IBM's megabrain Watson to make mobe, slab apps smarter? Not so fast IBM wants developers to build smartphone apps that use Big Blue's clever Jeopardy!-beating Watson software. But harnessing the TV star's silicon brain will require more than just invoking a few API calls with JSON: the app programmers will have to do a lot heavy lifting themselves to train Watson. The Watson Mobile Developer …
Typical marketing hack didn't get the reference.
In terms of 'hero who could do no wrong' that would be more L. Ron Hubbard 's style which plays in to his Scientology stuff. (You can replace stuff with harsher words if you'd like. I was focusing on the author's literary skills.)
In L. Ron Hubbard, the protagonist is the perfect Uber Mench. Always right, can do no wrong. I would say its probably due to his growing up on movie serials where the hero is alwasy a hero and has no flaws.
Robert A. Heinlein's main characters aren't perfect and do have some 'morale' flaws. But they tend to be smarter than the average bear, able to figure things out ahead of the rest of the crew, of course they do have help from friends. More of an adventure series but again the hero isn't flawed like today's Batman and others. (Choose a character out of Gibson's novels.)
To your point, I guess one could do a masters and even a PhD on the evolution of science fiction writers and their protagonists. But what do I know? ENG meant college of Engineering and not English. Although the only English class I ever got an A in was intro to Science Fiction.
I do like most of RA's work. It's just that he sometimes comes across as a bit elitist. Somewhat smug with it. He often suggests that anarchy is viable on the grounds that unfettered humans will do the right thing.
But it's fun enough. I re-read Podkayne of Mars, The Moon is a Harsh Mistress, Starman Jones, both volumes of Assignment in Eternity and Have Spacesuit, Will Travel (the latter being the first book I ever bought).
I considered reading the novel about a brain transplant but couldn't find it (my copy is hardback and not with the others - Fear no Evil?). Anyway that's a rather weird book so I called it quits and started on Niven instead.
Yeah... I think when you get to The cat who could walk through walls you hit the end.
And yes, if memory serves it was Fear no Evil however, it does provoke some thought about what if you could do a brain transplant.
I also do agree that Niven is a better read, however totally different and deals with more physics instead some of the social issues or questioned raised by Heinlein. Heinlein seemed more into Hippie movement.
A lot of the Sci-Fi Fantasy was written as a way to discuss social issues that could have been considered 'taboo'.
They would rename the machine from Watson to Mike.
Ha, yes. I re-read that novel last November. Actually one of his better ones although still has some of that annoying 'hero can do no wrong and knows everything' feel about it. Still, IBM's Watson is a fair attempt at a dinkum-thinkum :)
From the article: "most other app competitions involve the winner getting cold hard cash"...
If I understand correctly, "most other app competitions" shell cash for already developed apps. In this case contestants are supposed to "come up with an idea" and IBM will provide assistance (resources? expertise?) to develop a product based on the best ideas. Seems fair...
So they taught it all about cancer and then...what, it couldn't be arsed* so it's cocking about with apps now instead? What if it gets bored of apps as well, will it just go and play flappy birds instead?
*I'm assuming here that it didn't in fact cure cancer since I'm sure I'd have seen that in the news.
My understanding is that there is no "cure for cancer" because cancer is not a single disease. There are thousands of different forms, all with very different treatments. And there are no "cures" in the sense of "take these pills and call me in the morning", but instead a variety of approaches with varying probabilities of sending cancer into remission, and these probabilities vary from person to person.
So what may have happened is that Watson came up with some slightly-improved versions of existing treatments; or it made some interesting observations into the mechanisms of cancer that are not in themselves treatments, but were handed on to other researchers who might be able to use them to create new treatments somewhere down the line.
So, Watson did not find a magic bullet, not because they didn't run it for long enough, but because there is no magic bullet and never will be. But it probably made some useful, incremental discoveries nonetheless.
My understanding is that there is no "cure for cancer" because cancer is not a single disease.
Right. Cancer is essentially any instance where cytogenesis (cellular reproduction) outstrips apoptosis (cellular death) - whether that's because the former runs amok or the latter scales down. "Cancer" just means that one of the systems of a multi-cellular organism has gone out of equilibrium. You can't cure cancer as a whole for the same reason you can't "cure" mechanical breakdown, say.
More relevantly, what IBM are training Watson to do is recommend treatment regimes for individual patients, based on best probability of good outcome. It's mechanical evidence-based medicine (EBM).
This is a good use for the technology because one of the great problems with EBM is that clinicians don't have the time to stay on top of research in their field, particularly not with a very active field like oncology. And another problem is that humans make poor Bayesian reasoners, which means they mis-estimate probabilities. A big mechanical probabilistic modeler like Watson can handle both of those, and then present the highest-probability alternatives to the doctor, who can consider individual circumstances and work with the patient to determine a course of treatment. It's an approach that lets the machine do things it's good at, and the doctor do things that people are good at.
(It's also worth noting, as an aside, that medical research and data are more amenable to this kind of processing than many corpora. The former tends to use technical terms which are less ambiguous than general prose, and citations can be used to track influence, redundancy, etc. The latter is more and more coded for unambiguous machine processing. And there's been a lot of research in this area, so we know a lot about mechanical processing of medical information.)
Please can we stop with the "people in the 1400s thought that the earth was flat" trope? You'd have been hard-pressed to find an educated person back then who wouldn't laugh at you for suggesting a flat-Earth model.
And if you were asking uneducated peasants for help in generating your dataset, then you deserved to encounter the difficulties of updating your anachronistic thinking machine.
By example, if Watson had been fed a full dataset from the 1400s that stated unequivocally the world was flat, it would take it some time to adjust to new data coming in that stated the world was round, but adjust it would. This is fundamentally different to how current computers work and is a laudable, fascinating bit of technology.
What the every-lovin' fuck are you going on about here?
Watson is an application running on a bunch of conventional computers. You can put much of it together yourself and run it on any bog-standard Windows, Mac, or *ix machine, if you like: Hadoop, UIMA, etc are all there ripe for the picking. DeepQA itself might not be, but there are any number of ML algorithms you can run under UIMA - HMM, MEMM, SVG, "deep learning" ANNs, etc.
And - try to stay with me here - all machine-learning systems are capable of revising their models based on new data. That's what the "learning" part of "machine learning" means.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
