In early 2011, Ken Jennings seemed like humanity’s last hope.Watson, an artificial intelligence created by tech giant IBM, made fewer choices dangerous A player before the show’s All-Time Champion participates in a three-day exhibition match. At the end of the first game, Watson (a machine the size of 10 refrigerators) tied Jennings to the ropes to lead $35,734 to $4,800. On the third day, Watson finished his work. “I welcome our new computer overlord,” Jennings wrote on the Final Jeopardy video screen.
Watson did a better job than any AI to date at addressing a problem that has long puzzled researchers. In other words, how do you get a computer to correctly understand the clues presented in idiomatic English and spat out the correct answer (or dangerous, correct question)? Watson’s lead developer, David Ferrucci, told me, “It’s not a hit list of documents that might have the answer.” His team gave Watson over 200 million pages of documents, including dictionaries, encyclopedias, novels, plays, and the Bible, and created something that looked like a synthetic brain. And America went crazy about it. NPR asked in an article “Watson’s Dark Side.”4 months later dangerous This computer was named Person of the Year at the Webby Awards. (Watson’s acceptance speech: “Person of the Year: Ironic.”)
But now that people are once again facing doubts about seemingly omnipotent AI, Watson is clearly absent. When I asked Benedict Evans about Watson, he quoted Obi-Wan Kenobi. ChatGPT and other new generative AI tools can deliver pastiche poetry and popes wearing Balenciaga. These capabilities are far beyond what Watson was able to do a decade ago, but they are still based on the natural language processing ideas that took over Jennings. Watson needs to boast with a haughty voice instead of fading into irrelevance. But the trajectory is repeated over and over again. Some of what this technology was destined for is now sapping the potential of popular AI products.
The first thing to know about Watson is that he’s not dead. Machine models and algorithms are built into the body of B2B software. IBM now sells his Watson by subscription and embeds the code into applications such as Watson Assistant, Watson Orchestrate and Watson Discovery. These applications help automate back-end processes in customer service, human resources, and document entry and analysis. Companies like Honda, Siemens, and CVS Health have reached out to ‘Big Blue’ for help with AI in many of their automation projects. An IBM spokesperson said the company’s Watson tool is used by more than 100 million of his people. If he asks IBM to build an app that uses machine learning to optimize business, he says, “IBM would be happy to build it, and it would probably be perfectly fine.” He said Evans.
From the beginning, IBM wanted to turn Watson into a business tool.After all, this is IBM— International Business Machines Corporation — Company that pioneered a niche market for large enterprises that need IT assistance. But Watson has become much more modest than IBM’s initial pitches, such as unlocking machine fact-finding capabilities on topics as diverse as stock tips and personalized cancer treatments. And to remind people of how innovative Watson was, IBM put out a television commercial in which Watson cheerfully teases celebrities such as Ridley Scott and Serena Williams. The company soon landed his AI-centric deals with hospitals like Memorial Sloan Kettering and his MD Anderson Cancer Center. They were slowly established.Watson played by a machine dangerous At a very high level; Watson, the digital assistant, was basically a giant clippy ruled by corporate data and tech optimism that, let alone disrupting oncology, could barely read the doctor’s handwriting. .
Technique did not measure. “There was no intelligence there,” Evans said. Watson’s machine learning models were very advanced in his 2011, but they didn’t compare to bots like his ChatGPT, which incorporated much of what was published online. Watson was trained with far less information and was only good at answering factual questions like the ones you find on websites. dangerousThat talent included clear commercial potential, at least in certain areas such as searching. “I think what Watson was good at at the time turned into what Google is doing,” he said.
But the suit in charge pursued the larger, more technically challenging game of feeding machines with an entirely different kind of material. “There was a lot of exaggeration about it and a lot of lack of awareness of what it really could and could not do and ultimately what it took to effectively solve the business problem,” said Ferrucci. says Mr. He left his IBM in 2012 and has since founded his AI startup called Elemental Cognition.
An IBM spokesperson pointed to CEO Arving Krishna’s recent statement when asked what went wrong. Kareem Yusuf, head of product management for IBM’s software portfolio, told me that Watson is a “concept he’s a car.”
But to others, it may have seemed that IBM was more concerned with building a flashy convertible showroom than thinking about how to design next year’s model. Part of IBM’s problem was structural. Richer and more agile companies such as Google, Facebook and even his Uber were driving the most relevant AI research, developing their own algorithms and threading them into everyday software. “If you were a cutting-edge machine learning scientist,” Evans said. Company since the 70’s. By the mid-2010s, he told me, Google and Facebook were leading the pack in machine learning research and development and were betting big on his AI startups such as DeepMind. Meanwhile, IBM was producing his 90-second Academy Award spot featuring the voices of Watson, Carrie Fisher, and Steve Buscemi.
In a way, IBM’s vision for a suite of business tools built around machine learning and natural language processing has come true, but not thanks to IBM. AI is now powering search results, assembling news feeds, and alerting banks to possible fraud. Rosanne Liu, a senior research scientist at Google and co-founder of the research nonprofit ML Collective, said it rumbles behind “everything you deal with every day.” This AI moment has further fueled the corporate buzz over automation as every company wants its own bot.
Watson has been reduced to a historic footnote, but IBM is still in the action. The most advanced AI efforts aren’t made at IBM’s Westchester, NY headquarters, but many of them are open source and short-lived. Tailoring Silicon Valley hand-me-downs can be a profitable business. Yusuf summons a platoon of knowledge workers armed with 20th century tools. “There are people with PDFs and highlighters,” he said. IBM can provide programs that help you do your job better, whether it’s increasing productivity by a few points, lowering error rates, or finding problems like faults in production lines or cracks in bridges more quickly. increase.
Whatever IBM does next will not deliver on the promise implied in Watson’s early run, but that promise was in many ways misinterpreted. First and foremost, by IBM. Watson was a demo model that could garner a great deal of public interest, but that potential was blown away as soon as management tried to plug in the money. The same seems to apply to his new AI tools.Even high school students can make it another peace Sure, I write essays in the voice of Mitch Hedberg, but that’s not where the money is. As a result, run-of-the-mill consumer and enterprise software (the ability to find pictures of your dog or sell slightly better kibble) is just like all other data we passively consume. In fact, it can become invisible to us. In March, Salesforce introduced his Einstein GPT. This is a product that uses OpenAI technology to draft sales emails. It’s part of a trend Evans recently described as “boring automation of boring processes in boring back offices of boring companies.” A big name tied to a humble purpose, Watson’s legacy is unfolding once again.
The future of AI may prove to be truly world-changing, as Watson once suggested. But the only business IBM has successfully disrupted is its own business. On Monday, International Workers Day, it announced it would suspend the hiring of about 7,800 jobs that AI could do in the next few years. Freeing thousands of jobs in the name of cost-cutting measures rarely sounds so optimistic, but after years of positive spins, why pull back now? Yusuf asserts that IBM’s future is just around the corner, but not this time. “Look at this space,” he said.
