Ask HN: Is anybody doing something productive with IBM Watson or is it just BS? | Hacker News Reader

Ask HN: Is anybody doing something productive with IBM Watson or is it just BS?

62 pointsmaxxxxx51 comments

I have lately been through several pitches from companies and "architects" where they were talking about using IBM Watson but they rarely get into detail what they get from it. Is this just a buzzword for sales guys or are people doing something productive with it?

View on HN

I've posted my results with using their speech-to-text system:

https://github.com/dannguyen/watson-word-watcher

I think it's actually very good and could be of great use to people needing to mine and organize audio -- not for perfect transcriptions, but as a first pass solution...but I don't think that's the small vision that Watson purports to cater to...that is, if you fed the results of the speech-to-text into Watson's various language processing APIs, you'd get diminishing returns, though to be fair, I haven't really used those other APIs in bulk programmatic fashion yet [1]

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

This is great, thank you.

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...

You are on the Watson team, right?

sure am

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

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.

"helps with cancer research and patient care."

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

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

Yes, what does it do?

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).

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.

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...

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)

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...

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.

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

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.

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

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?

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?

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

"helps with cancer research"

Sure... Show us a single publication.

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...

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....

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

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?

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.

I've helped IBM explore some consumer applications around Watson.

We were basically using it as a recommendation engine. In contrast with regular recommendation engines which require other users to feed it data on preferences and then try to match you with similar users, the project we were working on had Watson build a feature space on its own (something it is very, very good at) then match a user profile within that space. In short, it cut out the 'similar person' middleman. It showed what 'shape' a user was in a feature space, the overall shape of the full feature space, and where that user was in that space. Thus, if a person wanted to get from where they are to somewhere else in that feature space, Watson provided a map, more or less.

Why is that important? Well, building a proper recommendation engine when it depends on a network of people is hard. Think Amazon and Netflix -- part of their competitive advantage is that recommendation network effect built by their users. The more people, the more value. The fewer people and it might be way off base and turn off your early users. What if you didn't need to move through the preliminary steps of building that network? What if you could jump to the end? Suddenly that moat dries up. That's what Watson offers, at least for that particular application type.

This sounds really interesting. Do you have a paper or more details?

Outside of some niche areas right now I think what you are hearing is people trying to pitch others to pay for their time to experiment and build something with Watson. Not that they already have a solution, again except for some niche industries and groups.

Overall, IBM is having to reinvent itself (again), and IBM people themselves are pushing in new directions and pitching new ideas which fit their longer term plan. At the same time they have this huge legacy dragging behind them, so it can't be easy. Personally, I have never been a huge fan of some of the total crap (consulting mainly) they put out there, but then they do some crazy amazing stuff with their research groups. And I have seen a few really stellar software consulting gigs they have done too, but that seems to be the minority. But at least I'll say for them, they aren't Oracle, and they will complete a project.

I am aware of some groups that are (at least planning to) use IBM Watson to analyse the outcomes and impacts of social policy research.

As a (made up) example: feeding the AI data-sets of various social investments made (eg: education, mental health, social & medical support) alongside health and societal outcomes (crime, hospital intake, suicides, etc..) then overlaying it on a map and maybe getting Watson to "tell you" which investments had the best impacts.

That particular example is actually an amalgam of a number of projects my girlfriend has undertaken recently in her career as a social-policy analyst BUT she is also enrolled in a university Master's program next year where one of the possibilities for her placement is to work supporting the Watson team in figuring out how to use Watson to improve lives & outcomes in the real world.

I've also heard about using Watson in analyzing health data (detecting cancers based on mass x-ray analysis) - but I don't know how much of that is currently working.

Hope that helps - I know it's all anecdotal from my side and I can't link you to anything published on the topic, but you can take the example for what it is worth :)

EDIT - I'll try and ask her if she got any more information about what she was going to do for Watson.

I had one company that used Watson flat out tell me that it was purely for marketing - they could have built their own solution just as easily, but used Watson because it had more marketability.

From what I understand, it's really just a wrapper around IBM, or a collection of things IBM has built or offers.

I have not yet seen anything particularly interesting - but i'm glad you asked this question! Hopefully someone will chime in with something worthwhile.

If what you're saying is correct then it could be seen as a sort of framework, though for marketing reasons that wouldn't sound as attractive I suppose? I have no experience in this field, though I would love to get into it someday.

I dunno about Watson, but it is my (inspecific) opinion that at least one state government and at least one upper-middle-sized corporation have a bunch of IBM big iron laying around that they don't need.

The real burn is, that's not JUST an upfront cost... You pay per month.

At the corp, I could not find anyone except the CEO that thought the investment was a good idea. I figure IBM gifted him a car or something. Complete speculation, but it's the only thing that makes sense!

Elsewhere, inspecifically... At least the machines are used. But honestly how many giant black modern IBM boxes does it take to fully emulate every single AS/400 machine that existed within the entity back in 1996? A quick look at the specs tells me: 1. There's no way it would take more than one.

My brother is creating new chocolate confections with Watson's ingredient pairing. :-D It's called Chocolat Monarque and it's in Montreal.

Can you share more please? merci

https://www.ibmchefwatson.com/community

Chef Watson is allegedly very good at what it does.

I've signed up for chef Watson as well! it's really cool.

https://www.ibmchefwatson.com/

I've used it for a prototype project for the news industry, though not in production.

IBM Watson: Building a Cognitive App with Concept Insights

http://www.primaryobjects.com/2016/02/01/ibm-watson-building...

From what I've seen watson is just the branding and they put a bunch of companies they've acquired working in the general data / analytics space under the hood.

It's not the solely the setup that won on jeopardy all those years ago.

I personally thought their tradeoff api was pretty interesting, but until I hear/see some real world results, I will continue to be skeptical. If anyone here could shine some light on something that would be awesome.

http://www.rossintelligence.com/ they use watson AI to simplify the legal system.

This should answer your questions: http://www.ft.com/cms/s/2/dced8150-b300-11e5-8358-9a82b43f6b...

Tldr: mostly hype. The "solve cancer" claims haven't worked out. Now there are lots of big data/machine learning APIs being branded as "Watson" and some are quite good.

I've looked into it for some past hackathon projects and a side project. Watson can take care of some of the heavy lifting of NLP, categorization, or parsing large volumes of text, et, in different ways, but it's not some jeopardy-winning AI you can just plug into something you already have (at least not yet). It can abstract away some tricky challenges if your needs fit what it can do.

Here is a recent Dallas R Users meetup on the very of topic of using IBM Watson

http://www.meetup.com/Dallas-R-Users-Group/events/229283377/

The coolest thing I've seen is an internal IBM project that organizes the news:

http://news-explorer.mybluemix.net/

It's just BS.

I'm successfully using it to supply a hedge fund with a new kind of edge.

Are you hiring? :)

ITT: People who have not used IBM Watson for something productive..

Reply on news.ycombinator.com