[ohhoe]

Yeah. People are using it.

-- For starters, Watson has taken residence at three of the top cancers hospitals in the US -- Memorial Sloan Kettering Cancer Center, University of Texas MD Anderson Cancer Center, and the Mayo Clinic -- where it helps with cancer research and patient care.

-- In Australia, the company ANZ Global Wealth is focused on the latter. The company uses the Watson Engagement Advisor Tool, an NLP SaaS offering, to observe and field customer questions. Similarly, DBS bank in Singapore uses Watson to ensure the proper advice and experience for customers of its wealth management business.

etc etc. -- http://www.zdnet.com/article/ibm-watson-what-are-companies-u...

[maxxxxx]

"helps with cancer research and patient care."

What does that mean? I only hear good sounding phrases but nobody ever gives me any detail.

[ohhoe]

What kind of details? Like exactly what it's doing?

[nkw]

I think that is what the above poster is asking and the root of the original inquiry. Most press and IBM originated information about Watson reads like vague sales collateral or PR puff. The first link I get on google when I type in "watson" is titled "IBM Watson - The platform for cognitive business" [1] which is pretty meaningless. I click "What is Watson?" and get this page [2] which says "IBM Watson is a technology platform that uses natural language processing and machine learning to reveal insights from large amounts of unstructured data" which is great but that could describe anything from Siri to Google to Yelp to Microsoft Word and Excel.

The above post saying "it helps with cancer research and patient care" is the perfect example of this. Yeah it sounds good, but what is it doing? Keeping patients warm? Writing publications for researchers? Churning through raw experimental data? Rationing healthcare (er, making treatment decisions) with sophisticated mathematics? Generating a quick summary of new research papers for treaters? I have no idea..

It may be super cool technology, but IBM has done an impressive job of completely obfuscating its nature and capabilities with sales speak gobbledygook.

[1] http://www.ibm.com/smarterplanet/us/en/ibmwatson/

[2] http://www.ibm.com/smarterplanet/us/en/ibmwatson/what-is-wat...

[tormeh]

The PR speak a while ago was that it would sift through research papers and books looking for relevant answers to a query - a much smarter Google Scholar. Nobody can keep track of all the cancer research, apparently, so machine analysis makes sense, IBM says.

[ChristianBundy]

Right, but is that what it's doing? It sounds like PR is all that exists.

[fabiobruna]

I've worked with some people from IBM, doing some cancer analyses with Watson. I think the goal was to compare US/EU results(cancer get different treatments in different regions I guess). We "fed" Watson with actual patient data / diagnostics(getting/selection data was my job) and reviewed the resulting suggested treatment plans. Watson would suggest different plans, with results stats, links to pubmed, datasets etc. I'm no medic, but it looked nice. I wouldn't mind a "second opinion" like that.

[FatAmericanDev]

Keeping patients warm, they actually put a bed in the server room.

[nharada]

My understanding is that Watson basically acts as a cloud machine learning platform and people can build on it as they see fit. In that sense, Watson is basically the "brand" of IBM's machine learning offerings, no matter how singular the advertising makes it sound (i.e. the ads make it sound like a singular AI platform).

Can you explain what differentiates it from Amazon Machine Learning, Microsoft Azure Machine Learning, etc?

[ohhoe]

In terms of some of the cancer stuff, this can explain it way better than I can.

"But it can take weeks to identify drugs targeting cancer-causing mutations. Watson can do it in minutes and has in its database the findings of scientific papers and clinical trials on particular cancers and potential therapies."

http://www.businessinsider.com/r-ibms-watson-to-guide-cancer...

[i000]

As someone who works at the bioinformatics end of a clinical cancer sequencing operation, and someone familiar with the MSKCC operation (they are doing good work with their IMPACT assay) I have to say that the landscape of matching patients with drugs is so sparse (few drugs, few FDA-approved links, not enough trials) that no Watson is necessary, just basic string matching and a comprehensive high-quality up-to-date database (which does not exist and currently there are 10+ companies working on it)

[gmarx]

One the things I have read it is intended for (and which strikes me as a plausible use of the technology) is to read the medical literature and direct the physician to it when relevant issues arise in patient care. The example they gave was a doc considering medication A for a patient who was allergic to medication B. The doc learned long ago that there was a high correlation between bad reactions to these two drugs so he was surprised when Watson recommended drug B anyway. Watson then gave him the link to a paper from 6 months previous debunking the correlation. I think Watson could be very helpful if this ever works reliably but I see several difficult problems.

Number one is, how does Watson judge the quality of a journal article when the overwhelming majority of them suck?

[et2o]

It is almost always poor practice to make clinical decisions based upon a single paper. Medicine is really complicated.

[maxxxxx]

Yes, what does it do?

[hyperbovine]

Like, what are the things which it does?

:-) I 100% know where you are coming from in this post. All I ever see are ads oriented at PHB types, shown in PHB-type venues (airport corridors, tennis tournaments).

[ohhoe]

http://www.ibm.com/smarterplanet/us/en/ibmwatson/developercl...

the 'learn more' part for each service should tell you more about what each service does.

hope this helps. it would take me a long time and a lot of copying and pasting to relay all the same stuff.

[et2o]

"helps with cancer research"

Sure... Show us a single publication.

[ohhoe]

Okay, here's two, then I gotta get off this thread and get back to work:

http://www.businessinsider.com/r-ibms-watson-to-guide-cancer...

https://www.mskcc.org/about/innovative-collaborations/watson...

[toast_coder]

You seem to be working against your own argument.

1) Business Insider is at best, simply not research journal, nor is it specifying results.

2) The medical center publication first statement says "A team of physicians and analysts at Memorial Sloan Kettering has been “training” IBM Watson for more than a year..."

Seems like like something productive would be going the other way....

[ohhoe]

No... it wouldn't be going the other way. You have to train with corpuses of data.

[ChristianBundy]

Right, but training isn't productive unless it's means to an end. Has there been a productive end, and can you share it with us?

[et2o]

Neither of these are a peer-reviewed publication.

As a bioinformatician, I'm really struggling to see what is new here. It seems like it's mostly a marketing thing.

[ryptophan]

How much of Watson's NLP success can be attributed to IBM acquiring AlchemyAPI (http://www.alchemyapi.com/)?

[ohhoe]

I honestly can't answer this, I've been here less than a year and am just a front end person. My team is v small.

[johansch]

You are on the Watson team, right?

[ohhoe]

sure am