I sat down with Steve Cowley, General Manager for IBM Watson, on Tuesday at IBM Insights to talk about Watson successes, challenges since the January launch, and what is in store. While the potential has always intrigued me, the initial use cases and message gave me more than a bit of pause: the daunting task to develop and train the corpus, the narrowness of the use cases, what would this actually cost? Jump ahead nine months and the IBM Watson world is in a very different place.
IBM is clearly in its market building phase. It is as much about what IBM Watson is and how IBM overall is repositioning itself as it is about changing the business model for selling technology. However, it is easy to get negative very fast on this strategy as seen with the tremors on Wall Street as IBM's stock has gone from a 52 week high of $199 to $164 at close on Friday 10/31, much of that happening in the past month since earnings release. Wall Street may not like company uncertainty during transitional periods, but enterprise architects care about what will make their organizations successful, make development and management of technology easier, and making sure costs don't sky rocket when new bright shiny objects come in. And, that is where IBM is headed with an eye toward changing the game.
IBM Watson delivers on information over technology.
Steve surprised me with this statement, "[With] traditional programmed systems, the system is at its best when it is deployed, because it is closest to the business need it was written for. Over time these systems get further and further away from the shifting business need and so either they fall in effectiveness, or require a great deal or maintenance." Steve pointed out that data is what is changing the game.*
This is interesting because in the last 30 years, artificial intelligence (AI) for the most part hasn't dramatically changed in what is available today (although we have yet to see the commercial results from research AI giants like Google, and the university labs). But, what has changed is the ecosystem to leverage AI – from the infrastructure to the data. In fact, when you look at Watson in particular, the value created is by having more and better data to feed insights. It is a system that craves experiences. The challenge for IBM is, if technology and algorithms are essentially all equal even if religiously argued, what makes IBM Watson stand out or will IBM define as the value to sell their broader technology portfolio?
Steve's point of view on that is the experience and value IBM Watson provides. In fact, where IBM Watson is getting traction is not in the bowels of CIO and enterprise architecture emerging technology watches. It is from CEOs and other top executives keen on reshaping their markets and business models. It has begun for IBM Watson in Healthcare where essentially leading cancer physicians extend their teachings from colleagues and students to include machines. Sloan Kittering and MD Anderson are scaling their experts through out physician networks trying to stop progression of cancer at the first visit, not the fifth. We see financial services making better sense of market conditions to guide investment decisions. And, we also see creativity in the culinary industry to determine recipes. And, Steve points to 5 CEOs initiating IBM Watson projects in the past month. This shows that IBM Watson is no longer a one trick pony but demonstrates how using artificial intelligence creates experiences as an expert (trained to answer), advisor (trained to suggest), and designer (trained to create).
There is value in this that you don't see in other AI implementations. Most are designed to deliver a single style of intelligence. The corpus and learning required for a single topic can still be daunting. But, IBM Watson does demonstrate that data, not technology, is at the heart because you train the systems to what is trusted and what will provide the most benefit. IBM is focusing this effort on "personalities" – who will be supported and how will they be supported. When lives are at stake, you need perfect answers. When investment decisions are required you work off less precise information. And when you create baseline knowledge is important but results are put into a more iterative process. The data has a continuum of trust.
Even if the underpinnings of these systems is complex, the experience of using IBM Watson is simple – deceptively so. There is still much hype into what IBM Watson can do. For example, if you want to use it for financial advise, it is only looking at financial conditions. It isn't taking into account the full range of investment profiles of those seeking advice. It doesn't incorporate wider socio-economic conditions. Essentially it is not yet thinking in multi-faceted ways across broader environmental experiences as humans do to make even the smallest decision. That is going to take time and require an evolution in artificial intelligence that combines several "thought" processes at once.
In the end, the success and failure for organizations will hinge on the training path chosen. The rules of information governance kick in to vet and classify the quality of data entering the system and tuning the systems for particular expectations of trust. That is an area that even outside of IBM Watson has been a significant barrier for organizations. We have designed systems for a very long time to serve A PURPOSE VS MANY PURPOSES.
We also have to consider what biases we introduce in the training process. If we rely on an expert in their field to vet results, how will we ensure it is not constrained by the expert's rules and ignore significant insights learned? Elon Musk has been quite vocal lately about the potential for AI catastrophe. The simple answer is to hard code in "moral" rules. But, at the same time, that can be a significant learning constraint. Better to control the action and not the learnings.
IBM strategy for purchase and adoption of IBM Watson is wise but risky
I asked Steve about IBM's go to market strategy and the building of the IBM Watson developer community. Here Steve said that IBM is not in the business of solely building IBM Watson solutions. IBM has taken this steps with Chef Watson and Watson Analytics to demonstrate the potential. However, the real potential he sees is that IBM Watson is a part of any and all technology solutions. We continually here about aspirations from companies to be the next "Intel Inside" with a graveyard of failures. This in someways is close to where IBM has moved – the infrastructure backbone and shaving off PC and printing businesses. But, the Watson in everything is clearly a much stronger step with the release of capabilities in BlueMix through the cloud, partnership building with Apple, and and the fact that in the past two weeks 1500 downloads and uses of Watson components have occurred.
For organizations looking to leverage IBM Watson immediately this means that it is the partners you will need to turn to. If you still want to go big and bad, for now IBM is there to partner, guide and deploy. In coming months as IBM Watson becomes more consumable this high touch engagement could very well move into their global services business as has been done elsewhere. Or, it could be primarily become a service through IBM partners. That is a risky bet if the task of consuming data and training systems becomes an onerous task when a corpus explodes to that multi-dimensional intelligence I mentioned above. The inside everything strategy is dependent on low touch ubiquity where the only thing a customer really knows is that IBM is the brand behind the brain. OEM is a precarious route for stakes this high, best to capitalize on IBM's current engaged position if the solution will do more than create your grocery list.
*Correction: Steve Cowley's comment clarified from original post.