The Dangers of Machine Learning

I am personally a huge fan of the British science fiction series Black Mirror , and every time I meet a hard-core technologist, I ask whether they watch the show. Wright doesn't have enough time to watch the series, but he has seen a few episodes. The reason I talk about Black Mirror is that we are living in a time when we are surrounded by the marvels of technology – machine learning, AI, Internet of Things (IoT), and whatnot – but the same technologies have almost become surveillance mechanisms, where the balance tips toward giving away more than is received in return. It's less about technology and more about business models and ethics. Would it be different if these technologies were developed by fully open source companies like Red Hat?

"It's not unique to now. There's a long history of powerful technologies being useful in really positive ways and also capable of having a negative impact on humanity," he said. "I think we always have to look at: How are we responsible with technology development? There is an interesting project called The Partnership on AI [1]; though Red Hat is not a member, I support the concept."

The project is trying to understand the social implications behind machine learning and data collection. "There are some potential, real negative ways that technology could be used and the implicit biases that can enter into models that we then assume are unbiased because they're generated by a machine," he said. "I think it's really important to understand the potential social impacts of technology that we invent."

That's where open source comes to the rescue. "What I like about open source is at least the whole technology development process is open, collaborative, visible, and transparent," he said, "but it still doesn't mean that the technology that we create couldn't be used in positive or negative ways."

The big challenge with machine learning is that it's not that much about framework and platform as it's about data. Most companies have open sourced their machine learning platforms, but the real value is in data collected by Microsoft, Google, Amazon, Uber, and so on. AI technologies feed on this data, and that data is not democratized, even if all the technologies consuming this data are.

"The open source development process creates technology that we understand; we know how to use it as code. That code may not be very useful without the data that's creating the data model," said Wright. "We have shifted that value from the code itself to the data. There is an interesting question about what data should be open and accessible. Especially when it's your data that's given to some service provider. A future vision that I would like to see is that I still retain a lot of control over my data – who has access to it; how it gets used. So you could imagine that there's a technology solution embedded in that problem, but first we have to recognize more at an industry level what it means to have all this value in data, and from an open source project point of view, we're going to keep building the technology. But it is true that some of that technology by itself is less useful without the data, which is providing a lot of the real value."

Other Areas of Interest

"There is a lot of working going on in and around containers and Kubernetes. Containers, container applications, and stuff that augment it is of interest to us – technologies that help us flush out the hybrid cloud story; it's about application portability. We also look at connectivity and storage," he said.

IoT is another area of interest, but Wright sees it more as an emerging market than an emerging technology. He looks at the set of use cases that employ technologies we already have. "We spend a lot of time in the NFV [network function virtualization] and telco space, looking at how that whole market is shifting to be fundamentally software driven. A lot of work is going on in preparation for 5G, which also has potential for edge computing and machine learning."

Conclusion

The takeaway from the interview, irrespective of whether you use Red Hat products or not, is that you should keep an eye on emerging technologies and markets. Try to see which are threats and which are opportunities; try to turn threats into opportunities, then bring those technologies to your customers in a way that they can invest their resources in adding value to their business instead of maintaining the stack – which is more or less a summary of Wright's job at Red Hat.