Wednesday, March 8, 2023

Seems like a good time to revisit IBM's war on cancer

Friday, September 15, 2017

A few notes on IBM's Watson, the battle against cancer, and what's wrong with the state of 21st century innovation

If you haven't read the Stat News article by Casey Ross and Ike Swetlitzon on IBM's project to revolutionize cancer treatment using Watson, you should do so as soon as possible. The piece raises all sorts of important points. I'll try to return and explore some of them in greater detail (perhaps even convince Joseph to join the conversation), but for now there are two or three I want to it while the topic is still fresh.

Before we go on, you should take a few minutes to listen to the following track from the Button-down Mind of Bob Newhart. It's a seminal comedy bit that also happens to be directly relevant to the conversation.

The joke here is not that using airplanes for passenger travel is absurd (it was the early 60s for Christ's sake); the joke is that the aircraft at Kitty Hawk was clearly not ready to be monetized in this way. A plan that will be viable sometime the future is not really viable.

Given the complexities of the problem and the mountainous quantities of research that need to be assessed, no one would argue that AI-based tools for diagnosis and recommending treatment wouldn't be tremendously valuable, probably even revolutionary, but IBM's Watson is looking definitely less than ready for prime time in this respect.

The stat news article also hits on a couple of long-standing threads here at the blog, starting with one of the best examples of ddulite reasoning we've seen in a long time.
IBM said in its statement that it has collaborated with the research community and presented data on Watson at industry gatherings and in peer-reviewed journals. Some doctors said they didn’t need to see more research to know that the system is valuable. “Artificial intelligence will be adopted in all medical fields in the future,” said Dr. Uhn Lee, who runs the Watson program at Gachon University Gil Medical Center in South Korea. “If that trend, that change is inevitable, then why don’t we just start early?”

Adopting technology because it supposedly will be the best choice sometime in the future is deeply flawed for at least a couple of reasons. First off, the future is, you know, in the future – – we don't know what we'll find there. While it is reasonably safe to assume that some kind of AI will play a vital role in medicine 20 or 30 years from now, it might have very little to do with the approaches currently being pursued. Furthermore, even early adopters who guess right about the direction of technology are often screwed by getting ahead of the functionality.

Another ongoing thread here at the blog is the ever increasing emphasis on hype and marketing at the expense of research and genuine innovation, particularly when combined with short-term thinking. Following speaks for itself:
“IBM ought to quit trying to cure cancer,” said Peter Greulich, a former IBM brand manager who has written several books about IBM’s history and modern challenges. “They turned the marketing engine loose without controlling how to build and construct a product.”

Greulich said IBM needs to invest more money in Watson and hire more people to make it successful. In the 1960s, he said, IBM spent about 11.5 times its annual earnings to develop its mainframe computer, a line of business that still accounts for much of its profitability today.

If it were to make an equivalent investment in Watson, it would need to spend $137 billion. “The only thing it’s spent that much money on is stock buybacks,” Greulich said.

1 comment:

  1. Point 0: There was a later STAT article providing a detailed autopsy of that train wreck after the Watson project was cancelled, but was behind their (expensive) paywall. The headline was something to the effect of "Watson Hardware Sold for Scrap".

    (IMHO, even $137 billion wouldn't have saved Watson: science and medicine are hard, and AI folks think they're smarter than real scientists and doctors.)

    Point 1: From that GPT in psych thread, it seems ELIZA was actually a better shink than GPT. Which is pretty funny. (I'm actually serious here: people really enjoyed typing at ELIZA back in the day.)