7 ms·

IBM Watson fires its own cancer-fighting ‘moonshot’

54 points by prateekj 13y ago

bglazer 13y ago
This is truly incredible stuff. Hats off to IBM for pushing the field of machine learning forward. This statement really struck me: "“In the Jeopardy days [in 2011], Watson was running off 90 servers and could store 15 terabytes of memory,” said Gold. That’s not your typical household integration. “Now it’s far more affordable and runs 240 times faster, so we can do far more with less.” Moreover, the size of IBM’s server is now smaller than a pizza box and can fit in any data center." How is this possible?
- Scryptonite 13y ago
  A stab in the dark would be chips made from Graphene or IBM's Silicene, if I am to understand what you quoted.
- wlesieutre 13y ago
  "Smaller than a pizza box and can fit in any data center" sounds like marketing for "We put it in a standard rackmount form factor"
- ancarda 13y ago
  I'm assuming they mean 1 server is a pizza box (ie. it's a 1U rack server), so you'd still need ~90U of space. Either that or IBM has made some breakthroughs in the last 2 years.
  - drblast 13y ago
    You can almost hear the article author completely misunderstanding what he was told. I'm betting it was something like, "For Jeopardy, Watson was using 15 Terabytes of data stored on 90 servers. And those servers were large back then, now they're your standard pizza box 1U's. And because of advances in storage technology we can store a lot more in the same space." Isn't there some sort of basic tech literacy class for people who write these articles? It's like reading the sports section and seeing "The Seattle Seahawks are a group of big men and they ran at another group of big men, and they've been doing that a lot, and faster lately. Some of the men in the Seahawks group are smaller and faster, so they are harder to catch by the other group of big men. The Seahawks grouping of men is like a successful business, because they've been winning competitions against other groups."
    - ghshephard 13y ago
      16 Terabytes of RAM on P750s. From http://en.wikipedia.org/wiki/Watson_(computer) http://en.wikipedia.org/wiki/Watson_(computer) "Watson is made up of a cluster of ninety IBM Power 750 servers (plus additional I/O, network and cluster controller nodes in 10 racks) with a total of 2880 POWER7 processor cores and 16 Terabytes of RAM. Each Power 750 server uses a 3.5 GHz POWER7 eight core processor, with four threads per core. "
    - cstross 13y ago
      There is some more prize gibberish in that article: Saxena said IBM is ready to make a big a bet on Watson, as it did in the 1970s when it invested in the emergence of the mainframe. Leaving aside the grammatical error (it's probably just an editing artefact), what bed did IBM make in the 1970s on the emergence of the mainframe? Is that a typo for 1950s? Or just unvarnished ignorance of computing history? It's stuff like this that makes me mistrust a news source: if they get it blazingly wrong about something I know enough about to have a working bullshit detector for, what are the chances that they're not also getting it wrong about everything else?
      - nl 13y ago
        To be fair, the System/360 (which was the first to run the classic OS/360 operating system) was released in '64[1], and was replaced in the '70s by the S/370. Certainly mainframes were in use prior to the '70s, but the S/360 was the first "classic" mainframe in that it was designed for both business & scientific use[2], and it wasn't until the 70's that they began to become more popular in areas outside their traditional use. [1] http://en.wikipedia.org/wiki/IBM_System/360 http://en.wikipedia.org/wiki/IBM_System/360 [2]
  - pkaye 13y ago
    The main change from the Jeopardy competitions is they don't need to find an answer in under a second. So you can probably reduce to computitional capability significantly.
- bratao 13y ago
  I would bet on optimization of the software, lower costs of memory, and database sanitization.
- gaadd33 13y ago
  I think prior to that it mentions its cloud based, so the head end is probably only 1U.
- mdisraeli 13y ago
  If we ignore cloud interfacing solutions[1], there is a simple answer. "Enterprise scale" solutions really mean "Enterprise SCALING". Pre-sales for 90 servers in many cases would be impossible. Pre-sales of a single rack mountable item that can then be expanded from depending on the workload is really attractive, however. Write the software right, and it should be able to cope. Pre-sales probably starts with a customer renting a single box to try out, possibly with some professional services time. The customer decides they like it, and commission a project to develop a Watson solution. IBM pre-sales then work out how many servers would actually be needed, once the customer is interested. As the solution can scale, when management goes "that cost looks little high" they can safely say "well, if we reduce it by sixteen servers, you'll be slightly slower but that might be ok. Then if the business decides to increase we can simply add more servers[2] at a later date". And, of course, adding more servers at a later date means more professional services time, meaning more money and more quality customer relationship building time. [1] Which for all kinds of commercial reasons, we have to discount a cloud solution being the default here. Buzzword, yes, but enterprise is slow and many really exciting uses cannot be based in the cloud. [2] Or, y'know, just turn this key....
PaulHoule 13y ago
Why is it always cancer? I'd be impressed if somebody was trying to cure ALS or Tourette's syndrome.
- teraflop 13y ago
  Maybe because cancer is ridiculously more common than both of those disorders combined?
  - PaulHoule 13y ago
    Perhaps cancer in general, but not particularly types of cancer, when you recognize that there are multiple forms of lung cancer, breast cancer, etc. -- common forms of cancer are not a single disease. The trouble with cancer drugs is that you take them for a short time and you either die or you survive. The real money is in drugs that a large segment of the population can take for an indefinite time such as Lipitor, blood pressure meds, asthma inhalers, and the perpetually elusive obesity cure.
  - thaumasiotes 13y ago
    Sure, but obviously curing cancer is a waste of research effort... what they should be trying to cure is disease.
    - foobarbazqux 13y ago
      Why is it a waste of research effort? We've made dramatic non-palliative strides in the area of cancer treatment, and most importantly have been able to induce complete remission in a significant number of cases.
      - thaumasiotes 13y ago
        see my sibling comment.
        foobarbazqux 13y ago
dnautics 13y ago
Instead of Jeopardy, I always wanted Watson to go on "are you smarter than a fifth grader" and throw questions like "think of the second color of the rainbow. What large bird starts with the same first letter?" Or: "If I call A 5, B 4, and C 3, then what do I call D?" As a scientist (working on cancer, no less), more than just churning through terabytes of data, it strikes me that THESE sorts of questions are the ones you'll need to be able to answer. There's also a lot of "reading between the lines" you'll need to do, and a lot of "small pattern extrapolation". For example, the drug candidate I'm working on is one of a family of about let's say, 16 known compounds. All of them kill cancer cells, fourteen are cardiotoxic. The two that are not, lack a single oxygen atom. Maybe that oxygen atom lets the molecule block the mitochondrion (sort of important for heart cells)? Why? Because the molecule family looks sort of like Coenzyme Q10 - if you use your imagination and remember some of your chemical reactivity rules from Grad-level Ochem - and supplementation of Q10 alleviated the cardiotoxicity in one study on rats for one of the molecules. Ultimately, my previous boss was bold enough to latch onto that and give it a shot - can we take this molecule that is the most potent of the group (but unfortunately cardiotoxic), delete the homologous oxygen atom and make something that still works and is less cardiotoxic?
- jknightco 13y ago
  I'm legitimately stumped. Which large bird starts with 'O'? Or is it second from the top.. so an Ibis?
  - ramchip 13y ago
    Ostrich?
    - jknightco 13y ago
      I'm going to go sit in the corner and cry now. Thanks!
  - protomyth 13y ago
    Only one I thought of relates to baseball. Didn't think it was large. Edit: for some reason ramchip I keep missing that one when the subject of birds comes up
JunkDNA 13y ago
As far as I'm aware there have been zero peer reviewed publications showing Watson actually doing something in medicine. When a doctor thinks he or she has a better approach than what is being done, they use the scientific method to find out of their hypothesis holds up. So far I've seen lots of IBM conference presentations and press conferences. But where is the data? Where are the increased survival rates? Where are the reduced medical errors compared to controls? That is science. Everything else is snake oil unless you have that.
- dbcooper 13y ago
  Indeed. It's a little annoying to keep seeing these IBM PR releases about fictional medical suppositions, when one is working on drug screening, diagnostics, and such from within a reality constrained lab!
- hershel 13y ago
  The way i understand it, this system offers data to physicians ,it doesn't offer treatment, so it doesn't have to be clinically tested. It is the the responsibility of the hospital and doctors to evaluate and see if it offer any benefit to them.That's the usual standard information systems are bound to. But watson is quite new. I'm pretty sure testing will come with time.
whiddershins 13y ago
Even if all some AI system did was take a list of symptoms, anonymize them, suggest a diagnosis, and then take the MD's actual diagnosis, track outcomes, and compare results over time, it would be a huge advancement. Right now, unless I misunderstand something structural about our medical system, most of the data gathered by MDs are wasted, because there is no digital record of much of it and patient privacy legalities prevent wholesale entering of symptoms and case outcomes in to any sort of freely available (inter)national database.
nnq 13y ago
Where can one learn more about how Watson actually works? ...starting from the IBM website and Wikipedia one mostly gets to whitepapers and high level descriptions. I hope at least bits of it are open source or at least have the math and logic underlying them well documented.

54 points by prateekj 13y ago

This is truly incredible stuff. Hats off to IBM for pushing the field of machine learning forward. This statement really struck me: "“In the Jeopardy days [in 2011], Watson was running off 90 servers and could store 15 terabytes of memory,” said Gold. That’s not your typical household integration. “Now it’s far more affordable and runs 240 times faster, so we can do far more with less.” Moreover, the size of IBM’s server is now smaller than a pizza box and can fit in any data center." How is this possible?

Scryptonite 13y ago
A stab in the dark would be chips made from Graphene or IBM's Silicene, if I am to understand what you quoted.
wlesieutre 13y ago
"Smaller than a pizza box and can fit in any data center" sounds like marketing for "We put it in a standard rackmount form factor"
ancarda 13y ago
I'm assuming they mean 1 server is a pizza box (ie. it's a 1U rack server), so you'd still need ~90U of space. Either that or IBM has made some breakthroughs in the last 2 years.
- drblast 13y ago
  You can almost hear the article author completely misunderstanding what he was told. I'm betting it was something like, "For Jeopardy, Watson was using 15 Terabytes of data stored on 90 servers. And those servers were large back then, now they're your standard pizza box 1U's. And because of advances in storage technology we can store a lot more in the same space." Isn't there some sort of basic tech literacy class for people who write these articles? It's like reading the sports section and seeing "The Seattle Seahawks are a group of big men and they ran at another group of big men, and they've been doing that a lot, and faster lately. Some of the men in the Seahawks group are smaller and faster, so they are harder to catch by the other group of big men. The Seahawks grouping of men is like a successful business, because they've been winning competitions against other groups."
  - ghshephard 13y ago
    16 Terabytes of RAM on P750s. From http://en.wikipedia.org/wiki/Watson_(computer) http://en.wikipedia.org/wiki/Watson_(computer) "Watson is made up of a cluster of ninety IBM Power 750 servers (plus additional I/O, network and cluster controller nodes in 10 racks) with a total of 2880 POWER7 processor cores and 16 Terabytes of RAM. Each Power 750 server uses a 3.5 GHz POWER7 eight core processor, with four threads per core. "
  - cstross 13y ago
    There is some more prize gibberish in that article: Saxena said IBM is ready to make a big a bet on Watson, as it did in the 1970s when it invested in the emergence of the mainframe. Leaving aside the grammatical error (it's probably just an editing artefact), what bed did IBM make in the 1970s on the emergence of the mainframe? Is that a typo for 1950s? Or just unvarnished ignorance of computing history? It's stuff like this that makes me mistrust a news source: if they get it blazingly wrong about something I know enough about to have a working bullshit detector for, what are the chances that they're not also getting it wrong about everything else?
    - nl 13y ago
      To be fair, the System/360 (which was the first to run the classic OS/360 operating system) was released in '64[1], and was replaced in the '70s by the S/370. Certainly mainframes were in use prior to the '70s, but the S/360 was the first "classic" mainframe in that it was designed for both business & scientific use[2], and it wasn't until the 70's that they began to become more popular in areas outside their traditional use. [1] http://en.wikipedia.org/wiki/IBM_System/360 http://en.wikipedia.org/wiki/IBM_System/360 [2]
- pkaye 13y ago
  The main change from the Jeopardy competitions is they don't need to find an answer in under a second. So you can probably reduce to computitional capability significantly.
bratao 13y ago
I would bet on optimization of the software, lower costs of memory, and database sanitization.
gaadd33 13y ago
I think prior to that it mentions its cloud based, so the head end is probably only 1U.
mdisraeli 13y ago
If we ignore cloud interfacing solutions[1], there is a simple answer. "Enterprise scale" solutions really mean "Enterprise SCALING". Pre-sales for 90 servers in many cases would be impossible. Pre-sales of a single rack mountable item that can then be expanded from depending on the workload is really attractive, however. Write the software right, and it should be able to cope. Pre-sales probably starts with a customer renting a single box to try out, possibly with some professional services time. The customer decides they like it, and commission a project to develop a Watson solution. IBM pre-sales then work out how many servers would actually be needed, once the customer is interested. As the solution can scale, when management goes "that cost looks little high" they can safely say "well, if we reduce it by sixteen servers, you'll be slightly slower but that might be ok. Then if the business decides to increase we can simply add more servers[2] at a later date". And, of course, adding more servers at a later date means more professional services time, meaning more money and more quality customer relationship building time. [1] Which for all kinds of commercial reasons, we have to discount a cloud solution being the default here. Buzzword, yes, but enterprise is slow and many really exciting uses cannot be based in the cloud. [2] Or, y'know, just turn this key....

PaulHoule 13y ago

Why is it always cancer? I'd be impressed if somebody was trying to cure ALS or Tourette's syndrome.

teraflop 13y ago
Maybe because cancer is ridiculously more common than both of those disorders combined?
- PaulHoule 13y ago
  Perhaps cancer in general, but not particularly types of cancer, when you recognize that there are multiple forms of lung cancer, breast cancer, etc. -- common forms of cancer are not a single disease. The trouble with cancer drugs is that you take them for a short time and you either die or you survive. The real money is in drugs that a large segment of the population can take for an indefinite time such as Lipitor, blood pressure meds, asthma inhalers, and the perpetually elusive obesity cure.
- thaumasiotes 13y ago
  Sure, but obviously curing cancer is a waste of research effort... what they should be trying to cure is disease.
  - foobarbazqux 13y ago
    Why is it a waste of research effort? We've made dramatic non-palliative strides in the area of cancer treatment, and most importantly have been able to induce complete remission in a significant number of cases.
    - thaumasiotes 13y ago
      see my sibling comment.
      - foobarbazqux 13y ago

dnautics 13y ago

Instead of Jeopardy, I always wanted Watson to go on "are you smarter than a fifth grader" and throw questions like "think of the second color of the rainbow. What large bird starts with the same first letter?" Or: "If I call A 5, B 4, and C 3, then what do I call D?" As a scientist (working on cancer, no less), more than just churning through terabytes of data, it strikes me that THESE sorts of questions are the ones you'll need to be able to answer. There's also a lot of "reading between the lines" you'll need to do, and a lot of "small pattern extrapolation". For example, the drug candidate I'm working on is one of a family of about let's say, 16 known compounds. All of them kill cancer cells, fourteen are cardiotoxic. The two that are not, lack a single oxygen atom. Maybe that oxygen atom lets the molecule block the mitochondrion (sort of important for heart cells)? Why? Because the molecule family looks sort of like Coenzyme Q10 - if you use your imagination and remember some of your chemical reactivity rules from Grad-level Ochem - and supplementation of Q10 alleviated the cardiotoxicity in one study on rats for one of the molecules. Ultimately, my previous boss was bold enough to latch onto that and give it a shot - can we take this molecule that is the most potent of the group (but unfortunately cardiotoxic), delete the homologous oxygen atom and make something that still works and is less cardiotoxic?

jknightco 13y ago
I'm legitimately stumped. Which large bird starts with 'O'? Or is it second from the top.. so an Ibis?
- ramchip 13y ago
  Ostrich?
  - jknightco 13y ago
    I'm going to go sit in the corner and cry now. Thanks!
- protomyth 13y ago
  Only one I thought of relates to baseball. Didn't think it was large. Edit: for some reason ramchip I keep missing that one when the subject of birds comes up

JunkDNA 13y ago

As far as I'm aware there have been zero peer reviewed publications showing Watson actually doing something in medicine. When a doctor thinks he or she has a better approach than what is being done, they use the scientific method to find out of their hypothesis holds up. So far I've seen lots of IBM conference presentations and press conferences. But where is the data? Where are the increased survival rates? Where are the reduced medical errors compared to controls? That is science. Everything else is snake oil unless you have that.

dbcooper 13y ago
Indeed. It's a little annoying to keep seeing these IBM PR releases about fictional medical suppositions, when one is working on drug screening, diagnostics, and such from within a reality constrained lab!
hershel 13y ago
The way i understand it, this system offers data to physicians ,it doesn't offer treatment, so it doesn't have to be clinically tested. It is the the responsibility of the hospital and doctors to evaluate and see if it offer any benefit to them.That's the usual standard information systems are bound to. But watson is quite new. I'm pretty sure testing will come with time.

whiddershins 13y ago

Even if all some AI system did was take a list of symptoms, anonymize them, suggest a diagnosis, and then take the MD's actual diagnosis, track outcomes, and compare results over time, it would be a huge advancement. Right now, unless I misunderstand something structural about our medical system, most of the data gathered by MDs are wasted, because there is no digital record of much of it and patient privacy legalities prevent wholesale entering of symptoms and case outcomes in to any sort of freely available (inter)national database.

nnq 13y ago

Where can one learn more about how Watson actually works? ...starting from the IBM website and Wikipedia one mostly gets to whitepapers and high level descriptions. I hope at least bits of it are open source or at least have the math and logic underlying them well documented.