Why the surge in machine learning? | Interview with Robbie Allen (Part 3)

DISCLAIMER: Below is an AI generated transcript. There could be a few typos but it should be at least 90% accurate. Watch video or listen to the podcast for the full experience!

Alexander Ferguson 0:00
Welcome to UpTech Report series on artificial intelligence. I’m Alexander Ferguson. This video is part of our deep dive interview series, where we share the wealth of knowledge by one of our experts in the field of artificial intelligence. This is the third part of my conversation with Robbie Allen, CEO of Infinia ml in Durham, North Carolina. Among his other business accomplishments, Robbie owns six patents and has authored eight books. Here, I ask why there’s been a surge in machine learning why data means everything, and how the AI community is different from the traditional software arena,

Robbie Allen 0:36
there’s really three things that ushered in this machine learning wave that we’re seeing. The first is data, which I mentioned before, and often say that the big data era, which you know, really kind of started around 2005, or six, you started hearing the term big data, you know, that was coined around that time, companies really started to take data collection more seriously, they started thinking about data as a competitive advantage. And so they at least started collecting it. And so that was necessary, because as I mentioned, machine learning is all about data. If you don’t have data, you can’t do machine learning. And so at least many companies had the data, there was GPUs, which you know, really is a special form of processor, you know, most computers, laptops have what they refer to a CPU or central processing units, GPUs or graphical processing units, they’re the things that oftentimes will, you know, power your monitor, or power gaming consoles. And in fact, GPUs became more popular due to the rise of gaming consoles, so often, you know, say that we, we have gamers to think and some part for, you know, some of the advances that we’ve seen with machine learning. And then there’s been advancements in the algorithms themselves, you know, specifically around deep learning. You know, there’s been just, there’s so much research going in now, in the machine learning community, and so many improvements that have been made, especially that are deep learning based that now, thanks to the processing, thanks to the data, we can apply more complex versions of machine learning, that can do all sorts of interesting things. I don’t think it’s a big jump to go from, you know, really basic statistical techniques to machine learning, at least the maybe the more straightforward techniques. And in fact, it really kind of depends on what the data looks like. Because the complexity of the data also oftentimes dictate the complexity of the algorithm or the things that you kind of have to keep an eye on. And that’s why deep learning, that’s really the step function of complexity in terms of, you know, it’s very difficult to go from, you know, basic regression to deep learning, without having, you know, more formal training or a lot more experience, because there’s more things that could go wrong. And if you’re not, you don’t have an awareness or an ability to understand where things went wrong, or why they went wrong, then, you know, you won’t be able to understand or interpret the results that you get. So I would say, it’s not too much of a stretch, to go to basic machine learning techniques. But once you want to get to something like deep learning that requires more of an understanding,

Alexander Ferguson 3:01
why is data so important to machine learning?

Robbie Allen 3:04
Machine learning, as I mentioned, is learning patterns in data. And if the data has garbage in it, or if it’s very noisy, or there’s problems, especially ones that you don’t anticipate, or you can’t see yourself, then the algorithm is going to learn those those, you know, learn that noise, or it’s going to start making misjudgments because it’s learned patterns that were incorrect. And so if the data is not in pretty good shape, you’re gonna have problems.

Alexander Ferguson 3:28
What are some of the challenges with machine learning?

Robbie Allen 3:31
One of the big challenges that many companies face is well, and even I think companies like mine, that are trying to really focus in on a couple of key use cases is we go and talk to companies, and we’re talking to, you know, some of the largest companies in the world all the way down to small product companies. And everybody has a different use case. And the reason for that is we’re so early on in the machine learning journey, that there’s not just one or two applications of it, in fact, you go into a large company, and there’s dozens of applications of it. And so right now, because people are just getting started, no two companies tend to have the exact same need with the exact same priority. Because machine learning is a real sort of general purpose tool that you can apply to lots of problems. So anyway, there’s that, but we’re seeing, you know, all sorts of interesting use cases for things like all sorts of document processing type of applications, whether that’s contract analysis, where you can automatically, you know, read a contract a legal document, and determine what kind of contract it is, you know, understand what kind of issues are in the contract, do what they refer to as q&a question and answering about the contract. So does this contract have a change of ownership clause in it? And it could tell you yes or no. You know, another one that you probably heard about his resume analysis. So can you automatically scan a resume and then tell me if that person is going to be potentially a good employee for me or not? You know, and there’s been some actually failed examples of that. In the past, you know, that have kind of injected bias. And this is a common thing that I get asked about all the time. Well, are we are you worried about the bias of the algorithms? And what I tell people is machine learning starts at neutral, right? Like there’s nothing biased about the algorithms. What’s biased is the data that it trains on. And so really, if it’s making a bias decision, it’s just reflecting what it was trained on, likely biases that were built into the data that it would that was collected. And so I think it’s overall a net positive, even with the bias built in, because at least we’re at a point where we can recognize the bias, people are talking about the bias, and we can do something about it, versus you know, how many years dozens, you know, decades, where people have been making these decisions, they never talked about bias, but guess what the bias was present, they just didn’t talk about it, we often say to that we you know, we’re not gonna apply machine learning to every problem. In fact, there’s many problems, there’s lots of low hanging fruit, in fact, inside of the the corporate world, that now that they have the data, you don’t need something like machine, you don’t need the big hammer of machine learning to go solve it, you can use simple statistical techniques, and get a lot of mileage out of them that way. So it’s not always the case that you need machine learning, there is a, you know, a very large set of statistical techniques that can do all sorts of processing and analysis, whether it’s regression or classification, that doesn’t require more complex, complicated techniques.

Alexander Ferguson 6:23
How is the AI community different than traditional software?

Robbie Allen 6:28
The great thing about artificial intelligence community is its academic base, which means, you know, most of you know, the great advances have been made or open and, you know, freely available. It’s not really, you know, kind of buried in patents for the most part. And so that’s one of the nice things about it. It’s not like a closed community, it’s very open here, more and more about companies trying to file patents, the USPTO recently came out with new guidance on filing AI oriented patents that will make it a little bit more difficult to do that. Again, the nice thing is that the credibility in the AI space comes with published papers, not published patents. And that’s another fundamental difference from traditional software, which, you know, most of the badges of honor were around how many patents Did you file? Now? It’s how many research papers have you published in my company? You know, we published a number of papers, Dr. Larry Caron, our chief scientist at one of the most prolific machine learning researchers in the world.

Alexander Ferguson 7:22
Lastly, I asked Robbie Allen for his final thoughts on the age of artificial intelligence.

Robbie Allen 7:27
You know, I think it’s super exciting. You know, I started my career During the.com, boom, and bust. And I always thought, you know, I was, you know, early 20s, during the late 90s, and early 2000s. And it was something a little sad about the fact that I thought maybe the most exciting point of my career happened when I was 21 years old. But now, I would revise that and think that now, potentially, will be the most exciting time of my career. This will be the time that, you know, one day I’ll tell you know, my grandkids will ask me about what was it like to be in the early days of, you know, the artificial intelligence revolution. And, and I do believe we’re just at the very beginning of that, so it’s a machine learning is while there may be, you know, some ups and downs as there is with any technology, I think overall, it’s a long term winner.

Alexander Ferguson 8:12
That concludes the audio version of this episode. To see the original and more visit our UpTech Report, YouTube channel. If you know a tech company, we should interview, you can nominate them at UpTech report.com. Or if you just prefer to listen, make sure you subscribe to this series on Apple podcasts, Spotify or your favorite podcasting app.