A laughable concept; absurd on its face. Just like the idea that uber is just suburban mums taking one or two trips on the way back from school.
It will absolutely be a full time, below minimum wage job that desperate people do. The same as uber, delivery drivers, and the entire rest of the gig economy
How does the entire textual corpus of say, new York times compare to all novels? Each article is a page of text, maybe two at most? There certainly are an awful lot of articles. But it's hard to imagine it is much more than a couple hundred novels. There must be thousands of novels released each year
LLMs are (apparently) massively used to get information about topics in the real world. Novels aren't going to be much help there. Journalism, particularly in written form, provides a fount of facts presented from different angles, as well as opinions, and it was all there free for the taking…
Wikipedia provides the scantest summary of that, fora and social media give you banter, fake news, summaries of news, and a whole lot of shaky opinions, at best. Novels give you the foundations of language, but in terms of knowledge nothing much beyond what the novel is about.
To begin with, your premise is that the only primary sources are press conferences and that press conferences only provide information in response to questions.
But even taking it literally, isn't that one of the things LLMs could actually do? You're essentially asking how a text generator could generate text. The real question is whether the questions would be any good, but the answer isn't necessarily no.
You used to need them, because journalists had the distribution and the sources didn't. In a word of printed newspapers, you couldn't get your story distributed nationally (much less worldwide) without the help of a journalist, doubly so if you wanted to stay anonymous.
Nowadays, you just make a Substack and there's that.
See that recent expose on the Delve fraud as just one example. No journalists were harmed in the making of that article.
Primary sources can and often are, very biased. Journalists are (supposed to be) doing fact checks and gathering multiple sources from all sides. Modern journalism is in a terrible state, but still important.
Imagine if all info about Facebook came from Facebook...
no. Scroll already does that. The header can stay where it is. The most intuitive thing you can do is have the content scroll in the direction the user asked for, when they asked for it.
If they want to go up to the top, they can already scroll. To the top.
The article's main idea is that for an AI, sycophancy or adversarial are the two available modes because they don't have enough context to make defensible decisions. You need to include a bunch of fuzzy stuff around the situation, far more than it strictly "needs" to help it stick to its guns and actually make decisions confidently
I think this is interesting as an idea. I do find that when I give really detailed context about my team, other teams, ours and their okrs, goals, things I know people like or are passionate about, it gives better answers and is more confident. but its also often wrong, or overindexes on these things I have written. In practise, its very difficult to get enough of this on paper without a: holding a frankly worrying level of sensitive information (is it a good idea to write down what I really think of various people's weaknesses and strengths?) and b: spending hours each day merely establishing ongoing context of what I heard at lunch or who's off sick today or whatever, plus I know that research shows longer context can degrade performance, so in theory you want to somehow cut it down to only that which truly matters for the task at hand and and and... goodness gracious its all very time consuming and im not sure its worth the squeeze
> I could give you a list of chess moves and force you to recover the complete board state from it, and you wouldn't fare that much better than an off the shelf LLM would
idk, I would expect anyone with an understanding of the rules of chess, and an understanding of whatever notation the moves are in, would be able to do it reasonably well? does that really sound so hard to you? people used to play correspondance chess. Heck I remember people doing it over email.
In comparison, current ai models start to completely lose the plot after 15 or so moves, pulling out third, fourth and fifth bishops, rooks etc from thin air, claiming checkmate erroneously etc, to the point its not possible to play a game with them in a coherent manner.
I would expect that off the shelf GPT-5.4 would be able to do it when prompted carefully, yes. Through reasoning - by playing every move step by step and updating the board one move at a time to arrive at a final board state.
On the other hand, recovering the full board state in a single forward pass? That takes some special training.
Same goes for meatbag chess. A correspondence chess aficionado might be able to take a glance at a list of moves and see the entire game unfold in his mind's eye. A casual player who only knows how to play chess at 600 ELO on a board that's in front of him would have to retrace every move carefully, and might make errors while at it.
Try to play a simple over the board style game with 5.4 with whatever notation you chose to use (or just descriptions, literally anything). Prediction: it will start out fine, but the mid game will be very hard to keep it on track, and the endgame will make you give up.
Generally, the more you write (and especially, the more you write long form content), the better your writing becomes. This also goes in reverse. Those who have great trouble writing, are unlikely to do much of it.
This alone can account for the seeming disparity. Though many people write poorly, they do not write much text for public consumption at all.
I imagine the person is talking about optimising along the lines you personally value for. Some people may value being with other interesting people (and being interesting themselves), wheras others prefer to be around smart people, or beautiful people, or rich people. Many see some of these desires as misguided or foolish or vain or whatever (often when it conflicts with their own values), but it is true that at least some people seem to want these.
I think this misses something, which is that there is absolutely the option to progress towards a region that is more "tool-like". See the difference between kimi k2 and many of the leading LLM providers. Its a lot better at avoiding sycophancy, avoiding emotive reasoning, etc. It's not as capable as others, and it is of course possible that thats why, but I find use for it regardless because of its personality
This matches what I observe in Claude and GPT Codex models when it comes to coding tasks. The personality differences go deeper; they relate to how each model approaches its work. Claude tends to communicate a lot by default, while GPT Codex simply executes tasks. System prompts and context files hardly change this behavior. Your point about Kimi K2 is intriguing because it suggests there's a real range here. Being "more tool-like" is a valid area to consider, not just a sign of a failed personality. The question is whether this area can still handle more complex tasks or if the article's argument holds true, meaning some capability is lost.
ChatGPT can do a terrific job of this, too— if you select the “Efficient” base style and tone, plus turn off the Warmth, Enthusiasm, and Emoji sliders.
So many people would benefit from this, I wish they advertised the config settings more
It will absolutely be a full time, below minimum wage job that desperate people do. The same as uber, delivery drivers, and the entire rest of the gig economy
reply