5 ms·

What makes you think data like this is readily available and of sufficient quality to provide nontrivial insights ? Just going off of failure of ML to produce m

What makes you think data like this is readily available and of sufficient quality to provide nontrivial insights ? Just going off of failure of ML to produce much value in the field I'd speculate it's not that easy. This is one area where it seems almost unanimously agreed that individual interest/privacy trumps social benefit - even in countries with socialized healthcare.

manderley 4y ago
Seems privacy is less of a concern in countries with privatized healthcare, I don't understand that last sentence. Is it some kind of political shibboleth?
- moonchrome 4y ago
  I'm just saying that considering society is providing for your healthcare, collecting socially valuable medical data doesn't seem like an unreasonable exchange to me - but I'm in the minority on that one I think.
- bertil 4y ago
  The original author answered, but I don’t think that answer is appropriate. I initially read it as a comment on how most large studies are based on national health care data in socialist countries (Scandinavia mostly, occasionally the UK’s NHS) because they are clean, consistent, and population-wide without imbalanced social class sampling. They also cover decades, making all sorts of key studies possible. Those are countries (the UK is the notable exception) where privacy is key and embedded in how those databases are structured and handled. In practice, those studies are possible because the health services only give information anonymized and relevant given an expressly spelled out research plan from a recognized team. That allows research without the semi-permanent health data leaks that health insurance companies have grown accustomed to. This is not a counter-balance for a free service: people in countries with national healthcare consider that they pay for top-of-the-line care through taxes. Just like people who pay for it indirectly through private insurance, they don’t believe that they deserve any less privacy for being prudent. The reason the data is shared is because this is seen as a valuable opportunity to support research. Any privacy breach would be far less acceptable, as such abuses don’t really happen in other aspects of their lives.
- ericd 4y ago
  I'm half convinced that a lot of the concern about privacy in the US is driven by orgs for whom the lack of transparency that permeates healthcare is a very helpful economic moat. IME, trying to get healthcare data is nightmarish, and companies try to extract incredible rates for sharing any of it.
capableweb 4y ago
Does it have to be "readily available" in order to train on it? For all we know, OpenAI could have pulled down the sci-hub/library genesis libraries and included everything from there in the training set. If they didn't, I hope they do it for GPT5.
- jacquesm 4y ago
  That's about 150T, so not beyond the capability of the companies we're talking about. The bigger question is how much real information is encoded in there and how many of the 'leaf' papers (the ones that build on other work but that so far are without further confirmation) are wrong. ChatGPT might give you good answers 9 out of 10 times and kill you on the 10th.
throwaway426079 4y ago
The UK has things like opensafely.org and other anonymous/pseudonymous analysis platforms with restricted access. Only as good as the structured, clinically coded data that gets entered. (I heard of a couple of projects around NLP applied to free text notes but that's not the usual approach).
bertil 4y ago
I expect that a lot of details are in doctor’s notes but not properly encoded, and most recent efforts have been through analysis of structured (categorical even) data. Having a natural-language tool parse could help notice more patterns. There have been many discoveries done by companies that started using NLP in the last 10 years. However, all the anecdotes that I‘ve heard about were not published but instead used by insurers to adjust their premiums.

manderley 4y ago

Seems privacy is less of a concern in countries with privatized healthcare, I don't understand that last sentence. Is it some kind of political shibboleth?

moonchrome 4y ago
I'm just saying that considering society is providing for your healthcare, collecting socially valuable medical data doesn't seem like an unreasonable exchange to me - but I'm in the minority on that one I think.
bertil 4y ago
The original author answered, but I don’t think that answer is appropriate. I initially read it as a comment on how most large studies are based on national health care data in socialist countries (Scandinavia mostly, occasionally the UK’s NHS) because they are clean, consistent, and population-wide without imbalanced social class sampling. They also cover decades, making all sorts of key studies possible. Those are countries (the UK is the notable exception) where privacy is key and embedded in how those databases are structured and handled. In practice, those studies are possible because the health services only give information anonymized and relevant given an expressly spelled out research plan from a recognized team. That allows research without the semi-permanent health data leaks that health insurance companies have grown accustomed to. This is not a counter-balance for a free service: people in countries with national healthcare consider that they pay for top-of-the-line care through taxes. Just like people who pay for it indirectly through private insurance, they don’t believe that they deserve any less privacy for being prudent. The reason the data is shared is because this is seen as a valuable opportunity to support research. Any privacy breach would be far less acceptable, as such abuses don’t really happen in other aspects of their lives.
ericd 4y ago
I'm half convinced that a lot of the concern about privacy in the US is driven by orgs for whom the lack of transparency that permeates healthcare is a very helpful economic moat. IME, trying to get healthcare data is nightmarish, and companies try to extract incredible rates for sharing any of it.

capableweb 4y ago

Does it have to be "readily available" in order to train on it? For all we know, OpenAI could have pulled down the sci-hub/library genesis libraries and included everything from there in the training set. If they didn't, I hope they do it for GPT5.

jacquesm 4y ago
That's about 150T, so not beyond the capability of the companies we're talking about. The bigger question is how much real information is encoded in there and how many of the 'leaf' papers (the ones that build on other work but that so far are without further confirmation) are wrong. ChatGPT might give you good answers 9 out of 10 times and kill you on the 10th.

throwaway426079 4y ago

The UK has things like opensafely.org and other anonymous/pseudonymous analysis platforms with restricted access. Only as good as the structured, clinically coded data that gets entered. (I heard of a couple of projects around NLP applied to free text notes but that's not the usual approach).

bertil 4y ago

I expect that a lot of details are in doctor’s notes but not properly encoded, and most recent efforts have been through analysis of structured (categorical even) data. Having a natural-language tool parse could help notice more patterns. There have been many discoveries done by companies that started using NLP in the last 10 years. However, all the anecdotes that I‘ve heard about were not published but instead used by insurers to adjust their premiums.