Where's Watson? IBM’s Tarun Chopra on the company’s AI strategy: past, present and future

Watson’s Jeopardy! victory cemented both the system and IBM into artificial intelligence (AI) history. For the first time, a computer had unequivocally bested the human brain in the realm of accurately responding to difficult questions. And it wasn’t the first time that an AI-powered computer system designed by IBM had outperformed humans in one of our most cherished intellectual domains: In 1997, the company’s Deep Blue supercomputer defeated chess grandmaster Garry Kasparov.

Around two years after Watson's historic Jeopardy! win, New York City’s Memorial Sloan Kettering Cancer Center integrated the computer system to help doctors make more accurate diagnoses and recommend more personalized treatment plans — making Watson the first AI model with a commercial application.

IBM, in other words, was quick to establish itself as a pioneering force in AI; the much-publicized Deep Blue and Watson moments were akin to moon landings for the then still-burgeoning field. But where has the company been since then?

Since the November 2022 release of ChatGPT — an AI model developed by OpenAI designed to respond to text-based prompts in natural language — the media landscape has been afire with chatter about AI. Every week, it seems, a new AI-powered product is released or a new dire warning is issued by experts, rekindling public interest in (and fears about) this rapidly advancing technology. We’re currently living through the early stages of what’s been referred to — perhaps hyperbolically, perhaps not — as the “AI revolution.”

Google, Microsoft, Meta and other tech giants are spending exorbitant amounts of money working to establish their footholds in the burgeoning AI industry. But where is IBM?

To find out, The Drum spoke with Tarun Chopra, IBM’s Vice President of Product Management, Data and AI.

It’s been 13 years since Watson was launched, and 10 years since the computer system was integrated into its first commercial use (by New York City’s Memorial Sloan Kettering Cancer Center). What has IBM’s AI strategy looked like since then? Where has the company mainly been focusing its efforts?

Watson was based on the techniques of machine learning—you feed the data, you build the model, so forth and so on. Then, in typical corporate fashion, we worked with thousands of clients to deploy AI into the [business] landscape. And we learned a ton. It’s not that easy to productize these capabilities in large-scale enterprise environments, for simple reasons: scales, corporate regulations, corporate mandate, privatization, return on investment, cost, a lot of factors go into mainstream, large-scale production.

The journey was fascinating; it moved at a pace that we didn't expect. And that's why we went into what we used to call ‘Trusted AI’ or ‘AI Governance,’ because customers were telling us, ‘Look, if I'm going to prioritize [AI], I need to be able to explain it to my shareholders — it can’t be a black box thing. So a lot of innovation took place with respect to the explainability of AI. A lot of IBM’s effort went into that, because we learned that just creating something and putting it out there isn’t going to work; customers need a lot of different tools in order to be able to productively deploy [AI] into their business models. So we were helping to productize AI — that didn’t really made the headlines, because it was not sexy. It was more like rolling up your sleeves and putting in the work to help customers adopt AI.

The other thing we learned is that you can’t just offer a bespoke AI solution for everybody — you need applications that have mass applicability in order to drive adoption.

OpenAI’s release of ChatGPT has in many important respects changed the ways in which most laypeople think about AI. How has IBM been positioning its AI products and services since then?

What ChatGPT really highlighted was the value of this new technique called foundation models, which allow you to apply whole models to other models, rather than just having to rely solely on data for models. This opens to the door to a more mainstream audience, because now you don't have to be an expert to use this technology; you just need to submit a prompt ... But most businesses aren’t going to adopt this technology because it’s still a giant black box. This poses an opportunity for IBM: we have 10 years of experience learning how to productize AI into enterprises, and now we can apply that to help clients adopt foundation models.

It takes a lot of work, skill and governance to help big enterprise clients adopt and scale AI. The promise the world is talking about right now is the mass-scaling of AI. After you take away all the hype, we still have a long way to go before we reach that goal. But I think IBM’s learnings over the past 10 years or so is going to be very beneficial for us, and could help to move the needle towards mass-scale.

In 2011, IBM made history when its Watson computer system defeated two top (human) Jeopardy! champions. As someone who works closely with AI, what did that moment mean to you?

One thing it highlighted for me was: what was new today is old tomorrow. It's fun to be in this field if you like that rapid pace of [technological] change.

It was also fascinating to keep seeing customer interest, because sometimes it’s tough to envision what one can really do with [a new technology]. When the internet was invented, people didn’t really know what to do with it before Netscape appeared. The same thing is now happening with AI, first with the Watson Jeopardy! moment, then with [Google’s] Deep Blue chess moment and now with the ChatGPT moment—now, a common consumer can think, ‘You can do this with this, or you can do that with that.’ But unless you really figure out your killer apps and know how you're going to help consumers uses these capabilities, it's tough to have mass adoption.

So to me, those [historic AI] moments again kept reflecting customer interest, but the same question keeps popping up: wow can you then bring that [technology] into an enterprise’s day-to-day environment? It's a very different task.

Recent advancements in AI have been sending shockwaves of anxiety — mainly about job displacement and existential risk — throughout our society. Are there any misconceptions about AI and its perceived risks that you commonly encounter? How do you think about the challenge of a brand like IBM communicating about AI in a manner which presents it not as something to be feared but as something that can legitimately benefit humanity?

To me, AI has huge benefits for our society. It’s really predicated on the business problems you're trying to solve. Can you infuse AI to help the human race become more productive? In the end, it's all about productivity. I believe in the power of AI to augment human capabilities and knowledge. Look at my own field: You can use AI to write better software. It’s not like tomorrow robots are just going to write the software and we won’t need any [human] programmers. It's more about asking, ‘How can I improve the overall quality and productivity of what we’re doing [using AI].’

If used correctly, and if we communicate in the right way to our consumers, to our government and to our marketplace, AI can become an augmentation of what we do today. All technologies throughout history are really about augmenting [human capabilities]. Right? AI is the next big evolution in that augmentation. That's how I view it, and that’s based on what I see happening in the marketplace today.

What do you consider to be the biggest opportunities and challenges that AI currently presents to marketers?

I think the biggest opportunity is integrated storytelling, in which [marketers] can leverage AI to bring inputs from hundreds of different sources, which they were not able to do before.

The big challenge for marketers is knowing how to distinguish fact from fiction, and not just believing all of the hype. They need to be able to do their own homework and know their facts. They must also be wary of promoting misinformation through AI. When an AI model starts to hallucinate, your company might do the same thing. The more your digital space gets flooded with AI, the more challenging it will become to separate fact from fiction. That’s where governance becomes very interesting, and very important.

What’s next for IBM’s AI research?

ChatGPT has brought the capabilities of foundation models to the world. Over the next 18 to 24 months, I expect us [at IBM] to hone in on those capabilities and those models, which can be applied to various industries. Large language models are just one flavor of foundation models. We’re working with various entities — including NASA and chemical companies — to expand the horizons and capabilities of foundation models. We’re also going to work with our clients to help them understand, govern, productize and cost-effectively deploy the capabilities of foundation models.

I gave you a short time horizon because the [AI] space is moving so fast; who knows what it will look like in ten years. But from a product leader perspective, those are the [projects] that I look forward to working on, while also — as I’ve said — leveraging the 10 years of learning we’ve acquired from Watson with respect to understanding what customers really need in order to productize these capabilities in their environments.

For more on the latest happening in AI, web3 and other cutting-edge technologies, sign up for The Emerging Tech Briefing newsletter here.