Writing is fundamentally the transfer of information from your brain to my brain. If you have 1000 bits of semantic information you want to transfer, you can't give 300 bits of semantic information to an LLM and have it fill in the remaining 700, because it doesn't know what those 700 bits are. If it's able to guess those 700 bits correctly, then they aren't true semantic information, and you really only have 300 bits you want to transfer. You might as well transfer those bits to me directly, rather than having the LLM add on an extra superfluous 700 bits that I then have to filter out.
For a long time after the internet arrived on the scene, a lot of online news stories would reference websites, papers, polls, etc. without linking to them. There are still news sources doing this today. Sometimes such articles interpret or place context around their hidden references, but a lot of the time they just summarize.
Giving someone the text output of a LLM is very similar to publishing a summary without links to the referenced material. When you were querying your LLM, you could have asked specific questions or asked for a custom focus or point of view. Your intended audience might have questions or different concerns, but they're unable to interact with your LLM. What you have delivered is static and unresponsive. It has all the disadvantages of being machine output without the advantage of being interactive, the way your LLM was for you.
It may have to wait until compute is cheap enough that tokens are essentially free, but we need a system to pass "hyperlinks" to LLM's primed with context, ready to be interactively queried on a chosen context. It's being overly generous to assume that people are putting even 300 bits into a LLM for every 1000 bits of regurgitated writing they try to pass off as their own. When people post LLM output as if it were their own, I have no choice but to assume they had zero knowledge of the subject, but this query taught them what they wanted to learn, and now they're sharing that. That's fine, but please pass an interactive LLM link rather than static text.
Once we have "hyperlinks" for LLM sessions, perhaps we can share LLM output a little more usefully and honestly.
I've seen professional journal pieces refer to science journal articles only to go and read the original article and find that it draws a different conclusion than what is implied by the journalist.
My father used to complain about that 40 years ago, though in his case he was reading newspapers rather than professional journals. But he'd point it out to me often enough that I started to see the pattern. Scientist publishes paper saying "We may have found evidence of X, which suggests the possibility that Y may also be occurring". Journalist: "Scientists find X which proves Y".
This has been happening for decades; I still see it happening today*. My cynical suspicion is that words like "maybe" and "suggests the possibility" don't sell enough papers.
* Worst offender I can remember was actually from the summary of a paper published on the research institution's own website, so I couldn't blame it on "Oh, the journalist misunderstood what the scientist wrote". Summary said "Exposure to X can, on average, cause a 40% higher chance of Y" (where Y was a negative health outcome). I clicked through to the study and read it. Turned out the confidence interval on that chance of Y was so wide, all you could say with 95% confidence was that exposure to X could do anything from reduce your chance of Y by 5 percent, or increase it by 85 percent, or somewhere in between. They had averaged -5 and +85 to get the scarier-sounding 40% number that they published in the summary, but the truth would have been far closer to "this confidence interval is so wide that we really can't conclude anything from this data". But that wouldn't be nearly as likely to get them grants, so they tortured the data in their summary so that it would look better.
> so I couldn't blame it on "Oh, the journalist misunderstood what the scientist wrote"
It's never the case that someone misunderstood what scientist wrote. Much like the scientific papers, news articles, including those reporting specifically on the discovery, have their own goals, and the paper being cited is used as evidence or argument for article's own "study". Except for press, the standard is rhetorical, not scientific, it's the conclusions and not the methods that are "pre-registered" at the start, and claims are defended by "hey it's just a point of view", not by statistical significance.
In your own example of worst offender: the scientific study was trying to establish and quantify the connection between X and Y. The summary article was trying to push the angle that "this institution is doing important work". It started with that conclusion, and the paper cited was just the first thing the author found that could be easily massaged into supporting that conclusions by rhetorical standards.
Same paper might get cited by journalist trying to push for "X is bad for you", and they'll do roughly the same as the summary article. And, same paper may be cited by someone claiming they have a miracle cure for Y, and they'll make a honest observation that "absence of X reducing Y is a common bullshit claim based on misunderstanding the paper [citation], that actually shows there's no correlation there, I mean look at the confidence intervals, even the author says that in text nobody bothers to read"... - citation may be honest, but the article itself is still using it to prop up a different flavor of bullshit.
TL;DR: don't believe news. It's bad for your mental and physical health (p<00.05).
hatthew · · focus · HN ↗
Writing is fundamentally the transfer of information from your brain to my brain. If you have 1000 bits of semantic information you want to transfer, you can't give 300 bits of semantic information to an LLM and have it fill in the remaining 700, because it doesn't know what those 700 bits are. If it's able to guess those 700 bits correctly, then they aren't true semantic information, and you really only have 300 bits you want to transfer. You might as well transfer those bits to me directly, rather than having the LLM add on an extra superfluous 700 bits that I then have to filter out.
beloch · · focus · HN ↗
Giving someone the text output of a LLM is very similar to publishing a summary without links to the referenced material. When you were querying your LLM, you could have asked specific questions or asked for a custom focus or point of view. Your intended audience might have questions or different concerns, but they're unable to interact with your LLM. What you have delivered is static and unresponsive. It has all the disadvantages of being machine output without the advantage of being interactive, the way your LLM was for you.
It may have to wait until compute is cheap enough that tokens are essentially free, but we need a system to pass "hyperlinks" to LLM's primed with context, ready to be interactively queried on a chosen context. It's being overly generous to assume that people are putting even 300 bits into a LLM for every 1000 bits of regurgitated writing they try to pass off as their own. When people post LLM output as if it were their own, I have no choice but to assume they had zero knowledge of the subject, but this query taught them what they wanted to learn, and now they're sharing that. That's fine, but please pass an interactive LLM link rather than static text.
Once we have "hyperlinks" for LLM sessions, perhaps we can share LLM output a little more usefully and honestly.
whycome · · focus · HN ↗
rmunn · · focus · HN ↗
This has been happening for decades; I still see it happening today*. My cynical suspicion is that words like "maybe" and "suggests the possibility" don't sell enough papers.
* Worst offender I can remember was actually from the summary of a paper published on the research institution's own website, so I couldn't blame it on "Oh, the journalist misunderstood what the scientist wrote". Summary said "Exposure to X can, on average, cause a 40% higher chance of Y" (where Y was a negative health outcome). I clicked through to the study and read it. Turned out the confidence interval on that chance of Y was so wide, all you could say with 95% confidence was that exposure to X could do anything from reduce your chance of Y by 5 percent, or increase it by 85 percent, or somewhere in between. They had averaged -5 and +85 to get the scarier-sounding 40% number that they published in the summary, but the truth would have been far closer to "this confidence interval is so wide that we really can't conclude anything from this data". But that wouldn't be nearly as likely to get them grants, so they tortured the data in their summary so that it would look better.
TeMPOraL · · focus · HN ↗
It's never the case that someone misunderstood what scientist wrote. Much like the scientific papers, news articles, including those reporting specifically on the discovery, have their own goals, and the paper being cited is used as evidence or argument for article's own "study". Except for press, the standard is rhetorical, not scientific, it's the conclusions and not the methods that are "pre-registered" at the start, and claims are defended by "hey it's just a point of view", not by statistical significance.
In your own example of worst offender: the scientific study was trying to establish and quantify the connection between X and Y. The summary article was trying to push the angle that "this institution is doing important work". It started with that conclusion, and the paper cited was just the first thing the author found that could be easily massaged into supporting that conclusions by rhetorical standards.
Same paper might get cited by journalist trying to push for "X is bad for you", and they'll do roughly the same as the summary article. And, same paper may be cited by someone claiming they have a miracle cure for Y, and they'll make a honest observation that "absence of X reducing Y is a common bullshit claim based on misunderstanding the paper [citation], that actually shows there's no correlation there, I mean look at the confidence intervals, even the author says that in text nobody bothers to read"... - citation may be honest, but the article itself is still using it to prop up a different flavor of bullshit.
TL;DR: don't believe news. It's bad for your mental and physical health (p<00.05).
etcetcetcetceta · · focus · HN ↗
TeMPOraL · · focus · HN ↗
elliotec · · focus · HN ↗