How IBM Watson Underdelivered On AI Healthcare
Artificial intelligence is a much-hyped technology, but as with all products, there is a need to separate the pipe dream from reality. Bloomberg mentions that AI might be considered the most significant transformative technology that mankind currently has access to, but warns that we may fall in love with the idea of what it does and miss the limitations of what it can do. As with most technologies that are hyped up as a massive game-changer and then end up in the rubbish bin of history, we must be wary of how we build up AI. Popular Science mentions a lot of these tech innovations that were hyped up by their parent companies but never made it to market or crashed when they entered it. As we're beginning to see, IBM's Watson might find itself in a similar situation.
Watson's Past Triumphs
Back in 2013, IBM developed an AI that could potentially think and answer in our language and set out to do something unusual with it, namely, challenge the two biggest winners in the history of the TV Show Jeopardy. As Tech Republic reports, Watson was a product of the same sort of program (called IBM's Great Challenge) that spawned Deep Blue, the computer that beat Grandmaster Garry Kasparov at chess in 1997. Watson was impressed into the psyche of the everyday human being, and noting that it had to think about answers the same way humans had to and managed to do so more efficiently and a lot faster impressed the importance of this technology on the rest of the world.
The Next Logical Step
After winning the show, the team that was tasked with improving Watson set their sights on something more impactful. They decided to add the expertise of their silicon child to the world of medicine. Things were going rather well at the start, with initial tests as reported by Futurism giving the same diagnosis as a real doctor in 99% of the cases tested. Watson was tipped to make the system easier to manage, and more efficient for both doctors and patients going forward. However as anyone who works in technology can attest, in many cases, the initial test of something doesn't necessarily make it viable in a real-world scenario. This is even true in the software development.
Deadly Medical Advice from your Friendly Neighborhood AI?
According to The Verge, IBM's Watson had a bit of a departure from proper functioning after handing out potentially deadly advice for treating cancer in July 2018. Luckily, the information was delivered to doctors working with a test case, and no one was harmed, but this incident shows the problems that still exist with Watson as a diagnostic tool. The more the AI learns, the more it must consider to offer advice, and it's this consideration that sometimes leads to disastrous situations. Watson's recent poor diagnosis history isn't the only shortcoming that the system has had.
Selling Dreams through Marketing
When Watson was initially tipped to help the medical sector, it was through its use as a tool for administrative functions in addition to taking the load off doctors through simple diagnosis. As IEEE Spectrum notes, Watson works well in its own enclosed dome and controlled environment, but it can't cope with the messy state of real-world medical systems. Many practitioners that have acquired the hardware admit to doing so in the hopes that it would help their business, but in fact wasn't doing much that they couldn't do themselves. In 2015, from IBM's site, they mentioned the formation of a dedicated Watson Health system to deal specifically with the issue of integrating Watson into the real world. While the promise remains, the actuality has come up far shorter than anyone would expect.
Separating Hype from Application
The Wall Street Journal states that Mark Kris, the oncologist that heads Watson's cancer training thinks that the system still has promise, citing the rapid changes that happen in the field to explain the AI's inability to do what was promised. Even regular doctors are having a hard time believing the hype anymore, with many of them calling for IBM to release peer-reviewed research that proves unequivocally that AI benefits doctors. AI has quite a long road to go before it becomes an everyday part of life, but in several fields, it is showing promise in how it simplifies things. It's very likely that if IBM were more in tune with the needs of doctors, they might have had a more welcoming and accepting audience for implementing AI in the medical field.