[Sixstrings]

IBM's Watson To Be Trained In Cancer

WHITE PLAINS, N.Y. — The medical training of IBM's speedy Watson computer will continue with a residency at a renowned Manhattan cancer hospital.

IBM and Memorial Sloan-Kettering Cancer Center said Thursday that they will add the latest in oncology research – and the hospital's accumulated experience – to Watson's vast knowledge base, and keep updating it.

The result should help the hospital diagnose and treat cancer more quickly, accurately and personally, they said.

"The capabilities are enormous," said Dr. Larry Norton, deputy chief for breast cancer programs at Sloan-Kettering. "And unlike my medical students, Watson doesn't forget anything."

Watson won fame by beating the world's best "Jeopardy!" players. Applying its speed and language skills to medicine was a longtime goal at IBM, and Watson went to work last year for the health insurer Wellpoint Inc.

The training at Sloan-Kettering will take time, and it may be the end of next year before patients at the hospital are benefiting from Watson's speed and depth, said Dr. Martin Kohn, chief medical scientist at IBM. If successful, the finished product could be used anywhere in the world to aid cancer treatment.

Kohn said there's a rule of thumb that it takes 15 years for breakthroughs in medicine to be disseminated around the world.

"So any process that can help get valuable information about choices and treatment out into general use more rapidly obviously is an improvement," he said.
http://www.huffingtonpost.com/2012/03/22/ibm-watson-cancer_n_1372426.html

This is a stepping stone to the singularity.

The supercomputer "Watson" can fully comprehend something like two million pages in three seconds. This is the computer that played and won at Jeopardy. The overall hope with the cancer project is that by taking in, digesting, and analyzing such vast volumes of cancer research, data, reports, patient records, EVERYTHING, the computer can do what no one person or large team could ever do -- comprehend and analyze EVERYTHING all at once and hopefully see connections between disparate data, and therefore maybe lead to a cure.

A cure for cancer would be great news. But the next step is a Watson supercomputer being able to put hundreds of thousands, maybe millions or tens of millions, of cubicle workers out of a job. If it cures cancer that's worth it, more socialism less work! As a society we'll have no other choice once computers and robotics get so good they can do anyone's job.

[PrestonSturges]

That sounds like pure desperation on IBM's part. Supercomputers are usually tragically underutilized - it's easy to make the sale but hard to get people to use it.

[Sixstrings]

That sounds like pure desperation on IBM's part. Supercomputers are usually tragically underutilized - it's easy to make the sale but hard to get people to use it.

Why do you say that?

Just to clarify, this particular project is aimed at suggesting and treatments, not finding a cure per se -- but that's where this is eventually headed.

The most powerful thing about Watson is that it can comprehend millions of pages, you could just dump the records from every doctor and hospital and insurer into it, then Watson cooly calculates the answer "it feels most confident about." As seen on jeapardy, it does pretty well.

That right there is very powerful, a computer that can truly comprehend language not just raw numbers. If you have a disease with multiple treatment options, and throw your data into Watson who can then compare it to millions of others just like you and how they turned out etc., the computer could come up with a nice list of recommendations with % chance of outcomes. It will really help diagnosis and treatment.

Even without supercomputer help, just using google to do searches, studies have shown doctors who use google make better diagnoses and there's less misdiagnosis and error.

[PrestonSturges]

Well I read the article and it looks like they are actually going to do the hard work involved with predigesting books, journals, and medical records. That sounds a bit more realistic. And there is nothing about "curing cancer," it's about automating diagnostics and the prescription of standard treatments. Cancer is tough to cure, but the classifications of tumors and cancer progression have unique degree of standardization because of all the work that pathologists have done for the last 50 years.

Now, if they could develop a good sysem for reading in old patient discharge and treatment records that would be nice, but this stuff has been bouncing around for decades. The point is that IBM exists to sell hardware, and they don't care about the human labor involved. I don't expect them to accomplish much over the next five years other than selling stuff, because that's all that matters. IBM does not care about cancer.

[Graeme]

http://the-scientist.com/2012/04/10/a-u ... r-vaccine/?

[Plantagenet]

it's about automating diagnostics and the prescription of standard treatments.

IBM is going to replace Dr. House and his crack diagnostic team with a supercomputer?


And who exactly will the Supercomputer get to break into the patients' homes to find all the clues?

[Keith_McClary]

Wasn't clear to me if the hospital is buying Watson or IBM is giving/lending it/him.

I thought the Space Station was supposed to give us all that singularity science, like curing cancer and stuff. Whatever happened to that?

[EnergyUnlimited]

Just searching a database.
Singularity?
Don't make me laugh.

Cannot see any creativity or AI in searching a database.

[Fishman]

Or
GREAT!!!! Sales job for Memorial Sloan-Kettering Cancer Center. I mean wow, every one who has watched Jeopardy will want to go there. Will my Obamacare fly me there?

[Beery1]

This is a souped-up adding machine - it cannot find a cure for anything. All it can do is assemble statistics on what is most likely to work, based on what has been done before. Expecting that this machine can work towards finding a cure for cancer is like expecting it to write a world class violin concerto. It's simply not going to happen.

[PrestonSturges]

It's very tough to sell a database because a database is a "cost center" not a "profit center," and adminstrative thugs don't care about the quality of the information in the sytem or the accuracy of the bill as long as the bills go out in vast numbers.

However, Monty Python nailed it where the hospital administrators are all giddy over the purchase of the "machine that goes ping."

[vision-master]

If Pot Prevented Cancer You Would Have Read About It, Right

And now for the study you haven't heard of. Writing in the August issue of the journal Cancer Prevention Research, investigators from Rhode Island's Brown University, along with researchers at Boston University, Louisiana State University, and the University of Minnesota reported that lifetime marijuana use is associated with a "significantly reduced risk" of head and neck squamous cell carcinoma.

Authors reported, "after adjusting for potential confounders (including smoking and alcohol drinking), 10 to 20 years of marijuana use was associated with a significantly reduced risk of head and neck squamous cell carcinoma (HNDCC)."

Perhaps even more notably, subjects who smoked marijuana and consumed alcohol and tobacco (two known high risk factors for head and neck cancers) also experienced a reduced risk of cancer, the study found.

"Our study suggests that moderate marijuana use is associated with reduced risk of HNSCC," investigators concluded. "This association was consistent across different measures of marijuana use (marijuana use status, duration, and frequency of use)....Further, we observed that marijuana use modified the interaction between alcohol and cigarette smoking, resulting in a decreased HNSCC risk among moderate smokers and light drinkers, and attenuated risk among the heaviest smokers and drinkers."

http://www.huffingtonpost.com/paul-arme ... 61157.html


Light up you squares.

[PrestonSturges]

Better safe than sorry

[Outcast_Searcher]

IBM's Watson To Be Trained In Cancer

http://www.huffingtonpost.com/2012/03/22/ibm-watson-cancer_n_1372426.html

This is a stepping stone to the singularity.

The supercomputer "Watson" can fully comprehend something like two million pages in three seconds. This is the computer that played and won at Jeopardy. The overall hope with the cancer project is that by taking in, digesting, and analyzing such vast volumes of cancer research, data, reports, patient records, EVERYTHING, the computer can do what no one person or large team could ever do -- comprehend and analyze EVERYTHING all at once and hopefully see connections between disparate data, and therefore maybe lead to a cure.

(Red bolded font in Six's commentary mine, for emphasis).

Um no, Six -- the "fully comprehend" words show you don't comprehend this AT ALL.

I was heavy into mainframes for 32 years (college and IBM), and both wrote chess programs that played more like people than anything like their peers in the early 80's, and had significant insight into IBM's advanced DB2 database system as a DB2 systems programmer for 17 years -- so I have some idea about the concepts involved here.

Just becase a program like Watson, or the Deep Blue chess program is EXTREMELY good at a very narrow domain (like database search or searching chess data trees) and can therefore produce absolutely amazing and world class results at that specific (narrow) domain -- does NOT in any way mean that the program can COMPREHEND anything.

Most importantly, it doesn't imply it can expand its scope ONE IOTA without specific programming and immense amounts of specific data input in the area one wishes to expand expertise into.

Let's face it. Doctors are pretty bad at diagnosing cancer. They are REALLY bad at curing it, unless they are LUCKY enough to catch it VERY early.

Now, a Watson like program may eventually be able to help quite a bit, after an extraordinary amount of time, expense, and effort is put into it -- just like the decades and $billions spent on chess programs.

That's nice. (If you have a nail, a hammer is nice). But is it anything remotely like a singularity or especially "comprehension"? Absolutely NOT.

[drgoodword]

Sounds like a potentially useful bit of hyper-data-mining. I've often thought (in my layman's semi-numerate way) that the clearest views of reality belong to the statisticians, actuaries and epidemiologists of the world.

[PrestonSturges]

Sounds like a potentially useful bit of hyper-data-mining. I've often thought (in my layman's semi-numerate way) that the clearest views of reality belong to the statisticians, actuaries and epidemiologists of the world.

Well it only took them 40 years to decide smoking causes cancer, so there's always that forest/trees vision problem.