AI tools cannot catch covid. Only humans can.
Don't forget bats.
Don't forget Chinese bats that Fauci paid millions to study bio-terrorism (excuse me, the p.c. term is "gain of function") on.
AI tools have been built to catch Covid, but none of them helped
21–30 of 136 posts
Re: AI tools have been built to catch Covid, but none of them helped
#22Earlier quoted context omitted.
You think a 100% failure rate on the hottest topic in the world right now is likely?
> You think a 100% failure rate on the hottest topic in the world right now is likely? Yeah. Attempts at powered flight had a 100% failure rate in the 1800s. It's kind of absurd to think that "AI" (especially in its current incarnation) must be able to solve any "hot" problem that it's thrown at.
You think AI is so primitive that we can't make any headway into the biggest problem we currently have? That seems like a fairly fringe view. Especially on a problem that has well defined data like medical imaging.
Re: AI tools have been built to catch Covid, but none of them helped
#23The silly thing here is that it’s very easy to train models nowadays, so the medical AI space is very noisy.
With the right data and the right application, this technology can definitely be helpful.
However, it is very hard to sort out signal from noise since any reasonably competent undergrad can throw a convnet at some images and get an AUC.
Actually, I have even seen a high school science fair where they did a medical computer vision application.
I think this speaks more to the accessibility and ease of use of the technology that is out there than it does to the inherent limitations of the tools we have.
I am reasonably confident that there is a group out there that was more careful and has something that might have been useful, but it’s just impossible for clinicians to know who to trust.
I feel that both uncritical, un-nuanced hype (Radiologists will be replaced in 5 years!) and uncritical, un-nuanced closed mindedness (AI will never be clinically useful) is harmful for tackling the biggest problem the field faces right now: trust.
Who to trust. When to trust.
Re: AI tools have been built to catch Covid, but none of them helped
#24Earlier quoted context omitted.
> You think a 100% failure rate on the hottest topic in the world right now is likely? Yeah. Attempts at powered flight had a 100% failure rate in the 1800s. It's kind of absurd to think that "AI" (especially in its current incarnation) must be able to solve any "hot" problem that it's thrown at.
You think AI is so primitive that we can't make any headway into the biggest problem we currently have? That seems like a fairly fringe view. Especially on a problem that has well defined data like medical imaging.
Medical diagnosis has been one of the core ai domains domains since the early 1960s so people have been breaking the picks at that coalface for almost 60 years.
And those older systems has more intelligence in them. Todays’s “AI” is a small number of tricks aimed at large amounts of data, with an unprecedented and enormous amount of marketing added.
Re: AI tools have been built to catch Covid, but none of them helped
#25> “The models are so similar—they almost all use the same techniques with minor tweaks, the same inputs—and they all make the same mistakes,” says Wynants. “If all these people making new models instead tested models that were already available, maybe we’d have something that could really help in the clinic by now.” Sounds about right.
I'm curious what people find is more likely given this evidence: - 100s of models were done incorrectly, and none even remotely correctly - there is no phenomenon to correctly model Both seem extremely unlikely. Not sure what the alternatives are.
100s of DL groups with little previous experience in this area set out to train similar models at the outset of the pandemic. Many rushed out models. I’m not surprised that they ran into similar pitfalls.
Re: AI tools have been built to catch Covid, but none of them helped
#26Yep, that's because they didn't use a blockchain. That would have solved everything.
Not sure why this was downvoted.
Re: AI tools have been built to catch Covid, but none of them helped
#27> “The models are so similar—they almost all use the same techniques with minor tweaks, the same inputs—and they all make the same mistakes,” says Wynants. “If all these people making new models instead tested models that were already available, maybe we’d have something that could really help in the clinic by now.” Sounds about right.
I'm curious what people find is more likely given this evidence: - 100s of models were done incorrectly, and none even remotely correctly - there is no phenomenon to correctly model Both seem extremely unlikely. Not sure what the alternatives are.
> there is no phenomenon to correctly model
COVID exists, it causes physical changes and affects the world outside the host. 100% it can be modelled somehow in theory.
That being said, there are harsh rules against collecting data (for example, it would be extremely illegal to create a sample set by taking a random group of people and infecting half of them with COVID on purpose). So I expect actually getting training data is a huge hill to climb. If most of the models used the same training data I assume there is a reason like that.
So in practice, I don't expect a model quickly.
Re: AI tools have been built to catch Covid, but none of them helped
#28The silly thing here is that it’s very easy to train models nowadays, so the medical AI space is very noisy. With the right data and the right application, this technology can definitely be helpful. However, it is very hard to sort out signal from noise since any reasonably competent undergrad can throw a convnet at some images and get an AUC. Actually, I have even seen a high school science fair where they did a med…
> I am reasonably confident that there is a group out there that was more careful and has something that might have been useful
What is that confidence based on? It seems like the researchers here made a reasonably thorough effort to find "something that might have been useful" and only turned up two that were even worth further investigation:
"She and her colleagues have looked at 232 algorithms for diagnosing patients or predicting how sick those with the disease might get. They found that none of them were fit for clinical use. Just two have been singled out as being promising enough for future testing."
So are you reasonably confident that these researchers just didn't look hard enough to find the good stuff?
Re: AI tools have been built to catch Covid, but none of them helped
#29Earlier quoted context omitted.
> You think a 100% failure rate on the hottest topic in the world right now is likely? Yeah. Attempts at powered flight had a 100% failure rate in the 1800s. It's kind of absurd to think that "AI" (especially in its current incarnation) must be able to solve any "hot" problem that it's thrown at.
You think AI is so primitive that we can't make any headway into the biggest problem we currently have? That seems like a fairly fringe view. Especially on a problem that has well defined data like medical imaging.
Hey alexa win the war on drugs.
Re: AI tools have been built to catch Covid, but none of them helped
#30The silly thing here is that it’s very easy to train models nowadays, so the medical AI space is very noisy. With the right data and the right application, this technology can definitely be helpful. However, it is very hard to sort out signal from noise since any reasonably competent undergrad can throw a convnet at some images and get an AUC. Actually, I have even seen a high school science fair where they did a med…
> I am reasonably confident that there is a group out there that was more careful and has something that might have been useful What is that confidence based on? It seems like the researchers here made a reasonably thorough effort to find "something that might have been useful" and only turned up two that were even worth further investigation: "She and her colleagues have looked at 232 algorithms for diagnosing patie…
Why not? Plenty of researchers aren’t thorough. I’m not saying that’s the case here but it’s entirely plausible from what I know about academia and wouldn’t be a crazy claim to make.