5 ms·

ML for healthcare. The whole industry is so far behind everything else, it is not even funny. Problem is, Just like Airbnb or Uber, the biggest road blocks are

by screye 7y ago

ML for healthcare. The whole industry is so far behind everything else, it is not even funny. Problem is, Just like Airbnb or Uber, the biggest road blocks are not technological, they are organizational and political. It is entirely possible, we might need some major privacy / security related leaps in ML to convince the medical industry to adopt it fully. ML in movies and rendering. Deep Fakes is just an amateur's tech demo. In a few years, I expect to see render rights actually become a thing. Like, you don't act in the movie, but give a company rights to use a 3d render of you in it. real time and full time holograms / AR. With remote work becoming a thing, I see a huge market for full 3d renders of the person presenting or even completely virtual AR work places where people check in to work. Maybe not for a decade or two, but whoever builds the flagship product will make a ton of money.

hokumguru 7y ago
This is already a thing. In the upcoming video game Death Stranding, Guillermo del Toro and Nicolas Winding Refn both play physical roles without having taken part in the motion-capture or voice acting. They gave the rights for their likenesses to be used without actually taking part.
- burntoutfire 7y ago
  There's an ocean of quality difference between a video game and a (non-CGI) movie.
  - Miraste 7y ago
    There really isn't, any more. A few more polys in the models and a few more rays in the path tracing aren't a big difference. Check out what Blender can do with Eevee, which is more-or-less a realtime engine. Better physics simulations are still a big differentiator, but that doesn't matter for characters. The only reason rendered likenesses don't appear in films more (because they already do, look at Peter Cushing) is that they're still expensive and take a lot of manual work, and both of those factors are decreasing rapidly.
    - phaus 7y ago
      The CGI Peter Cushing is obviously CGI and non CGI Peter Cushing is obviously not. Its impressive but its not remotely close to being there yet. Is there a better example? You can probably make some CGI stills that fool people pretty well, but so far I haven't seen a convincing animated character. Death Stranding is even more glaringly obvious. However to be fair to the devs they aren't under the impression that it looks like anything other than a videogame. A very pretty videogame, but still obvious CGI.
      - Miraste 7y ago
        > CGI Peter Cushing is obviously CGI and non CGI Peter Cushing is obviously not. It's interesting you say that, because I've found there are two camps of people on CGI Cushing: those who say the CGI is blindingly obvious and nowhere near realism, and those who didn't even notice. It is already good enough to fool some of the population. I'm also not saying Death Stranding is photorealistic--unlike Tarkin it wouldn't fool anyone. I'm saying there's nowhere near an "ocean" of quality difference between the best video games and film CGI. Death Stranding isn't an example I'd use because it runs on PS4 and thus lacks the raytracing and high poly counts of state-of-the-art PC games.
achillesheels 7y ago
Years ago I was shared an anecdote of researchers affiliated with Scripps Research were able to predict a cholera outbreak in subsaharan Africa from reports in another country using analytics.
dcolkitt 7y ago
Healthcare tech has been the "next big area" for the past 40 years. Maybe this time is different, but historically it never pans out. Healthcare's an extremely regulated, extremely change-averse area of the economy where entrenched insiders will always close ranks to protect their own.
- loufe 7y ago
  I wouldn't be so sure. My millionaire uncle built his fortune on medical software, and I have a friend who works in the field with a degree just for it (medical technologies).
  - abrichr 7y ago
    I’d love to learn more about your uncle’s story! Can you provide any additional details? What does the software do?
- mtnGoat 7y ago
  the sales lifecycle in healthcare is a tough nut to crack, most healthcare orgs arent reviewing software every year, they are signing multi year contracts and allergic to change(for a number of systemic reasons). Plus the conglomeration of syndicates has made fewer clients to sell to, at least on the higher end). There are companies making fortunes in the space for sure, but I think many just sell to mckesson after reaching a decent size. It is 100% a field that needs disrupting... someone will figure it out and become very wealthy.
- david_shaw 7y ago
  > Healthcare tech has been the "next big area" for the past 40 years. Maybe this time is different, but historically it never pans out. How do you figure that it never pans out? Healthcare providers use a lot more technology today than they did in 1979. From medical devices to billing software to electronic health records, I think this is an area that will continue to grow.
- amatecha 7y ago
  "healthcare tech never pans out" -> BRB while I check my heart rate and take an electrocardiogram reading on my watch... Looks like the [10000 steps] I've been averaging per day are improving my resting heart rate and I see my daily weight scale shows I've lost [5lbs]! :) Yeah, that's kinda-surface-level stuff, but these kind of health metrics are already extremely commonplace and helping open the door to further tech<->healthcare integration.
applecrazy 7y ago
> ML for healthcare. I'm kind of apprehensive about this advancement. There's many ways to mess this up, from intentional dataset poisoning (to encourage patients to get unnecessary procedure) to just terrible errors made with just a shift in data capture methodology. This paper has a lot of detail on how such systems can fail (https://arxiv.org/pdf/1804.05296.pdf https://arxiv.org/pdf/1804.05296.pdf).
- JMTQp8lwXL 7y ago
  People are already encouraged to get procedures or medications they don't need. A data set isn't going to change that.
  - m0zg 7y ago
    Not to mention quarter of a million dead per year due to medical errors. If we could even just reduce the number of medical errors by 10 percent, that'd be HUGE. Source: https://www.cnbc.com/2018/02/22/medical-errors-third-leading-cause-of-death-in-america.html https://www.cnbc.com/2018/02/22/medical-errors-third-leading... Unfortunately, it's a multi-trillion dollar industry with insanely entrenched and deep pocketed players, and the old boy network the likes of which us software people can't even imagine. And that's before you consider the regulatory framework and the cost of compliance with it.
    - zelly 7y ago
      Solving medical errors would be huge. The problem is that it's a long-tailed problem with no clear way to solve with software. Every error is different, and there are already lots of bureaucratic mitigations in place every step of the way. There are no algorithms capable of the general reasoning required to broadly spot all types of errors including de novo ones. The only way to truly guarantee no medical errors is to replace every agent in a hospital with a machine.
      - m0zg 7y ago
        I bet just even solving the errors in diagnostics and/or catching stuff before it's too far gone would be huge in itself. I'm not talking AGI here, just bog standard perceptual stuff: looking at xrays, MRIs, mining medical records for patterns, handing unstructured medical records better, low hanging stuff like that.
carapace 7y ago
Ages ago Glyph (he of Twisted Python fame) said we would get to the point where we can e.g. use the Indiana Jones movies to build a computer model of Indiana Jones and make more movies...
dugditches 7y ago
What you're talking with 'render rights' is a major plot point in: https://en.wikipedia.org/wiki/The_Congress_(2013_film) https://en.wikipedia.org/wiki/The_Congress_(2013_film)
- Mortos3 7y ago
  Also, the Michael Crichton movie Looker: https://en.wikipedia.org/wiki/Looker https://en.wikipedia.org/wiki/Looker
arcadeparade 7y ago
RE: deep fakes, in a few years a movie director will be able to say to his editor: "increase Tom Cruise's rage and facial contortions in this scene by 50%"
al_chemist 7y ago
> give a company rights to use a 3d render of you in it There's reason why actors earn money and it's not due to perfect face or body.
Wh1skey 7y ago
DIY healthcare

by screye 7y ago

hokumguru 7y ago

This is already a thing. In the upcoming video game Death Stranding, Guillermo del Toro and Nicolas Winding Refn both play physical roles without having taken part in the motion-capture or voice acting. They gave the rights for their likenesses to be used without actually taking part.

burntoutfire 7y ago
There's an ocean of quality difference between a video game and a (non-CGI) movie.
- Miraste 7y ago
  There really isn't, any more. A few more polys in the models and a few more rays in the path tracing aren't a big difference. Check out what Blender can do with Eevee, which is more-or-less a realtime engine. Better physics simulations are still a big differentiator, but that doesn't matter for characters. The only reason rendered likenesses don't appear in films more (because they already do, look at Peter Cushing) is that they're still expensive and take a lot of manual work, and both of those factors are decreasing rapidly.
  - phaus 7y ago
    The CGI Peter Cushing is obviously CGI and non CGI Peter Cushing is obviously not. Its impressive but its not remotely close to being there yet. Is there a better example? You can probably make some CGI stills that fool people pretty well, but so far I haven't seen a convincing animated character. Death Stranding is even more glaringly obvious. However to be fair to the devs they aren't under the impression that it looks like anything other than a videogame. A very pretty videogame, but still obvious CGI.
    - Miraste 7y ago
      > CGI Peter Cushing is obviously CGI and non CGI Peter Cushing is obviously not. It's interesting you say that, because I've found there are two camps of people on CGI Cushing: those who say the CGI is blindingly obvious and nowhere near realism, and those who didn't even notice. It is already good enough to fool some of the population. I'm also not saying Death Stranding is photorealistic--unlike Tarkin it wouldn't fool anyone. I'm saying there's nowhere near an "ocean" of quality difference between the best video games and film CGI. Death Stranding isn't an example I'd use because it runs on PS4 and thus lacks the raytracing and high poly counts of state-of-the-art PC games.

achillesheels 7y ago

Years ago I was shared an anecdote of researchers affiliated with Scripps Research were able to predict a cholera outbreak in subsaharan Africa from reports in another country using analytics.

dcolkitt 7y ago

Healthcare tech has been the "next big area" for the past 40 years. Maybe this time is different, but historically it never pans out. Healthcare's an extremely regulated, extremely change-averse area of the economy where entrenched insiders will always close ranks to protect their own.

loufe 7y ago
I wouldn't be so sure. My millionaire uncle built his fortune on medical software, and I have a friend who works in the field with a degree just for it (medical technologies).
- abrichr 7y ago
  I’d love to learn more about your uncle’s story! Can you provide any additional details? What does the software do?
mtnGoat 7y ago
the sales lifecycle in healthcare is a tough nut to crack, most healthcare orgs arent reviewing software every year, they are signing multi year contracts and allergic to change(for a number of systemic reasons). Plus the conglomeration of syndicates has made fewer clients to sell to, at least on the higher end). There are companies making fortunes in the space for sure, but I think many just sell to mckesson after reaching a decent size. It is 100% a field that needs disrupting... someone will figure it out and become very wealthy.
david_shaw 7y ago
> Healthcare tech has been the "next big area" for the past 40 years. Maybe this time is different, but historically it never pans out. How do you figure that it never pans out? Healthcare providers use a lot more technology today than they did in 1979. From medical devices to billing software to electronic health records, I think this is an area that will continue to grow.
amatecha 7y ago
"healthcare tech never pans out" -> BRB while I check my heart rate and take an electrocardiogram reading on my watch... Looks like the [10000 steps] I've been averaging per day are improving my resting heart rate and I see my daily weight scale shows I've lost [5lbs]! :) Yeah, that's kinda-surface-level stuff, but these kind of health metrics are already extremely commonplace and helping open the door to further tech<->healthcare integration.

applecrazy 7y ago

> ML for healthcare. I'm kind of apprehensive about this advancement. There's many ways to mess this up, from intentional dataset poisoning (to encourage patients to get unnecessary procedure) to just terrible errors made with just a shift in data capture methodology. This paper has a lot of detail on how such systems can fail (https://arxiv.org/pdf/1804.05296.pdf https://arxiv.org/pdf/1804.05296.pdf).

JMTQp8lwXL 7y ago
People are already encouraged to get procedures or medications they don't need. A data set isn't going to change that.
- m0zg 7y ago
  Not to mention quarter of a million dead per year due to medical errors. If we could even just reduce the number of medical errors by 10 percent, that'd be HUGE. Source: https://www.cnbc.com/2018/02/22/medical-errors-third-leading-cause-of-death-in-america.html https://www.cnbc.com/2018/02/22/medical-errors-third-leading... Unfortunately, it's a multi-trillion dollar industry with insanely entrenched and deep pocketed players, and the old boy network the likes of which us software people can't even imagine. And that's before you consider the regulatory framework and the cost of compliance with it.
  - zelly 7y ago
    Solving medical errors would be huge. The problem is that it's a long-tailed problem with no clear way to solve with software. Every error is different, and there are already lots of bureaucratic mitigations in place every step of the way. There are no algorithms capable of the general reasoning required to broadly spot all types of errors including de novo ones. The only way to truly guarantee no medical errors is to replace every agent in a hospital with a machine.
    - m0zg 7y ago
      I bet just even solving the errors in diagnostics and/or catching stuff before it's too far gone would be huge in itself. I'm not talking AGI here, just bog standard perceptual stuff: looking at xrays, MRIs, mining medical records for patterns, handing unstructured medical records better, low hanging stuff like that.

carapace 7y ago

Ages ago Glyph (he of Twisted Python fame) said we would get to the point where we can e.g. use the Indiana Jones movies to build a computer model of Indiana Jones and make more movies...

dugditches 7y ago

What you're talking with 'render rights' is a major plot point in: https://en.wikipedia.org/wiki/The_Congress_(2013_film) https://en.wikipedia.org/wiki/The_Congress_(2013_film)

Mortos3 7y ago
Also, the Michael Crichton movie Looker: https://en.wikipedia.org/wiki/Looker https://en.wikipedia.org/wiki/Looker

arcadeparade 7y ago

RE: deep fakes, in a few years a movie director will be able to say to his editor: "increase Tom Cruise's rage and facial contortions in this scene by 50%"

al_chemist 7y ago

> give a company rights to use a 3d render of you in it There's reason why actors earn money and it's not due to perfect face or body.

Wh1skey 7y ago

DIY healthcare