades, whereas general is decades and weak AI is what we have today.
I’m kind of talking about all of this from the standpoint of talking about where humans are valuable. It’s in that we understand a broader context, we’re not just one tool. We get the world in a bigger way. A weak AI, a good weak AI even today is going to smoke a human in a very limited thing that allows it just to have blinders on and not know or pay attention to anything else. Otherwise, humans are infinitely more powerful and effective. In the article that you had, we didn’t go into the specifics of the kind of work that was being done there, but what was critical and important was that broader view, that broader focus where it’s not just “this is a search for the specific thing that is only going to be found in this way.” It was instead taking a certain environment as a starting point and looking for the unexpected, looking for more.
The way that humans and AI are going to work together is that AI, and this is again, in the short term weak AI where we are today, AI is providing a tool that can go very deep, very fast, very hard, in super narrow applications. And then humans can take that and build upon it. Yeah.
As I was thinking about that, the crowd source in it of itself, you know you’re using software, right, because you’re accessing this crowd via a platform, presumably, that allows a person or a company with a problem to define it and give it to not just one human mind but many human minds linked up by whatever this platform can do. So you can see whether that’s an advanced, very high-level mechanical turk type of program, versus feeding it into IBM lots and where a lot of already digitized data is available to be searched.
Those two models, the big pile of data versus the crowd source research, I think there are interesting ways that that could come together, depending on how much the crowd is able to digitize their discoveries as they’re doing it, versus doing all the analysis and synthesis in their heads and just giving the customer or a client the end results, right?
The question is then, are there more things that are being found that are thinking things, that are real knowledge work for those individuals to do, or is there just less work for that type of a person? And there’s just right now some unknowns there. I mean, for me at least, I haven’t studied it or thought about it enough, but it could go one of a couple of things at the extremes. We start to do a lot more thinking knowledge stuff, and we’re putting our minds to work in sort of exhilarating ways that have unpredictable and potentially wonderful impacts in the world. Or on the other extreme, there’s just not enough work and designers are going to have to go to, I was going to say slinging coffee or something, but those jobs are going to be gone too, and quickly. I don’t know where they’d go, but it won’t necessarily be the exciting tra-la-la of “Jeez we’re all using our big brains in bigger ways.”