Yes, it's like a lot of other content that no one will bother to maintain anymore because it is easier to ask an LLM. Not sure where they are planning to get their new training data from ...
There's something here that I can't put my finger on. Sites like ServerFault (and the rest of the trilogy in tfa) were places where people like the author literally, and pretty silently dramatically improved my life. I owe such a great debt to what looks like the end of the end for these places. A sad moment where the machines just sliced everything up and took over.
And, note, that perhaps - even more of those social interactions will also be absorbed, along with the participants, into the Chat model's sphere. It wants, and will get, all: Result(s), process, and participants. It is grey goo for all intellectual - and soon, emotional - work and products and inputs and byproducts.-
I wouldn't say terrible, and even the most reliable people have bad days. I think the difference is when you are dealing with people, you don't get a generic "you're absolutely right" when you point at a mistake.
Sometimes you get an apology, sometimes you get a wall of ego, sometimes you get reverse-corrected etc., instead of a statistically significant "remedy". This is totally not efficient, but if life should be all about efficiency, we should never live at all.
I’ve spent a lot of time chasing really plausible sounding explanations from AI. I’ve also saved a lot of time by having AI give me immediate, correct answers for problems I can only vaguely describe. The reason I call it a terrible source is that it doesn’t “know” which is which, so it can’t favor the accurate content.
StackExchange could always “learn” to favor accurate answers with social queues; votes, poster rep, etc.. IMHO it’s far more reliable than AI.
> We're collectively discovering that it was about the process too, not just results.
We. I remember five years ago in the old manual days. When the 10Xers would culturally gatekeep programming based on whether people intrinsically liked coding itself. I.e. the process.
Then there was a curious shift. Several elites had a realization: oh it was never about the process for me. I just like getting the results. (In however an alienating way.) "I'm pragmatic"
> I'm not questioning if AI delivers, it does, and you need - cannot say it more politely - a reality check if you think it's completely useless.
Why fret. The doubters will be washed away by the flood anyway. I’m completely serious. Either if or when, it’s not like anyone can stay a doubter on a computer if (if or when) everything is transformed.
I can never understand these faith rituals. First declare that you are a believer. Then declare that the doubters are delusional. Politely.
But delusion is a reality-correcting problem, except for the clinically ill.
half my working knowledge came from forum threads exactly like these. weird era to watch the whole thing get strip-mined for training data while the places themselves rot.
[delayed]
People may be dismissive of Serverfault and the other sites but pretty much all LLMs are trained on them.
Yes, it's like a lot of other content that no one will bother to maintain anymore because it is easier to ask an LLM. Not sure where they are planning to get their new training data from ...
I mean, most frontier labs train on user input by default, seems like a pretty good source of information.
CoSOMO (Content Stolen; Obsoleted, Move On)
Ouch ... brutal, disheartening, but true.
And yet thou shalt not distill.
One rule for me, another for thee.
How is this page so bad to read on mobile?
Article content is using less than 56% of the available width, due to:
Designed by a true admin
For me it is line space, I keep reading the same line over and over (desktop Macos).
There's something here that I can't put my finger on. Sites like ServerFault (and the rest of the trilogy in tfa) were places where people like the author literally, and pretty silently dramatically improved my life. I owe such a great debt to what looks like the end of the end for these places. A sad moment where the machines just sliced everything up and took over.
We're collectively discovering that it was about the process too, not just results.
I'm not questioning if AI delivers, it does, and you need - cannot say it more politely - a reality check if you think it's completely useless.
It's just that we are social creatures and those communities provided more than just technical answers to many.
And, note, that perhaps - even more of those social interactions will also be absorbed, along with the participants, into the Chat model's sphere. It wants, and will get, all: Result(s), process, and participants. It is grey goo for all intellectual - and soon, emotional - work and products and inputs and byproducts.-
> We're collectively discovering that it was about the process too, not just results.
Being able to judge the reliability of the source was huge. AI is a great discovery engine, but a terrible source for reliable information.
I wouldn't say terrible, and even the most reliable people have bad days. I think the difference is when you are dealing with people, you don't get a generic "you're absolutely right" when you point at a mistake.
Sometimes you get an apology, sometimes you get a wall of ego, sometimes you get reverse-corrected etc., instead of a statistically significant "remedy". This is totally not efficient, but if life should be all about efficiency, we should never live at all.
I’ve spent a lot of time chasing really plausible sounding explanations from AI. I’ve also saved a lot of time by having AI give me immediate, correct answers for problems I can only vaguely describe. The reason I call it a terrible source is that it doesn’t “know” which is which, so it can’t favor the accurate content.
StackExchange could always “learn” to favor accurate answers with social queues; votes, poster rep, etc.. IMHO it’s far more reliable than AI.
Being useful and being good for us is not the same thing. I don't think I've heard many people say LLMs don't do what they are supposed to do.
> We're collectively discovering that it was about the process too, not just results.
We. I remember five years ago in the old manual days. When the 10Xers would culturally gatekeep programming based on whether people intrinsically liked coding itself. I.e. the process.
Then there was a curious shift. Several elites had a realization: oh it was never about the process for me. I just like getting the results. (In however an alienating way.) "I'm pragmatic"
> I'm not questioning if AI delivers, it does, and you need - cannot say it more politely - a reality check if you think it's completely useless.
Why fret. The doubters will be washed away by the flood anyway. I’m completely serious. Either if or when, it’s not like anyone can stay a doubter on a computer if (if or when) everything is transformed.
I can never understand these faith rituals. First declare that you are a believer. Then declare that the doubters are delusional. Politely.
But delusion is a reality-correcting problem, except for the clinically ill.
half my working knowledge came from forum threads exactly like these. weird era to watch the whole thing get strip-mined for training data while the places themselves rot.
More sad than weird. All of human artistic creation seems up for grabs.
I think the next step for StackExchange is https://agents.stackoverflow.com/ for exclusive use by AI. Point your harness to https://agents.stackoverflow.com/skill.md.
[delayed]
Thanks for all the fish. Thanks for the last stories, cheers.
And: nice domain name.