IBM's Watson and the future of Medicine

[SeekerOfTheTree]

---

Members do not see this ad.

Remember, it's not an 'algorithm'. Nothing so simple could be used to diagnose so many conditions and variables. That's what makes Watson unique. It's not a pre-determined algorithm. It's a more general purpose problem solving tool based on certain guidelines and strategies. Like a human is (though it works differently).

Algorithm - a precise rule (or set of rules) specifying how to solve some problem

I'm sorry to respectfully disagree here but the IBM folks are trying to sell garbage to the masses. The computer has an algorithm. It has a syntax to understand natural language which in this case would be proper English. It has rules it must follow which language follows. I did this kind of programming before too so I have a little understanding about it. For different languages or lingo it would have to be changed. It doesn't learn from what it sees. It calculates probabilities and chances and wagers based on those results. As you noticed it still gave the same answer after the challenger before him gave it and it was wrong. It is not learning anything. It is not interacting with its environment anymore than a cin or cout command would in C++. They can claim they have found the holy grail of computing but it is algorithm that is designed on the same principle as search engines. More than likely it has a tree database structure where it can break down nouns, adjectives, verbs and perform its queries within its database. I would only agree that there was a leap forward if they built a machine that could understand anyone speaking English. From the Chinese immigrant who came here yesterday to the kid who grew up in Compton. I would be impressed if it understood what they were trying to say even if it wasn’t in perfect English. I would be more impressed if it could have a conversation and react to what I say rather than search for the thing to say.

It is nothing close to a human. That comparison is insulting. I’m much hotter than a box of transistors.

 

[Lokhtar]

The computer has an algorithm. It has a syntax to understand natural language which in this case would be proper English. It has rules it must follow which language follows.

Yes.

As you noticed it still gave the same answer after the challenger before him gave it and it was wrong. It is not learning anything.

That is because according to the rules of the game, it was not fed wrong answers. It was however, fed right answers after someone gave it. In one category, it realized all the right answers were decade, so it took that into account and changed answers late in the category.

It is not interacting with its environment anymore than a cin or cout command would in C++. They can claim they have found the holy grail of computing but it is algorithm that is designed on the same principle as search engines. More than likely it has a tree database structure where it can break down nouns, adjectives, verbs and perform its queries within its database.

No no, it's much more than that. I posted a PBS documentary with a lot more details, of you're interested.

I would only agree that there was a leap forward if they built a machine that could understand anyone speaking English. From the Chinese immigrant who came here yesterday to the kid who grew up in Compton. I would be impressed if it understood what they were trying to say even if it wasn't in perfect English. I would be more impressed if it could have a conversation and react to what I say rather than search for the thing to say.

This has nothing to do either 1) Speech Recognition, or 2) Passing the Turing Test. No one is claiming this is a significant step towards sentience or anything....

 

[SeekerOfTheTree]

Yes.
This has nothing to do either 1) Speech Recognition, or 2) Passing the Turing Test. No one is claiming this is a significant step towards sentience or anything....

That's why I consider it insignificant. Till it can do this it can't replace vital functions in our society that require human interaction.

A the end of the day this thing is a program. Someone wrote it using a language out there or a combination of many that the computer can understand. A feel good special about Watson isn't going to make me be impressed by IBM's techies. Now if this came from DARPA that would be another thing. I would buy it to be more meaningful of a discovery.

 

[Lokhtar]

That's why I consider it insignificant. Till it can do this it can't replace vital functions in our society that require human interaction.

Well, if you notice, neither I or anyone said it'll replace human interaction 🙂.

 

[SeekerOfTheTree]

I'm sorry if I misunderstood but from your tone you imply that the replacement of physicians won't happen just yet but this is a step towards a technology that could bring about that change. My only point is that I don't think this is that stepping stone. Just like IBM Deep Blue this will go down as a pointless machine with limited potential. If you take away it's ability to have human interaction then it just becomes an expensive piece of machine whose results are at the mercy of the person who collects and inputs the information into it. So making it a complicated search engine that can give answers that sound more natural.

So what is so great about Watson other than it being a publicity gimic by IBM?

 

[VTbiochemist]

That's why I consider it insignificant. Till it can do this it can't replace vital functions in our society that require human interaction.

You have to remember that advancement takes place in small steps, you aren't going to go from your every day desktop to a sentient machine. Watson just happens to be a big step, I think that it is pretty significant. Like Lokhtar said earlier, now that people know it can be done, others will do it and improve upon it.

 

[JaggerPlate]

My plan for 2050:

1. Watson's prices drop like those of personal computers (better, faster, smaller, and cheaper) and I pick up 6 - 8 at Best Buy.

2. I open a clinic with 5 - 7 rooms and an office.

3. I put one Watson behind the desk ... this is receptionist Watson. She's awesome, deals with the patients perfectly, never misses a day of work, and is not a sexual harassment risk.

4. I put the rest of the Watsons in the exam rooms.

5. Receptionist Watson welcomes the patients, gives them instructions, and they go to an exam room.

6. The patients go into the room, explain to Watson what's wrong, he ddx, explains what's up, gives them a printout with all the relevant information and tells them to go to my office.

7. The patient enters my office with the print outs ... I'm wearing sweats and watching Dexter on the flat screen (these were on sale at Best Buy as well) when the patient arrives. I open the chart, make sure Watson isn't malfunctioning and printed out some binary junk, review the relevant information, sign off on any necessary scripts, lab orders, etc, and tell them to pay receptionist Watson on the way out.

8. Rinse and repeat ... all day.


Best part - I'm only half kidding.

 

[Lokhtar]

2050? You could probably afford Watson by 2018 - with a much better algorithm too.

Eight years after deep blue vs kasparov, chess programs were capable of beating/matching the best in the world running on personal computers (it was mostly improved algorithms).

 

[JaggerPlate]

2050? You could probably afford Watson by 2018 - with a much better algorithm too.

Eight years after deep blue vs kasparov, chess programs were capable of beating/matching the best in the world running on personal computers (it was mostly improved algorithms).

Awesome ... going through as planned.

 

[Lokhtar]

I'm sorry if I misunderstood but from your tone you imply that the replacement of physicians won't happen just yet but this is a step towards a technology that could bring about that change.

When I say replace, I meant it could better their diagnostic abilities based on information like test results, scans, medical literature and physical (which would have to be done by humans still).

So what is so great about Watson other than it being a publicity gimic by IBM?

No computer has ever been able to do what Watson did by a long shot. Unlike information stored for computer use, Watson reads documents (like wikipedia) that is written in natural language. It then understands the question that is extremely complex (read my earlier post) and parses the natural language pages, weighs millions of pieces of evidence, ranks it based on a multitude of factors and comes up with a single answer with a certain confidence level.

Meaning, you don't have to store data in a database for computers - the computer can simply read the studies published in NEJM or Nature (or wikipedia), and apply that knowledge in answering your question.


It's Wolfram Alpha times about a million because it can handle immense amounts of natural language complexity.

As I said hospital trials are starting in 2012. By the time I graduate and start practicing 🙂xf🙂, I bet this will start to be decent at many tasks (and a while after that for it to rival physicians in certain tasks - again, it can't talk to a patient and ask it symptoms, understand non verbal cues, do a physical, etc....so I don't see it replacing physicians on that front, as I mentioned several times 😛). In the beginning (say 10-12 years), I see it as a backup tool/double check as in 'Hey based on past cases and all the medical literature, there is a 5% chance that this patient has X or we should order Y. Recommend: x'. If x or y is stupid, the physician can ignore it (e.g Toronto moment). But eventually it'll get better and better at that.

 

[SeekerOfTheTree]

You have to remember that advancement takes place in small steps, you aren't going to go from your every day desktop to a sentient machine. Watson just happens to be a big step, I think that it is pretty significant. Like Lokhtar said earlier, now that people know it can be done, others will do it and improve upon it.

I understand that but my point is this is not a leap in any technology. This is not like a microchip array or finding how to do PCR. This is just writing code differently. Something that already exists modified to serve a different purpose, which in this case is entertainment.

I know the computer's predecessors were vacuum tube but comparing Watson in such a manor would be an insult to those pioneers. This thing is nothing more than a glorified search engine with better syntax laws. This thing is not the next step in the computing age.

We have supercomputers that can model nuclear explosions and weather patterns and you guys are impressed because this thing plays Jeopardy well.

 

[SeekerOfTheTree]

When I say replace, I meant it could better their diagnostic abilities based on information like test results, scans, medical literature and physical (which would have to be done by humans still).
As I said hospital trials are starting in 2012. By the time I graduate and start practicing 🙂xf🙂, I bet this will start to be decent at many tasks (and a while after that for it to rival physicians in certain tasks - again, it can't talk to a patient and ask it symptoms, understand non verbal cues, do a physical, etc....so I don't see it replacing physicians on that front, as I mentioned several times 😛). In the beginning (say 10-12 years), I see it as a backup tool/double check as in 'Hey based on past cases and all the medical literature, there is a 5% chance that this patient has X or we should order Y. Recommend: x'. If x or y is stupid, the physician can ignore it (e.g Toronto moment). But eventually it'll get better and better at that.

I am sorry my friend but you are dreaming. I will bet you money (100 dollars, I'm a poor med student afterall) and be willing to pay up by 2012. 2050, maybe there will be something like this but in 2 years, you are dreaming. Plus where do you think this money is going to come from? Did you forget our healthcare system is going broke? You really think hospitals are going to pay a doctor 200,000 a year to go around diagnosing and then on top of that pay probably a million for the machine and then another monthly licensing fee. Please tell me where the business sense in that will be?

I read the article about IBM partnering up to do trials at the hospital; it's more of a publicity stunt. I am not going to go into all the kinks this thing will have trying to search through relevant databases, articles, websites and determining how to resolve conflicting information. Trust me these will not be resolved in 2 years. You would want a failsafe system when this thing is providing you information on treatments to save lives.

The role you are defining for it to serve in the next ten to twelve years is nothing more than a glorified search engine with a better definition around it's syntax for searching. Having this make it into the hospital and someone actually paying for it where its only benefit is it helps docs do less thinking...good luck.

 

[SeekerOfTheTree]

2050? You could probably afford Watson by 2018 - with a much better algorithm too.

Eight years after deep blue vs kasparov, chess programs were capable of beating/matching the best in the world running on personal computers (it was mostly improved algorithms).

You are comparing an algorithm that is designed to go through a matrix to determine all possible moves the user will make and determine a strategy based on that by picking one branch of that tree to being a diagnostician? Interpreting lab results, X-rays, vitals, patient history....right. in 2018 this will happen, right when we are able to make antimatter engines.

 

[Lokhtar]

You are comparing an algorithm that is designed to go through a matrix to determine all possible moves the user will make and determine a strategy based on that by picking one branch of that tree to being a diagnostician? Interpreting lab results, X-rays, vitals, patient history....right. in 2018 this will happen, right when we are able to make antimatter engines.

No, I don't think this will happen in 2018 since Watson can't do that now. I said in ten years they will be backups and give some suggestions but will be wrong a lot more than doctors. I've said that many times that I don't see them becoming 'better' than doctors for many decades, if ever (or ever in the context of my lifetime of hopefully at least 50 more years ).

I understand that but my point is this is not a leap in any technology. This is not like a microchip array or finding how to do PCR. This is just writing code differently. Something that already exists modified to serve a different purpose, which in this case is entertainment.

I know the computer's predecessors were vacuum tube but comparing Watson in such a manor would be an insult to those pioneers. This thing is nothing more than a glorified search engine with better syntax laws. This thing is not the next step in the computing age.

We have supercomputers that can model nuclear explosions and weather patterns and you guys are impressed because this thing plays Jeopardy well.

Well not all breakthroughs are hardware. This is a big breakthrough in programming and pushes the boundaries on what computers can do. I've only spoken to my old comp. sci professor (the guy who gave me a C- btw 👎 😛) and he was quite impressed. However, this is of course n=1, and anecdote is not data, so it can be ignored...and in anycase I'm sure you might be able to find professors who feel the same way you do.

I think I've tried to explain several times why this is not a glorified search engine with better syntax laws (that PBS documentary goes into a bit of detail too).

I simply respectfully disagree with your argument. However, since I can't be proven right or wrong, I suppose we'll just agree to disagree here 🙂. In ten years if I continue to have too much time and a lack of any social life, maybe I'll dig up this thread 😛.

 

[Lokhtar]

Or maybe Watson will dig it up and gloat for itself 😛.

 

[SeekerOfTheTree]

I think I've tried to explain several times why this is not a glorified search engine with better syntax laws (that PBS documentary goes into a bit of detail too)

Lokhtar, I'm not asking you to reference a documentary or find more facts, I just want you to think for yourself what it does and compare it to what we have. Don't get sold by what IBM is spinning, trust me I was once going to work for their program and changed my mind to one of their competitors before my med school days. This to me is moving horizontally in the progress line to serve a different function, not vertically towards new changes.

But on a sidenote I do like you. You are fun to talk to and you are smarter than the average med student. I hope we do dig up this thread in a few years from now. I will be using Watson to help me find robot porn. Oh Yeah!

 

[SeekerOfTheTree]

No, I don't think this will happen in 2018 since Watson can't do that now. I said in ten years they will be backups and give some suggestions but will be wrong a lot more than doctors. I've said that many times that I don't see them becoming 'better' than doctors for many decades, if ever (or ever in the context of my lifetime of hopefully at least 50 more years ).

Now this I agree with.

 

[SeekerOfTheTree]

Or maybe Watson will dig it up and gloat for itself 😛.

If Watson tries to do this to me I will throw water on it, then we will see who is boss. Stupid Watson, always talkin **** behind my back.

 

[DocEspana]

0:15 in: "you don't want your doctor to guess. You want him to have confidence in his answer before he decides to give you a treatment."

This woman clearly has no clue how medicine works. The protocol for meningitis? Give the patient (assuming they can handle both drugs) acyclovir and ceftriaxone immediately. Before even the gram stain is done. Who cares if its viral or bacterial? You for sure don't have any clue. But you gotta make a decent guess and/or just treat for both. So you cover your ass and treat for everything. Once the gram stain, culture, CSF analysis and possible PCR are done you can discontinue the one or both of those that are redundant.

The first time watson diagnoses viral meningitis and its miliary tb he's in some deep doo doo (mind you, not that either of the drug regimens listed above would do anything for miliary tb, but at least humans would be pumping you full of drugs as is, so they'd have no problem switching you over to a rifampin and isoniazid bath when they realized they were wrong. Watson would continue acyclovir and/or supportive treatments.)

 

[lachlan]

anyone know when watson will be on jeopardy again? this was so fun to watch!

 

[TDX]

Thats the way i feel. It will hapen eventually, but not for us.
Watson's CPU fills and entire room. it looks more like the first computers that were invented. What you saw on screen was just the monitor. Its going to be a while till they can fit all those processors into a manageable size enough to make wide use in different fields feasible.

Desktop computers today have the same power as room-sized computer had 20 years ago. At the rate things are going right now, it will be less than 30 years before mobile phones have Watson-like power.

Keep in mind, it was only about 15 years ago that a computer first won a chess match against Kasparov. Today, mobile devices can consistently perform on a level comparable to that of grandmaster chess players.

 

[lildave2586]

Watson won't be able to perform surgery.

 

[Strontium]

Related:

http://www.physorg.com/news/2011-03-watson-ability-illness.html

 

[Lokhtar]

http://news.yahoo.com/s/ap/us_doctor_watson

 

[HH Holmes]

I just saw this on Yahoo! and they (IBM) are now saying that Watson can synthesize diagnoses in seconds due to previous case files, medical literature and other online sources. I'm calling bullsh*t, but if so, medicine is DOOMED! 😛

 

[Orbo]

I am 87.6% confident that Watson will change the face of medicine in 12.64 years.

 

[shaggy alfresco]

I just saw this on Yahoo! and they (IBM) are now saying that Watson can synthesize diagnoses in seconds due to previous case files, medical literature and other online sources. I'm calling bullsh*t, but if so, medicine is DOOMED! 😛

Well it depends. Who knows what will happen in the future but right now it'd be a tool to aid, not replace. You still need someone inputting all the data into Watson.

 

[deleted746658]

bump.

can anyone comment on watson's status/ timeline for integration into the medical world?

 

[Rekt]

bump.

can anyone comment on watson's status/ timeline for integration into the medical world?

Pretty simple. Never.
Patient's lie. Sometimes they just want drugs or want a place to stay. Sometimes they present with atypical symptoms or are asymptomatic. The machine can't predict anything with that. Not to mention, being a physician is mostly managing cases. Every patient isn't an episode of house.

 

[deleted746658]

Pretty simple. Never.
Patient's lie. Sometimes they just want drugs or want a place to stay. Sometimes they present with atypical symptoms or are asymptomatic. The machine can't predict anything with that. Not to mention, being a physician is mostly managing cases. Every patient isn't an episode of house.

never? are you serious? you're clearly uninformed. watson will have a serious role as part of the healthcare team, just a matter of when.

 

[PL198]

never? are you serious? you're clearly uninformed. watson will have a serious role as part of the healthcare team, just a matter of when.

whole architecture of society will be so different by then that there's no use speculating about it. you think docs will be the first one to be replaced? what about garbage-men, fast food workers, store clerks, etc etc. there are so many incredibly low skill jobs that will be easily replaceable and result in massive shifts in employment, before machines are ever good enough to do healthcare.

the biggest obstacle is liability. who is liable when machines make mistakes? this is also biggest obstacle of autonomous cars. they are going to have to be drastically superior to human drivers in accident rates before those liabilities being on the manufacturer will be worthwhile for the manufacturers to take on that immense responsibility.

 

[DRsmartypants]

Here is my two cents
First I will list my medical concerns as to why I don't think it will work(by itself), secondly I will list the political ones.
Watson is right now being trained in medicine and radiology(this is first because it is image based)
1.Patients--Assuming this would have to be tried and put into a clinical study obviously before it was implemented in any fashion, informed consent would be required obviously. As someone who has had many radiology studies before, I cannot imagine wanting an unconfirmed computer(even backed by a radiologist in case it doesn't work) to "read" my scan. To what I know of radiology, at least half of radiology is not about reading the scan, but about interpreting it. It is not the hardest thing in the world to see a nodule and decide what it is. What makes the radiologist worth something is his clinical recommendation, what the next steps should be, are there other possibilities?
-To my knowledge, the watson ability is mainly to make very straight foward diagnoses, ones NPs or PAs could probably make if they were trained in it.
-I do think watson has potential to be an AID to patient care, but not a replacement.

2. Politics, the AMA, congress will never allow watson to replace doctors because they would be loosing millions of dollars in tax revenue from doctors and the very idea of your congressman replacing your doctor with a machine will not go over well.

 

[deleted746658]

i think most should be more concerned than you two are. i think most PCP's could struggle to find work. watson + midlevel... yikes

 

[PL198]

i think most should be more concerned than you two are. i think most PCP's could struggle to find work. watson + midlevel... yikes

again, if they can replace doctors, they're going to replace a whole lot more professions as well and likely before. so I can't really fathom the landscape at that time to think how I would respond.

 

[Psai]

i think most should be more concerned than you two are. i think most PCP's could struggle to find work. watson + midlevel... yikes

Watson would replace a midlevel earlier than it would replace a physician

 

[Rekt]

again, if they can replace doctors, they're going to replace a whole lot more professions as well and likely before. so I can't really fathom the landscape at that time to think how I would respond.

This. We'll be in a mid-dystopian generation by the time machines can replace physicians due to the sheer amount of lost jobs due to automation.