This sounds like one of those dumb ideas that government services come up with when they're forced to privatize.
In come the corporate consultants, they toss around buzzwords, they brainstorm, they come up with brilliant ideas, and they get paid and get out before anyone can see whether their advice worked.
Well, I can't think of anyone who understands public health better than a pair of economists funded by a right-wing free-market anti-regulatory think tank. http://perc.org/about-perc/perc-board
Free market good, government bad. What more do you need to know?
They far outperform humans in finding the best, likeliest match given the current symptoms and the response to treatment so far.
You have still not given an example of a program that outperforms humans.
I know something about the problems in diagnosing necrotizing fasciitis. As a couple of my medical textbooks say, it's difficult to diagnose early, before a lot of the damage has been done, because there are no clear symptoms. There are just vague, generalized symptoms, the kind that patients get every day, and one in 100,000 turns out to be necrotizing fascitiis. The specific symptoms appear when the infection is already widespread and they have to do emergency surgery.
The reason there are vague, generalized symptoms is that those are the symptoms caused not by the specific disease, but by the body's immune response, primarily interferon.
Doctors have checklists and algorithms already, if they need them. They don't need to put them in a computer. They don't have to look it up in a computer. There are no checklists or algorithms for identifying necrotizing fasciitis early.
No one claims that this program would have diagnosed anyone with necrotizing fasciitis sooner. There is no pattern for a computer to find.
The big question for me is whether his daughter was vaccinated for chicken pox. The guidelines say that infants should have 2 shots, starting at age 1 year. If Jason Maude didn't vaccinate his daughter against chicken pox, then he is responsible for her injuries.
It's interesting that the program was written by a money manager. I don't have expertise in financial trading, but I think the record is that money managers have written lots of algorithms that turned out not to work. They think that somewhere in those stock charts there's a pattern that will enable them to make money, but the best traders still underperform the S&P 500 mutual funds, on average, in my reading.
You said:
We'll continue with this discussion when you learn how to read. In the meantime I see no reason to waste my time with you.
I agree with you. You should shut up. As the Talmud says:
If silence becomes the wise, how much more does it become the foolish?
They far outperform humans in finding the best, likeliest match given the current symptoms and the response to treatment so far.
I said:
I challenge you to give me a citation to a real medical journal article that describes such a program that was actually validated in the real world. Show me 100 patients who were diagnosed with something by a computer that a doctor missed.
You have not done so. Bullshit confirmed.
(1) The NYT article says the opposite of what your'e claiming:
Dr. Henry Lowe, an internist at Stanford University and director of its Center for Clinical Informatics, doubts that a computer could ever replace a diagnostic wizard like Dr. Dhaliwal, or even a competent clinician.
“Designing computer systems that work well with incomplete or imprecise information is challenging,” Dr. Lowe said. “Particularly in medicine, where the consequences of defective decision-making may be catastrophic.”
(2) NYT article doesn't say anything about DiagnosisPro or Isabel Health Care.
(3) The NYT isn't a medical journal.
(4) Since you didn't give me a medical journal article, I tried to find one myself, to see if -- possibly -- there actually was something to your claims. I don't like to dismiss things merely because they sound ridiculous, I need to see the evidence.
When I searched for "DiagnosisPro" in PubMed, http://www.ncbi.nlm.nih.gov/pubmed/?term=DiagnosisPro which is the best medical database available, I didn't get any results. Interestingly, when I did a Google search, I did find a couple of aticles in medical journals. That shows you that even the best medical database -- Medline on PubMed -- is incomplete. (Actually, DiagnosisPro itself uses Medline.)
(5) Here's a review of DiagnosisPro that I found with a Google search:
I think that this program represents a useful advance over the previous version, 4.0. The greatest utility of the program will be to advance the learning curve of neophyte physicians or medical students; however, even an experienced physician contemplating a difficult case can meaningfully use this program. The program is comprehensive; after becoming familiar with it, the user realizes that the program can be used to expand the scope of differential diagnosis for any patient.
The practicality of finding time to use the program and to reflect on the depth and profundity of its information is another issue, however. Limited time is available to see a patient in the office or even during inpatient rounds: Considering the patient's diagnosis, ordering laboratory tests, prescribing treatment, and discussing these matters with the patient is often done in less than 15 minutes. To use a computer program to review one case and to research the diagnostic possibilities takes considerably longer, although with increased familiarity--especially if quickly viewing the leading differential diagnostic possibilities--the clinician may find this program excellent for the purpose.
(6) You said:
They far outperform humans in finding the best, likeliest match given the current symptoms and the response to treatment so far.
That's the issue. DiagnosisPro doesn't outperform humans. What the doctors who review it say is that it's interesting. It's a good learning tool for medical students. It's interesting to try it on a difficult case and see what it does. However, DiagnosisPro itself doesn't claim to outperform humans.
On their website, they give a bibliography of the sources they incorporate for their diagnoses. ht
but in fact your use of Hollywood-style medibabble indicates that you haven't really bothered to learn much about the subject
The use of babble is a lame attempt at a joke and it indicates nothing. It truly shows your lack of arguments that you have to base your reply on that irrelevant detail.
No, your medibabble indicates that you don't understand the subject. Doctors make jokes all the time. But a doctor's joke, like a programmer's joke or an engineer's joke, is based on an insight and understanding of the discipline. Your medibabble is based on ignorance.
I often see lame attempts at humor about science written by people who don't know anything about science. It comes off as really stupid. Their message is, "There's all this science stuff that none of us understands. It's OK to be ignorant. Nudge, nudge, wink, wink."
No, it's not OK to be ignorant.
I'm a bioinformaticist, which means I have a pretty good idea how hard it is to model living systems in silico.
Maybe you are confused by the latest, state of the art software packages that try to read radiology directly. Those are still in development.
I'm talking about diagnostic software based on expert systems applying deduction rules over noisy data. The doctor enters the data and the computer produces all possible matches with probabilities attached to them. They far outperform humans in finding the best, likeliest match given the current symptoms and the response to treatment so far.
I call bullshit. There is no such diagnostic software. I challenge you to give me a citation to a real medical journal article that describes such a program that was actually validated in the real world. Show me 100 patients who were diagnosed with something by a computer that a doctor missed.
As other people have pointed out, TFA doesn't explain how they determined that their algorithms worked, in other words, how they validated their claims. They don't even define what they mean by "a 30-35% increase in patient outcomes."
Seriously, think about it, who will be better: Dr. House trying to rake his brains for what you could possibly have or a computer with access to a complete database containing every possible disease with their typical AND atypical presentations?
You don't understand how a diagnosis works. There is no complete database of every possible disease with typical and atypical presentations. There couldn't be. Doctors don't even agree.
Here's an example: How do you diagnose prostate cancer? Do you think you can just put PSA tests and pathology reports into an algorithm and get an answer? Some urologists will tell you that you have cancer and need surgery (which has a 50% risk of leaving you impotent and a 50% risk of urinary incontinence) immediately. Other doctors will tell you that your so-called cancer is growing so slowly that you'll die of old age before it ever bothers you. How does your database deal with that?
Since the VA has a very successful automated records system, and the military might have one too, it's possible that the nurse might be able to pull up your medical records. But as a practical matter, a lot of these records will be a hundred pages or more, and no doctor or nurse has time to paw through them. It's easier to just ask you what she needs to know. One of the easy things to do is to automate the appointment scheduling. But automating clinical decision-making is pretty difficult.
I've dealt with triage nurses (and on-call doctors) and they're very good. But an RN doesn't need a computer. They're trained to ask a few perceptive questions, and they can easily decide whether your problem is going to get better by itself, whether you need the next available appointment, whether you need an urgent appointment, or whether you should go to the emergency room immediately.
I once had a book with pages in flow-chart form that was written for laymen to help them decide when they needed a doctor. It was about 150 pages. That's about as much information as it takes. Do you have chest pains? If yes, call 911 immediately. Do you have a cold and sniffles? If yes, drink some tea.
A nurse has all that information in her head. She has to. She has to make decisions immediately. She can't go looking things up on a routine basis on the job.
If they did work, doctors would use them. Doctors would love them. Medical students love iPhones. They use them to store textbooks, to look things up on PubMed, and to download journal articles.
There was an article in NEJM by a medical school professor who said his student solved a difficult diagnosis by looking up the symptoms with a Google search. The professor was crushed.
That works sometimes -- assuming you have an intelligent doctor or medical student doing the search. If you're not a doctor, and you look up your own symptoms on Google, you're liable to decide that you have a fatal disease with 6 months to live. Google hypochondria.
There are fairly complicated tasks in medicine that can now be automated, and it is impressive. But there's no computer system that can duplicate a doctor's thinking as that article claims. Given infinite resources and time, I'm sure there will be, but not in my lifetime.
The linked articles don't tell you what a "30-35% increase in patient outcomes" is, what kind of decisions they made, or how they decided that the algorithm's decisions were better than the doctor's decisions. It seems to be strictly a simulation. They're asserting that their algorithm is better than a doctor's decision, without any supporting evidence.
One of the tip-offs for me is that they don't give the predetermined outcome in the abstract or anyplace else. That's what medical journal articles do. How do they define success? I dunno, maybe computer journals are different.
If somebody could tell me what a "30-35% increase in patient outcomes" is, I'd be very grateful.
I invite anybody to explain that to me. What do they mean by a "30-35% increase in patient outcomes"?
For example, the press release says:
"This was at the same time that the AI approach obtained a 30 to 35 percent increase in patient outcomes," Bennett said. "And we determined that tweaking certain model parameters could enhance the outcome advantage to about 50 percent more improvement at about half the cost."
What does that mean -- "a 30 to 35 percent increase in patient outcomes"? Does that mean the program treated patients with diabetes and got 30% fewer foot ulcers? Or 30% lower blood sugar? Or 30% longer survival? Or did they reduce the weight of overweight patients by 30%? Did they get 30% more patients to stop smoking? Did they diagnose 30% more cases of colon cancer?
They don't seem to have defined their outcomes or endpoints.
This is one of those times when you wish they had to publish in a peer-reviewed journal where an editor would have made them answer some obvious questions.
It looks like an entirely theoretical study. I don't see where they compared their predictions to real-world data. And if they did, how would they decide that they're right and the doctors are wrong?
They're like economists who come up with clever theories that ignore the real world.
This reminds me of the story about the efficiency expert who heard a symphony orchestra. There's nothing here to indicate that they understand anything about medicine.
I have engaged in plenty of arguments with conservatives, on the Wall Street Journal comments pages and elsewhere.
I started out thinking, "They read the Wall Street Journal, how stupid can they be." Boy, was I wrong. I picked out people who seemed to be more intelligent. It was hopeless.
In the area that I know a lot about, health policy, they were repeating Republican and conservative talking points over and over again. I used to link to peer-reviewed journals. They would just dismiss it. ("Hah! The New England Journal of Medicine! That liberal rag!" "Science magazine!...") There have been a lot of studies about conservative thinking in the peer-reviewed literature, and they found that, with conservatives, giving them more evidence merely makes them argue their beliefs more strongly. Conservatives and liberals are not symmetrical. Liberals do that, but significantly less.
The other thing I did on the Wall Street Journal comments pages was to track down their sources and see what they actually said. I was really surprised to see how often their facts didn't hold up. You'd think they'd be concerned about their credibility. Boy, was I wrong.
And for supporting evidence, I repeat that a Google search Chris Mooney is a good place to start.
Why not minimum income? Why not government guaranteed basic income?
Because that would be socialism:
There is no reason why, in a society which has reached the general level of wealth ours has, the first kind of security should not be guaranteed to all without endangering general freedom; that is: some minimum of food, shelter and clothing, sufficient to preserve health. Nor is there any reason why the state should not help to organize a comprehensive system of social insurance in providing for those common hazards of life against which few can make adequate provision.
If you think Obama is breaking the law, give solid examples. If you think he is lying, give facts that attempt to prove your case. If and when your facts are shown to be lacking, acknowledge the fact and come up with a different argument. At the moment the people that don't like Obama are throwing words around like rocks but I never, EVER see any facts coming about.
You sound like somebody who learned how to make a logical argument based on facts and arguments. You assume that everybody else is doing that. But they're not. Right-wing crackpots, who comprise most of the Republican Party, don't understand what a fact is or what an argument is.
They think that things are true because they believe them strongly. They think that argument consists of saying, "Obama is a socialist" and "Good post!" They think that argument consists of shouting louder.
They're like football hooligans who tap into primitive emotions by rooting for their team, getting into fights with the other team, fighting with the cops, and tearing up the town.
Trying to reason with them is a waste of time.
There are a few good psychological studies of conservatives. I would start with Chris Mooney.
a lot of relatively dangerous experiments could be carried out in a virtual lab
Virtual labs, where experiments always work the way it says in the textbook, where nothing unexpected happens, where things are simplified over the real world.
One of the values of doing a lab is that you can do an experiment and actually see it happen. It works. I remember a freshman physics lab in which we took a cardboard tube a meter long, and poured in a kilogram of lead shot (in the days before OSHA). We turned the tube over a hundred times to drop the lead through a meter. Then we measured the rise in temperature. Sure enough, work was converted to heat. And we could estimate the constants.
I could write an instructional program that would let you drop a virtual kilogram of lead through a virtual meter, and raise the virtual temperature. Or it could lower the temperature. It could make smoke and dragons come out. It could do anything I wanted -- it wouldn't have the limitations of the physical world.
That's the point of physics labs in the real universe. They do have the limitations of the physical world. If that lead got warm, it demonstrated a physical principle. The principle worked. It didn't have to work. It worked because kinetic energy was converted to heat. That was the lesson. You can't do that with a virtual experiment.
Not sure about English composition, but there are other subjects that can benefit from technology: visualisation, learning with feedback outside the classroom, gamification... and other than just improving learning effectiveness, could you think of a way where technology could help a teacher effectively teach a class of 1000 rather than 30 or so? Or reduce the cost of learning so you can justify the expense for a far larger group? I can... and I am not the only one. We're not there yet, though.
Teach a class of 1000 rather than 30? In a class of 30, a teacher can get to know every student by the end of the year. Students get to know each other. A class of 1000 is an assembly line. It's a mob. What's your measure of success? Students per dollar?
I took a class in modern poetry, and I still remember a guy who was a car designer, who was taking classes in his retirement. He would tell us obscure things about poems by Wallace Stevens and Ezra Pound that in the news when the poems were written. In my freshman humanities course, one guy was an atheist. One guy was a Jesuit-educated Catholic. There were marxists and army veterans. After a while you could get to know how these people approached the world.
I also took lecture classes of 300 in physics. The teacher basically read his notes. He answered questions, but it wasn't the same.
Humans evolved in the last 100,000 years or whatever to deal with each other in family-sized groups of about 6 to 30. You can't have the same kind of communications and interactions in groups much larger than that.
I once heard a lecture by James Allen Fox, the criminologist.
He was complaining that the penalties for petty crimes in the U.S. are irrationally severe. The example he gave was, somebody breaks into your car. It's a crime, and he deserves to be punished, but ten years? "Get a life."
He was talking about sending somebody to jail for ten years. You're talking about killing somebody, just because he broke into your car. There's such a thing as proportion. There's no justification for killing somebody just because he committed a minor crime.
There are a lot of people who are going around with guns looking for an excuse to kill somebody. They're called psychopaths. I'd rather be around the car thieves.
The shooter in Philadelphia was convicted, of murder or homicide, and he was sentenced to a long prison term, 20 years I think. It seems a little harsh, but he didn't have any mercy so he didn't get any.
You want to go around shooting people? You're going to wind up in jail. Say hello to George Zimmerman.
When I studied history in college, our textbooks and source books had lots of stuff about the arguments for and against God. Considering how many people have been killed over those arguments. it's hard to study history without knowing what they were killed over.
I think it's pretty easy to prove that God doesn't exist in a science class. The Greek philosophers came up with an argument 2500 years ago. He's supposed to be omnipotent and benevolent. But if he exists, he permits all this evil in the world. Therefore, either he's not omnipotent and he's not God), or not benevolent (and he's evil). QED there is no God.
Slashdot Mirror
This sounds like one of those dumb ideas that government services come up with when they're forced to privatize.
In come the corporate consultants, they toss around buzzwords, they brainstorm, they come up with brilliant ideas, and they get paid and get out before anyone can see whether their advice worked.
These guys are economists. They're not going to say anything against money.
Well, I can't think of anyone who understands public health better than a pair of economists funded by a right-wing free-market anti-regulatory think tank. http://perc.org/about-perc/perc-board
Free market good, government bad. What more do you need to know?
Charlie Brown: Where did you get those statistics, Lucy?
Lucy: I made them up!
You said, originally:
They far outperform humans in finding the best, likeliest match given the current symptoms and the response to treatment so far.
You have still not given an example of a program that outperforms humans.
I know something about the problems in diagnosing necrotizing fasciitis. As a couple of my medical textbooks say, it's difficult to diagnose early, before a lot of the damage has been done, because there are no clear symptoms. There are just vague, generalized symptoms, the kind that patients get every day, and one in 100,000 turns out to be necrotizing fascitiis. The specific symptoms appear when the infection is already widespread and they have to do emergency surgery.
The reason there are vague, generalized symptoms is that those are the symptoms caused not by the specific disease, but by the body's immune response, primarily interferon.
Doctors have checklists and algorithms already, if they need them. They don't need to put them in a computer. They don't have to look it up in a computer. There are no checklists or algorithms for identifying necrotizing fasciitis early.
No one claims that this program would have diagnosed anyone with necrotizing fasciitis sooner. There is no pattern for a computer to find.
The big question for me is whether his daughter was vaccinated for chicken pox. The guidelines say that infants should have 2 shots, starting at age 1 year. If Jason Maude didn't vaccinate his daughter against chicken pox, then he is responsible for her injuries.
It's interesting that the program was written by a money manager. I don't have expertise in financial trading, but I think the record is that money managers have written lots of algorithms that turned out not to work. They think that somewhere in those stock charts there's a pattern that will enable them to make money, but the best traders still underperform the S&P 500 mutual funds, on average, in my reading.
You said:
We'll continue with this discussion when you learn how to read. In the meantime I see no reason to waste my time with you.
I agree with you. You should shut up. As the Talmud says:
If silence becomes the wise, how much more does it become the foolish?
Remember what we're talking about. You said:
They far outperform humans in finding the best, likeliest match given the current symptoms and the response to treatment so far.
I said:
I challenge you to give me a citation to a real medical journal article that describes such a program that was actually validated in the real world. Show me 100 patients who were diagnosed with something by a computer that a doctor missed.
You have not done so. Bullshit confirmed.
(1) The NYT article says the opposite of what your'e claiming:
Dr. Henry Lowe, an internist at Stanford University and director of its Center for Clinical Informatics, doubts that a computer could ever replace a diagnostic wizard like Dr. Dhaliwal, or even a competent clinician.
“Designing computer systems that work well with incomplete or imprecise information is challenging,” Dr. Lowe said. “Particularly in medicine, where the consequences of defective decision-making may be catastrophic.”
(2) NYT article doesn't say anything about DiagnosisPro or Isabel Health Care.
(3) The NYT isn't a medical journal.
(4) Since you didn't give me a medical journal article, I tried to find one myself, to see if -- possibly -- there actually was something to your claims. I don't like to dismiss things merely because they sound ridiculous, I need to see the evidence.
When I searched for "DiagnosisPro" in PubMed,
http://www.ncbi.nlm.nih.gov/pubmed/?term=DiagnosisPro which is the best medical database available, I didn't get any results. Interestingly, when I did a Google search, I did find a couple of aticles in medical journals. That shows you that even the best medical database -- Medline on PubMed -- is incomplete. (Actually, DiagnosisPro itself uses Medline.)
(5) Here's a review of DiagnosisPro that I found with a Google search:
http://xnet.kp.org/permanentejournal/sum02/diagnosis.html
I think that this program represents a useful advance over the previous version, 4.0. The greatest utility of the program will be to advance the learning curve of neophyte physicians or medical students; however, even an experienced physician contemplating a difficult case can meaningfully use this program. The program is comprehensive; after becoming familiar with it, the user realizes that the program can be used to expand the scope of differential diagnosis for any patient.
The practicality of finding time to use the program and to reflect on the depth and profundity of its information is another issue, however. Limited time is available to see a patient in the office or even during inpatient rounds: Considering the patient's diagnosis, ordering laboratory tests, prescribing treatment, and discussing these matters with the patient is often done in less than 15 minutes. To use a computer program to review one case and to research the diagnostic possibilities takes considerably longer, although with increased familiarity--especially if quickly viewing the leading differential diagnostic possibilities--the clinician may find this program excellent for the purpose.
(6) You said:
They far outperform humans in finding the best, likeliest match given the current symptoms and the response to treatment so far.
That's the issue. DiagnosisPro doesn't outperform humans. What the doctors who review it say is that it's interesting. It's a good learning tool for medical students. It's interesting to try it on a difficult case and see what it does. However, DiagnosisPro itself doesn't claim to outperform humans.
On their website, they give a bibliography of the sources they incorporate for their diagnoses. ht
but in fact your use of Hollywood-style medibabble indicates that you haven't really bothered to learn much about the subject
The use of babble is a lame attempt at a joke and it indicates nothing. It truly shows your lack of arguments that you have to base your reply on that irrelevant detail.
No, your medibabble indicates that you don't understand the subject. Doctors make jokes all the time. But a doctor's joke, like a programmer's joke or an engineer's joke, is based on an insight and understanding of the discipline. Your medibabble is based on ignorance.
I often see lame attempts at humor about science written by people who don't know anything about science. It comes off as really stupid. Their message is, "There's all this science stuff that none of us understands. It's OK to be ignorant. Nudge, nudge, wink, wink."
No, it's not OK to be ignorant.
I'm a bioinformaticist, which means I have a pretty good idea how hard it is to model living systems in silico.
Maybe you are confused by the latest, state of the art software packages that try to read radiology directly. Those are still in development.
I'm talking about diagnostic software based on expert systems applying deduction rules over noisy data. The doctor enters the data and the computer produces all possible matches with probabilities attached to them. They far outperform humans in finding the best, likeliest match given the current symptoms and the response to treatment so far.
I call bullshit. There is no such diagnostic software. I challenge you to give me a citation to a real medical journal article that describes such a program that was actually validated in the real world. Show me 100 patients who were diagnosed with something by a computer that a doctor missed.
As other people have pointed out, TFA doesn't explain how they determined that their algorithms worked, in other words, how they validated their claims. They don't even define what they mean by "a 30-35% increase in patient outcomes."
Seriously, think about it, who will be better: Dr. House trying to rake his brains for what you could possibly have or a computer with access to a complete database containing every possible disease with their typical AND atypical presentations?
You don't understand how a diagnosis works. There is no complete database of every possible disease with typical and atypical presentations. There couldn't be. Doctors don't even agree.
Here's an example: How do you diagnose prostate cancer? Do you think you can just put PSA tests and pathology reports into an algorithm and get an answer? Some urologists will tell you that you have cancer and need surgery (which has a 50% risk of leaving you impotent and a 50% risk of urinary incontinence) immediately. Other doctors will tell you that your so-called cancer is growing so slowly that you'll die of old age before it ever bothers you. How does your database deal with that?
Since the VA has a very successful automated records system, and the military might have one too, it's possible that the nurse might be able to pull up your medical records. But as a practical matter, a lot of these records will be a hundred pages or more, and no doctor or nurse has time to paw through them. It's easier to just ask you what she needs to know. One of the easy things to do is to automate the appointment scheduling. But automating clinical decision-making is pretty difficult.
I've dealt with triage nurses (and on-call doctors) and they're very good. But an RN doesn't need a computer. They're trained to ask a few perceptive questions, and they can easily decide whether your problem is going to get better by itself, whether you need the next available appointment, whether you need an urgent appointment, or whether you should go to the emergency room immediately.
I once had a book with pages in flow-chart form that was written for laymen to help them decide when they needed a doctor. It was about 150 pages. That's about as much information as it takes. Do you have chest pains? If yes, call 911 immediately. Do you have a cold and sniffles? If yes, drink some tea.
A nurse has all that information in her head. She has to. She has to make decisions immediately. She can't go looking things up on a routine basis on the job.
They have been combined and they don't work.
If they did work, doctors would use them. Doctors would love them. Medical students love iPhones. They use them to store textbooks, to look things up on PubMed, and to download journal articles.
There was an article in NEJM by a medical school professor who said his student solved a difficult diagnosis by looking up the symptoms with a Google search. The professor was crushed.
That works sometimes -- assuming you have an intelligent doctor or medical student doing the search. If you're not a doctor, and you look up your own symptoms on Google, you're liable to decide that you have a fatal disease with 6 months to live. Google hypochondria.
There are fairly complicated tasks in medicine that can now be automated, and it is impressive. But there's no computer system that can duplicate a doctor's thinking as that article claims. Given infinite resources and time, I'm sure there will be, but not in my lifetime.
The linked articles don't tell you what a "30-35% increase in patient outcomes" is, what kind of decisions they made, or how they decided that the algorithm's decisions were better than the doctor's decisions. It seems to be strictly a simulation. They're asserting that their algorithm is better than a doctor's decision, without any supporting evidence.
One of the tip-offs for me is that they don't give the predetermined outcome in the abstract or anyplace else. That's what medical journal articles do. How do they define success? I dunno, maybe computer journals are different.
If somebody could tell me what a "30-35% increase in patient outcomes" is, I'd be very grateful.
Those are all very good points. I just spent half an hour going through the articles, press release, and article itself (which is available here http://arxiv.org/abs/1301.2158 http://www.caseybennett.com/uploads/Bennett_AI_ClinicalDecisionMaking__Article_in_Press_.pdf ) trying to figure out how they determined that the program diagnosed patients better than doctors. I couldn't do it. And it didn't look like it was worth another hour of trying to figure it out.
I invite anybody to explain that to me. What do they mean by a "30-35% increase in patient outcomes"?
For example, the press release says:
"This was at the same time that the AI approach obtained a 30 to 35 percent increase in patient outcomes," Bennett said. "And we determined that tweaking certain model parameters could enhance the outcome advantage to about 50 percent more improvement at about half the cost."
What does that mean -- "a 30 to 35 percent increase in patient outcomes"? Does that mean the program treated patients with diabetes and got 30% fewer foot ulcers? Or 30% lower blood sugar? Or 30% longer survival? Or did they reduce the weight of overweight patients by 30%? Did they get 30% more patients to stop smoking? Did they diagnose 30% more cases of colon cancer?
They don't seem to have defined their outcomes or endpoints.
This is one of those times when you wish they had to publish in a peer-reviewed journal where an editor would have made them answer some obvious questions.
It looks like an entirely theoretical study. I don't see where they compared their predictions to real-world data. And if they did, how would they decide that they're right and the doctors are wrong?
They're like economists who come up with clever theories that ignore the real world.
This reminds me of the story about the efficiency expert who heard a symphony orchestra. There's nothing here to indicate that they understand anything about medicine.
I have engaged in plenty of arguments with conservatives, on the Wall Street Journal comments pages and elsewhere.
I started out thinking, "They read the Wall Street Journal, how stupid can they be." Boy, was I wrong. I picked out people who seemed to be more intelligent. It was hopeless.
In the area that I know a lot about, health policy, they were repeating Republican and conservative talking points over and over again. I used to link to peer-reviewed journals. They would just dismiss it. ("Hah! The New England Journal of Medicine! That liberal rag!" "Science magazine! ...") There have been a lot of studies about conservative thinking in the peer-reviewed literature, and they found that, with conservatives, giving them more evidence merely makes them argue their beliefs more strongly. Conservatives and liberals are not symmetrical. Liberals do that, but significantly less.
The other thing I did on the Wall Street Journal comments pages was to track down their sources and see what they actually said. I was really surprised to see how often their facts didn't hold up. You'd think they'd be concerned about their credibility. Boy, was I wrong.
And for supporting evidence, I repeat that a Google search Chris Mooney is a good place to start.
Why not minimum income? Why not government guaranteed basic income?
Because that would be socialism:
There is no reason why, in a society which has reached the general level of wealth ours has, the first kind of security should not be guaranteed to all without endangering general freedom; that is: some minimum of food, shelter and clothing, sufficient to preserve health. Nor is there any reason why the state should not help to organize a comprehensive system of social insurance in providing for those common hazards of life against which few can make adequate provision.
http://en.wikipedia.org/wiki/Road_To_Serfdom#A_role_for_government
Oh, wait a minute. Friedrich Hayek isn't a socialist. He's a free-market conservative.
"Nobody has lost a job in the US because someone else could do the job in another country for less."
Hrmm. Obviously, you've never worked in IT.
Or the manufacturing industry.
Or the garment industry.
If you think Obama is breaking the law, give solid examples. If you think he is lying, give facts that attempt to prove your case. If and when your facts are shown to be lacking, acknowledge the fact and come up with a different argument. At the moment the people that don't like Obama are throwing words around like rocks but I never, EVER see any facts coming about.
You sound like somebody who learned how to make a logical argument based on facts and arguments. You assume that everybody else is doing that. But they're not. Right-wing crackpots, who comprise most of the Republican Party, don't understand what a fact is or what an argument is.
They think that things are true because they believe them strongly. They think that argument consists of saying, "Obama is a socialist" and "Good post!" They think that argument consists of shouting louder.
They're like football hooligans who tap into primitive emotions by rooting for their team, getting into fights with the other team, fighting with the cops, and tearing up the town.
Trying to reason with them is a waste of time.
There are a few good psychological studies of conservatives. I would start with Chris Mooney.
The constitution does not apply to non-citizens, just as our other laws do not.
Where does the text of the constitution say that?
A British popular science lecturer said, "Nothing gets the attention of the audience like an indoor explosion."
pushing IT off on every supposed problem (what was wrong with the classroom that we are trying to fix) does not make things better.
Or as Isaac Asimov said: Need a doorstop? Get a robot with a big foot.
a lot of relatively dangerous experiments could be carried out in a virtual lab
Virtual labs, where experiments always work the way it says in the textbook, where nothing unexpected happens, where things are simplified over the real world.
One of the values of doing a lab is that you can do an experiment and actually see it happen. It works. I remember a freshman physics lab in which we took a cardboard tube a meter long, and poured in a kilogram of lead shot (in the days before OSHA). We turned the tube over a hundred times to drop the lead through a meter. Then we measured the rise in temperature. Sure enough, work was converted to heat. And we could estimate the constants.
I could write an instructional program that would let you drop a virtual kilogram of lead through a virtual meter, and raise the virtual temperature. Or it could lower the temperature. It could make smoke and dragons come out. It could do anything I wanted -- it wouldn't have the limitations of the physical world.
That's the point of physics labs in the real universe. They do have the limitations of the physical world. If that lead got warm, it demonstrated a physical principle. The principle worked. It didn't have to work. It worked because kinetic energy was converted to heat. That was the lesson. You can't do that with a virtual experiment.
Not sure about English composition, but there are other subjects that can benefit from technology: visualisation, learning with feedback outside the classroom, gamification... and other than just improving learning effectiveness, could you think of a way where technology could help a teacher effectively teach a class of 1000 rather than 30 or so? Or reduce the cost of learning so you can justify the expense for a far larger group? I can... and I am not the only one. We're not there yet, though.
Teach a class of 1000 rather than 30? In a class of 30, a teacher can get to know every student by the end of the year. Students get to know each other. A class of 1000 is an assembly line. It's a mob. What's your measure of success? Students per dollar?
I took a class in modern poetry, and I still remember a guy who was a car designer, who was taking classes in his retirement. He would tell us obscure things about poems by Wallace Stevens and Ezra Pound that in the news when the poems were written. In my freshman humanities course, one guy was an atheist. One guy was a Jesuit-educated Catholic. There were marxists and army veterans. After a while you could get to know how these people approached the world.
I also took lecture classes of 300 in physics. The teacher basically read his notes. He answered questions, but it wasn't the same.
Humans evolved in the last 100,000 years or whatever to deal with each other in family-sized groups of about 6 to 30. You can't have the same kind of communications and interactions in groups much larger than that.
Or explosions.
That's why students don't learn science any more. Teachers can't set off explosions in their lecture rooms.
I once heard a lecture by James Allen Fox, the criminologist.
He was complaining that the penalties for petty crimes in the U.S. are irrationally severe. The example he gave was, somebody breaks into your car. It's a crime, and he deserves to be punished, but ten years? "Get a life."
He was talking about sending somebody to jail for ten years. You're talking about killing somebody, just because he broke into your car. There's such a thing as proportion. There's no justification for killing somebody just because he committed a minor crime.
There are a lot of people who are going around with guns looking for an excuse to kill somebody. They're called psychopaths. I'd rather be around the car thieves.
The shooter in Philadelphia was convicted, of murder or homicide, and he was sentenced to a long prison term, 20 years I think. It seems a little harsh, but he didn't have any mercy so he didn't get any.
You want to go around shooting people? You're going to wind up in jail. Say hello to George Zimmerman.
Given that people do steal shit, that doesn't give you the right to kill them.
In a civilized society, that gets you charged with murder.
Capitalists steal stuff all the time. That doesn't give me the right to shoot them. Although there are some anarchists who will disagree.
When I studied history in college, our textbooks and source books had lots of stuff about the arguments for and against God. Considering how many people have been killed over those arguments. it's hard to study history without knowing what they were killed over.
I think it's pretty easy to prove that God doesn't exist in a science class. The Greek philosophers came up with an argument 2500 years ago. He's supposed to be omnipotent and benevolent. But if he exists, he permits all this evil in the world. Therefore, either he's not omnipotent and he's not God), or not benevolent (and he's evil). QED there is no God.
Yes, it means the freedom to put 13 shots into an 18-year-old who is breaking into your car. http://articles.philly.com/2010-09-16/news/24979574_1_gun-hill-philadelphia-police