Just as the human organism can slowly self-heal, so do new artificial-intelligence models. AI can be trained to watch and fix its own problems. Undoubtedly, there are many dangers inherent in AI algorithms — from self-driving commercial truck fleets, trading and financial predictions, or even health decisions, the risk is there. So how are companies dealing with these unintended consequences?

“We use AI to modify the data, to train the models, and then we set some of the models up as self-training,” said Joe Selle (pictured, right), global advanced analytics and cognitive lead at IBM. “So they’re learning on their own. They’re enhancing data sets. And once we turn them on, we can go to sleep so they do their own thing. You could think of it as an AI watching over the other AI.” 

Selle and Jerry Gupta (pictured left), senior vice president and digital catalyst at the Swiss Re Group, spoke with Dave Vellante (@dvellante), host of theCUBE, SiliconANGLE Media’s mobile livestreaming studio, during the IBM Chief Data Officer Summit in San Francisco. Swiss Re, a reinsurance and insurance company, has implemented IBM’s Watson cognitive-computing technologies, zSecure, and analytics accelerator. Selle and Gupta discussed new data themes, solving the inherent algorithm problems in AI, and real-life applications (see the full interview with transcript here). (* Disclosure below.)

[Editor’s note: The following answers have been condensed for clarity.]

Vellante: The [data] themes over the last couple of years, how have they changed? 

Gupta: So two years ago, when I first attended this conference, it was about cyber, and when we went out in the market, there were not too many companies talking about cyber. And so you come to a place like this, and you’re sort of blown away by the depth of knowledge that IBM has. And that really helped us to gain information about the cyberrisk .. and evolve a little bit. 

Vellante: Joe … how do you scale this [event] globally? 

Selle: The themes now are around AI at scale. How do you do that? We have a remit at IBM to get 100,000 IBMers … on our Cognitive Enterprise data platform by the end of this calendar [2019] year, and we’re making great progress there. But … the key question [is], how do you do that? And it involves cultural issues of teams and business process owners being willing to share the data, which is really key. And it also involves technical issues around cloud-computing models, hybrid public and private clouds, and multicloud environments. 

Vellante: You’ve got machine intelligence or AI that you now can apply to that data. And then for scale, you’ve got cloud. Does that sort of innovation cocktail … make sense in your business? 

Gupta: From the perspective of my company, it’s not the shiny, new toy that’s attractive or that really moves the needle for us. It is the underlying risk. And Joe alluded to AI and algorithms and stuff. And it’s clearly starting to become a very big topic globally. Even people are starting to talk about the risks and dangers inherent in algorithms and AI. And, for us, that’s an opportunity that we need to study more, look into deeply to see if this is something that we can help address and solve.  

We look at not just blind spots — risks that we don’t even know about that we are facing — but also look at known risks. 

Vellante: So you got the potentially unintended consequences of algorithms. So how do you address that? 

Selle: So the name of the game is really understanding how as we all move towards a digital enterprise with big data streaming in, in every format. So we use AI to modify the data, to train the models, and then we set some of the models up as self-training. So they’re learning on their own. They’re enhancing data sets. And once we turn them on, we can go to sleep so they do their own thing. 

Then what? We need a way to understand how these models are producing results. Are they results that we agree with? So we want to monitor, understand and remediate, but it’s at scale.

Vellante: Jerry, you mentioned autonomous vehicles before that’s obviously a potential disruptor to your business. What can you share about how you guys are thinking about that?

Gupta: So, the journey towards full automation is a series of continuous steps, right? So it’s a continuum and to a certain extent. We are in a space now where even though we may not have full autonomy while we’re driving, there is significant feedback and signals that a car provides and acts on in an automated manner that eventually move us towards full autonomy. And so the question then becomes: Where does liability lie — owner, manufacturer or driver? 

Vellante: You guys are doing a panel on making customers the king. What’s that all about? 

Selle: We’re using some of our AI to create an alert system that we call Operations Risk Insights. We’ve been giving this away to nonprofit relief agencies who can deploy it around a geo-fenced area. And if you’re a relief agency providing flood relief or services to people affected by floods, you can use our solution to understand the magnitude and the potential damage impact from a storm. And we use AI to do that. Those are happy customers, and I’ve delivered “wow” to them. 

Watch the complete video interview below, and be sure to check out more of SiliconANGLE’s and theCUBE’s coverage of the IBM Chief Data Officer Summit. (* Disclosure: TheCUBE is a paid media partner for the IBM Chief Data Officer Summit. Neither IBM, the sponsor for theCUBE’s event coverage, nor other sponsors have editorial control over content on theCUBE or SiliconANGLE.)

Photo: SiliconANGLE