A Fanciful Explanation of the Expensive Failure of IBM Watson
February 19, 2020
I love the idea of revisionist history. I associate the method with an individual named Ioseb Besarionis dze Jughashvili.
Alleged Stalin quote: It is not heroes that make history, but history that makes heroes.
You may know this allegedly competent leader as Joseph Stalin. Changing history is one way to make sure the present comes out in a way that is more satisfying — at least to some people.
I read “IBM Watson And The Value Of Open.” I thought of Jughashvili in the terms my former history professor (Dr. Philip Miller Crane) explained the revisionist thing.
My analysis of IBM Watson included information I obtained when I was researching my various and sundry books about search and retrieval. I did not include IBM as a “recommended” solution for three reasons:
- Watson was a marketing confection which conflated a range of technologies: Some developed by IBM and others obtained via an open source download or by paying money for technology; for example, Vivisimo, a metasearch and clustering system
- Training “Watson” required programmers to interview subject matter experts, create specific content domains, test, do more interviews, retrain, and test. Once the content domain was in hand, Watson would crunch away to locate an answer. Many companies do a similar expensive process. IBM was unique in making Watson seem something other than what other vendors offered. By sweeping the time and cost of training under the digital rug, Watson was cut loose from reality.
- Question answering systems work when certain conditions are met; for example, content, response expected, handcrafted rules that mostly work. Toss the system questions based on new content, and the responses are going to be interesting if not off base a certain percentage of the time.
To sum up, the cost and unreliability of Watson were wildly out of step with the marketing of cognitive computing. IBM’s billions made it possible for search and retrieval carpetbaggers to describe their systems as “cognitive”; that is, infused with artificial intelligence, predictive linguistics, and my favorite bit of jargon natural language processing.
The article’s explanation of the failure of IBM’s billion dollar bet, the office near NYU, and the absolutely bonkers ad in the New York Times for Watson as a collection of digital molecules is at odds with my assessment.
That’s okay. Let’s look at a couple of the “revelations” in this Forbes’ article.
The Texas Fold
The write up explains the outright failure of Watson as a useful medical tool for cancer doctors says:
But with the passage of more time, it must be said that IBM Watson has not delivered the results that IBM expected. One particular moment was the decision of MD Anderson’s Cancer Center to withdraw from its partnership with IBM in 2017. An internal audit by the University of Texas found that the university had spent over $62 million dollars (not counting internal staff time) and did not meet its goals.[i] Other health partners soon followed.
Yep, to summarize. Watson did not work. In fact, I heard from a reliable source that cancer doctors in New York City refused to answer endless programmer questions. The message for me was, “Cancer doctors don’t want to teach programmers how to be cancer doctors.” Hasta la vista to Texas.
The Wrong Explanation: Vertical Integration
Why did IBM Watson succumb to its self generated cancer. Here’s what the Forbes’ write up asserts:
Being vertically integrated gave IBM complete end-to-end control over Watson. But it condemned Watson to being applied in only a few areas. IBM essentially had to guess where this powerful technology could best be applied. Even within health care, some likely areas for Watson like radiology were not pursued in its early years. Because of the limited number of areas IBM was able to explore for using Watson, we will never know whether there were other areas where Watson might have performed beautifully.
Okay, this means in my opinion that IBM engineers and scientists wanted to run the show. There was, therefore, one throat to choke. That throat was IBM Watson’s. The fall out continues. A new CEO, hoots of laughter when I tell people about IBM’s Watson ads, and the loss of shareholder value. I would roll in the weird layoffs as a somewhat desperate way to slash costs too.
Alleged Stalin quote: Death is the solution to all problems. No man – no problem.
Forget vertical integration. The reason for failure was that the system and method did not work.
The Reality
Mr. Jughashvili would be proud of this analysis. It rewrites history. But like Mr. Jughashvili’s, Watson’s actions live on. Changing the words does not alter the reality.
Don’t believe me? Just ask IBM Watson. Is IBM “open”?
Stephen E Arnold, February 19, 2020
Comments
-
- Subscribe to Beyond Search
Feature archive
News archive
-
