The only machine working over at Watson is the PR machine.
It's astonishing how much IBM has spent to achieve so little. Per the Jefferies report, IBM spent $15 billion in Cognitive from 2010 to 2015, not including an additional $5 billion in data acquisitions. Even if Cognitive weren't a marketing buzzword invented by IBM to describe if-then statements, that would still be deeply embarrassing. (You could probably double the number of startups funded in Silicon Valley in a year with $20 billion dollars, and I'll bet you'd get much higher returns.)
When you look at the main open-source AI frameworks, none of them were created by IBM, because IBM was incapable of building them. There are individual startups that have done more in AI, using far fewer resources.
And the real shame is that IBM is convincing rubes to buy their cognitive trash, and the rubes are doing it. And when the rubes' bosses realize that the rubes blew the AI budget and the "AI" delivered nothing, those rubes are getting canned. You can get fired for buying IBM. I've seen it.
There's a Madoff-like quality to IBM's Watson sales strategy. The more they fail, the wider they circle to depend on dumber and dumber money from abroad. They're going after rural Scandinavia. They're selling to 3rd-tier cities in China. Just imagining targeting a buyer with the sophistication of the mayor of Mobile, Alabama. That's where IBM Watson is selling strong.
IBM has been poisoning the well in the AI industry for years, teaching customers to be gun-shy and scared of an important technology, just because they blew their money on a dog to begin with. I don't believe an AI winter will come, but if it did, IBM would be the main culprit.
Even if Watson's many mysterious technologies worked, they would still have a lame business model, because most of AI, and the best of AI tooling, is open source. IBM's trying to sell a secret sauce, even as it declines to participate in competitions that would should how Watson stacks up against other tech. Oren Etzioni has been calling them out for years. [1,2,3]
Anyone wondering how IBM came to this sorry pass should read Robert Cringely's "The Decline and Fall of IBM." It's a company run by a clueless management class, who probably believe the smoke they're blowing up everyone else's ass. That's the scary part: The C-suite at IBM actually bought that hype about Watson, until they started to see the numbers, and the numbers sucked. That's why they don't break them out in the quarterly earnings reports. IBM is not signing up the Watson partners it thought it would. [0]
Watson makes claims to solve everything. But when you dig in with the sales people, the only success story they cite is Memorial Sloan Kettering (MSK). It's suspicious, considering how many other organizations they must have worked with. H&R block is bragging about Watson technology, too, but that's just because they think the marketing gimmick will work for them as well as it's worked for IBM.
The real problem in a situation like this is that very few people have an incentive to get the truth out. IBM doesn't want its stock price to dive. And all the customers are mortified that they failed and wasted their money. And they probably had to sign some non-disparagement clause to boot.
I was actually thinking of the main deep-learning/RL frameworks: TensorFlow, Keras, Theano, Torch, MxNet, Chainer, Caffe. That's where the action is. And if you're talking about AI today, you need to be talking about deep neural networks.
It's astonishing how much IBM has spent to achieve so little. Per the Jefferies report, IBM spent $15 billion in Cognitive from 2010 to 2015, not including an additional $5 billion in data acquisitions. Even if Cognitive weren't a marketing buzzword invented by IBM to describe if-then statements, that would still be deeply embarrassing. (You could probably double the number of startups funded in Silicon Valley in a year with $20 billion dollars, and I'll bet you'd get much higher returns.)
When you look at the main open-source AI frameworks, none of them were created by IBM, because IBM was incapable of building them. There are individual startups that have done more in AI, using far fewer resources.
And the real shame is that IBM is convincing rubes to buy their cognitive trash, and the rubes are doing it. And when the rubes' bosses realize that the rubes blew the AI budget and the "AI" delivered nothing, those rubes are getting canned. You can get fired for buying IBM. I've seen it.
There's a Madoff-like quality to IBM's Watson sales strategy. The more they fail, the wider they circle to depend on dumber and dumber money from abroad. They're going after rural Scandinavia. They're selling to 3rd-tier cities in China. Just imagining targeting a buyer with the sophistication of the mayor of Mobile, Alabama. That's where IBM Watson is selling strong.
IBM has been poisoning the well in the AI industry for years, teaching customers to be gun-shy and scared of an important technology, just because they blew their money on a dog to begin with. I don't believe an AI winter will come, but if it did, IBM would be the main culprit.
Even if Watson's many mysterious technologies worked, they would still have a lame business model, because most of AI, and the best of AI tooling, is open source. IBM's trying to sell a secret sauce, even as it declines to participate in competitions that would should how Watson stacks up against other tech. Oren Etzioni has been calling them out for years. [1,2,3]
Anyone wondering how IBM came to this sorry pass should read Robert Cringely's "The Decline and Fall of IBM." It's a company run by a clueless management class, who probably believe the smoke they're blowing up everyone else's ass. That's the scary part: The C-suite at IBM actually bought that hype about Watson, until they started to see the numbers, and the numbers sucked. That's why they don't break them out in the quarterly earnings reports. IBM is not signing up the Watson partners it thought it would. [0]
Watson makes claims to solve everything. But when you dig in with the sales people, the only success story they cite is Memorial Sloan Kettering (MSK). It's suspicious, considering how many other organizations they must have worked with. H&R block is bragging about Watson technology, too, but that's just because they think the marketing gimmick will work for them as well as it's worked for IBM.
The real problem in a situation like this is that very few people have an incentive to get the truth out. IBM doesn't want its stock price to dive. And all the customers are mortified that they failed and wasted their money. And they probably had to sign some non-disparagement clause to boot.
[0] http://www.cringely.com/2016/09/07/carrie-underwoods-america...
[1,2,3] https://twitter.com/etzioni/status/654486809825570817
https://twitter.com/etzioni/status/785639012161630208
https://twitter.com/etzioni/status/861231962853224448