Slashdot Mirror
IBM Pitched Its Watson Supercomputer as a Revolution in Cancer Care. It's Nowhere Close (statnews.com)
IBM began selling Watson to recommend the best cancer treatments to doctors around the world three years ago. But is it really doing its job? Not so much. An investigation by Stat found that the supercomputer isn't living up to the lofty expectations IBM created for it. It is still struggling with the basic step of learning about different forms of cancer. Only a few dozen hospitals have adopted the system, which is a long way from IBM's goal of establishing dominance in a multibillion-dollar market. And at foreign hospitals, physicians complained its advice is biased toward American patients and methods of care. From the report: The interviews suggest that IBM, in its rush to bolster flagging revenue, unleashed a product without fully assessing the challenges of deploying it in hospitals globally. While it has emphatically marketed Watson for cancer care, IBM hasn't published any scientific papers demonstrating how the technology affects physicians and patients. As a result, its flaws are getting exposed on the front lines of care by doctors and researchers who say that the system, while promising in some respects, remains undeveloped. [...] Perhaps the most stunning overreach is in the company's claim that Watson for Oncology, through artificial intelligence, can sift through reams of data to generate new insights and identify, as an IBM sales rep put it, "even new approaches" to cancer care. STAT found that the system doesn't create new knowledge and is artificially intelligent only in the most rudimentary sense of the term.
Until now.
I'm curious, after a few dozen hospitals adopted Watson, did the "revolution" it created cause a negative impact to the massive amounts of revenue created by maintaining the status quo within the Cancer Industrial Complex?
If so, you have your answer as to why adoption would die off faster than someone mentioning a cure...
And at foreign hospitals, physicians complained its advice is biased toward American patients and methods of care.
Are you seriously telling me that they sold a multi-million dollar machine and didn't even include a goddamn machine learning step to adapt to local variations? Aren't the IBM guys supposed to be experts? Or at least guys that know how to pick up a fucking phone and dial an expert?
This is the kind of rookie mistake I see in my undergrads...
I sure hope I'm reading this wrong, because it sounds like people might die from maltreatment over this.
At its heart, Watson for Oncology uses the cloud-based supercomputer to digest massive amounts of data — from doctor’s notes to medical studies to clinical guidelines. But its treatment recommendations are not based on its own insights from these data. Instead, they are based exclusively on training by human overseers, who laboriously feed Watson information about how patients with specific characteristics should be treated.
Ahh I guess I was wrong. There is no machine learning at all yet.
In the case of Watson for Oncology, those human operators are a couple dozen physicians at a single, though highly respected, U.S. hospital: Memorial Sloan Kettering Cancer Center in New York. Doctors there are empowered to input their own recommendations into Watson, even when the evidence supporting those recommendations is thin.
But hey, looks like the dying part could be correct. I only hope those doctors know what the variations across the world requires, because they will be giving recommendations both for japanese highschool girls and African village elders without even knowing it, and I don't think those groups have the same contextual issues.
Despite the hype, computers don't learn anything. They run programs in the same way they always have. The term "AI" is just hype to attract venture capital and avoid doing real work.
I know it's AI Doom and Gloom, but Winter is Coming
The people at IBM thought it was about "caring about people who are cancer (astrological sign)".
#DeleteFacebook
FTA: https://www.statnews.com/2017/...
Pilar Ossorio, a professor of law and bioethics at University of Wisconsin Law School, said Watson should be subject to tighter regulation because of its role in treating patients. “As an ethical matter, and as a scientific matter, you should have to prove that there’s safety and efficacy before you can just go do this,” she said.
Norden dismissed the suggestion IBM should have been required to conduct a clinical trial before commercializing Watson, noting that many practices in medicine are widely accepted even though they aren’t supported by a randomized controlled trial.
“Has there ever been a randomized trial of parachutes for paratroopers?” Norden asked. “And the answer is, of course not, because there is a very strong intuitive value proposition. So I believe that bringing the best information to bear on medical decision making is a no-brainer.”
What an ass-hole. If the good doctor has a license to practice medicine, it should be revoked! Yes ass-hole they do clinical trails for medical treatments but thanks for the Faulty Comparison...
When captains of industry are talking about cancer treatment in terms of "establishing dominance in a multibillion-dollar market", does any rational person believe we're going to have an actual cure for cancer any time soon?
I've calculated my velocity with such exquisite precision that I have no idea where I am.
For something this critical, I would want the AI to explain itself. If an "expert" tells me that I need to that I need to take a pill for my cholesterol (intentionally choosing something rather minor), I will first ask her why this medicine and how does it work. I will expect to get a cogent explanation before paying for the drug. If I'm being asked to pay thousands of dollars for a cancer treatment, I'd expect someone to explain how the medicine works, and show me how it has been successful in other cases.
There are AI solutions that will show how they arrived at a recommendation. Intel has some AI that uses the feature as a selling point. Why would anyone just say, "Hmm? The computer says to give 'em hypocholoroacetiminophin. Wonder why? Where's my needle.", without getting an explanation?
Aah, change is good. -- Rafiki
Yeah, but it ain't easy. -- Simba
No. There is no such thing as "AI" (yet). It's possible but probably not within our lifetimes. It's taken us well over 30 years since functional MRI came into the market and we're just beginning to understand what general areas of the brain are involved in doing "something", let alone individual neurons and synapses.
To claim the ability to "create intelligence" when we don't even understand the question yet is hubris (and salesmanship) on the side of IBM. Trust me, there will be several boom-bust cycles before AI becomes 'intelligent'.
Custom electronics and digital signage for your business: www.evcircuits.com
From the 2017 Amazon review it appears that all of the URLs in the text are now bad, after one year. Not sure how severely that impacts the value of the book but the reviewer only gave it two stars. So that "four star" average rating should be taken with a grain of salt. I certainly would not buy it new, you can get it for a little more than half the (allegedly) discounted Amazon new price.
Starships were meant to fly, Hands up and touch the sky - Nicky Minaj
First of all, Jeopardy Watson has almost nothing to do with the Watson product IBM is selling. That system was largely NLP based and the team disbanded afterwards. From what I can tell, the only thing still alive from that project is Apache UIMA. Watson isn't even a single product. It's a collection of about a dozen disjoint products with the word "Watson" in front of their name.
The current iteration of "Watson" is not interesting at all. Their machine learning portion is just SPSS. Their next-gen machine learning is just Apache Spark. Their UI to setup hardware and submit spark jobs is very unreliable. When you get an error, it's generally a "An error has occurred" or something just as useless.
Jeopardy Watson was interesting, but the big players are doing much more interesting things these days. Google and Microsoft have very good public machine learning and AI platforms. Amazon's is OK, nothing special. If you want to work at a lower level, Stanford maintains a set of libraries with implementations of their cutting edge algorithms. Especially their NLP group. Theirs are actually user-friendly compared to many other research level projects.
Most of the complaints are that you have to give Watson the answer first and then it gives you the question. Doctors were hoping for something the other way around.
And looking at the contents of the book with Amazon preview it looks like it is entirely about using Watson's services, instead of being about Watson's architecture, implementation or technology - on other words, understanding anything about Watson itself.
In that context, where web access is the whole point of the book, bad URLs would make it next to useless.
Starships were meant to fly, Hands up and touch the sky - Nicky Minaj
Criemer, could you please stop posting these. Just because you're posting anon doesn't mean we don't know it's you.
You, Sir, are one of the few people I've ever encountered on the Internet who sounds like they actually understand this subject.
AI has been used to refer to computer controlled opponents in video games for so long, that is what I associate AI with. A program that can react according to very specific data sets. Otherwise just brain dead stupid. Makes me appreciate the level of intelligence my pets have. When I think of AI, I might think of Siri, or Cortana (Microsoft, not UNSC.), or Alexa, or Google's assistant. I don't think of AI in real life as a Star Wars droid, K.I.T.T., or Skynet, or hologram like Voyagers Doctor, or Halo's Cortana, or Ultron or Vision. Iron Man's Jarvis or Star Trek's "computer" seem to be idealized AI.
First wave. This is how it starts. We learn from this. The question is how much of a diagnosis requires the human element? Are there clues from reading body language or other non-verbal cues that the human gut/intuition would diagnose against? Or does it truly all boil down to data sets, and getting accurate data? Because a machine can hold a larger repository of data, and cost less to maintain than army of doctors, and handle the same number of patients. Do doctors perform better when following a system? Systems favor AI implementation. Or do doctors work better "cowboying" it, and shooting from the hip? Human intuition is much more accurate than randomly generated reactions from an AI.
Search on "oncologists would not have chemotherapy".
Boosting the bodies own defenses against cancer in various ways (including nutrition, intermittent fasting, immune-system tuning, etc.) is another approach at least generally without negative side effects -- wonder if Watson has been fed enough alternative data to recommend it (especially for prevention)?
Example: https://www.drfuhrman.com/lear...
"Cancer screening is promoted as preventive health, and while this may detect early forms of cancer so it can be treated earlier, it does not prevent the development of cancer and has minimal effects on reducing cancer deaths. A Nutritarian diet has the power to repair defects that can lead to cancer, detoxify carcinogens, cause cancer cell death, cut off blood supplies to growing tumors , and stimulate the immune system to recognize, repair abnormalities, and even fight and kill cancer cells. The vitamins, minerals, phytochemicals, and antioxidants found in a diet rich in vegetables, fruits, beans, nuts, and seeds is the key to prevention and even can play an important role in the treatment of various cancers."
Good luck with your own health care choices. It is hard to wade through all the conflicting information and conflict-of-interest. I wanted to make free software to help people make sense of conflicting health information -- but just not enough time given a need to earn money in other ways. What I could do with Watson hardware and that project's budget... (When I was at IBM Research around 2000 I proposed making an interactive display wall powered by an AI-like system to help people make complex decisions and better designs -- but as a contractor the idea did not go that far beyond a proof-of-concept with nine old Thinkpads that looked a lot like a Jeopardy screen, made when my supervisor went on a long vacation...)
https://web.archive.org/web/20...
https://github.com/pdfernhout/...
A 21st century issue: the irony of technologies of abundance in the hands of those still thinking in terms of scarcity.
all posts about AI are completely overblown.
Machine learning algorithms can be powerful when used on a narrowly focused problem or goal, but curing cancer is definitely not a narrowly focused problem. There are aspects of machine learning that might prove to be useful but it's not a catch all solution whereby we simply say, "Hey Watson, cure cancer" and expect it to churn out a cure. What is sounds like they are doing is inputting doctor notes and hoping for Watson to apply treatment plans based on historical success of past similar patients. Firstly, let me say that that approach is flawed due to the simple fact that it's data can be biased or inaccurate. Without unbiased inputs machine learning is useless. Their approach is an attempt to deliver a quick marketing gimmick in order to profit from it. A better approach would be to find measurable unbiased inputs and use the outputs based on those findings to build on top of a larger neural network. Your average computer scientist would be able to write the code but they probably don't have the Medical expertise to make it useful. They would need to work with experts in the Oncology field to solve simple problems with a larger goal in mind. A deeper understanding on the subject might find that mapping genetic sequences and modeling chemical reactions based on a brute force of all possible chemical compounds would be the way to go. However, machine learning wouldn't need to be used in this situation and current computing power wouldn't be able to process usable results within a realistic time frame, unless of course you consider quantum computers which last I checked were still in their infancy.
Raises the question if AI is sufficiently advanced, and actually is able to think for itself -- will it operate in a kind of a black-box fashion, and if so, how will we know it's arrived at 'AI' ?
Well, if it's a black box without input or output, then you can say whatever you want about the computer, it doesn't make any difference whether it has if it doesn't "do" anything.
To arrive at "AI" you need to be able to evaluate it's inputs and it's outputs and see whether they are sufficiently intelligent. As I said, we haven't defined intelligence ourselves yet but it seems to involve being able to make progressively 'better' decisions based on past experiences, and we know we can do that (somewhat). The other factor in intelligence is being able to act with foresight, being able to predict and we can make statistical predictions already fairly accurately.
The final point about intelligence is being able to "apply" what you've learned in the past to a totally new data set and be correct about it most of the time but also discard data either new or old if it doesn't fit a particular model. That's what Watson and pretty much every other "AI" out there has trouble with. It can act statistically perfect on pretty much any data set, but give it incomplete or brand new information that does not closely fit anything it has seen before and it will go haywire, acting upon the new information and even classifying things as correct/incorrect which, when you keep iterating on it, you just start getting garbage, intelligence is able to discern and adapt on such information.
Our brain doesn't actually store a whole "lot" of information. It stores models and transformations, a perfect example is when you remember something, colors, cars etc. can change from memory to memory which implies we're not storing a picture of the car in our brain but rather a bunch of pointers to models of cars we have in our brain and over time this information ages out and/or gets replaced and/or optimized. Our brain also doesn't do a supercomputer worth of information processing, we're just very good at applying various reduction filters and compression and then when we have to recall or reclassify, we need to do minimal work.
Once you can run a model like "cancer research" on a desktop computer, you will be able to talk about an 'intelligence'.
Custom electronics and digital signage for your business: www.evcircuits.com
An AI system delivering well below what was promised. Must be a world first.