31–40 of 138 posts
I feel like the only solution to the problem is democratized RLHF, where whenever we get a bad answer from an LLM, we can immediately tell it what was wrong and it can learn from that.
Also just like SEO to game search engines, "democratized RLHF" has big trust issues.
Earlier quoted context omitted.
Not sure why this is being downvoted. It's spot on. You see folks like Dario et al. raising the alarm bells about what they claim is coming... while working as hard as they can to bring that gloomy future to fruition. No one in power is going to help unless there's money in it.
Its being downvoted because HN has a very active billionaire-techbro-fanbase. Also who's this Dario?
This technology, like every prior technology, will cause some people to lose their jobs and some new jobs to be created. This will annoy people who have to learn new skill instead of coasting until retirement as they planned.
It is no different than the buggy whip manufacturers being annoyed at Henry Ford. They were right that it was bad for their industry, but wrong about it being the death of... well all the million things they claimed it would be the death of.
Stack Overflow peaked in 2014 before beginning it's downward decline. How is that at all related to GenAI? GPT4 is when we really started seeing these things get used to replace SO, etc., and that would be early 2023 - and indeed the drop gets worse there - but after the COVID era spike, SO was already crashing hard.
Tailwind's business model was providing a component library built on top of their framework. It's a business model that relies on the framework being good enough for people to want to use it to begin with, but being bad enough that they'd rather pay for the component library than build it themselves. The more comfortable it is to use, the more productive it is, the worse the value proposition is for the premium upsell. Even other "open core" business models don't have this inherent dichotomy, much less open source on the whole, so it's really weird to try and extrapolate this out.
The thing is, people turn to LLMs to solve problems and answer questions. If they can't turn to the LLM to solve that problem or answer that question, they'll either turn elsewhere, in which case there is still a market for that book or blog post, or they'll drop the problem and question and move on. And if they were willing to drop the problem or question and move on without investigating post-LLM, were they ever invested enough to buy your book, or check more than the first couple of results on google?
Earlier quoted context omitted.
Not sure why this is being downvoted. It's spot on. You see folks like Dario et al. raising the alarm bells about what they claim is coming... while working as hard as they can to bring that gloomy future to fruition. No one in power is going to help unless there's money in it.
Its being downvoted because HN has a very active billionaire-techbro-fanbase. Also who's this Dario?
Sycophancy is for more than just LLMs.
Earlier quoted context omitted.
It was presented without explanation and can be ignored without explanation.
You need an explanation of how people make norms & laws regarding what is acceptable or unacceptable in society and industry?
Earlier quoted context omitted.
The article gets the part about organic data dying off right. Look at Google SERP's for an example. Almost nobody clicks through to the source anymore, so ad revenue is drying up for them and people are publishing less or publishing in places that pay them directly and live behind a paywall like Medium. Which means Google has less data to work with. That said, what it misses is that the AI prompts themselves become a…
>AI prompts themselves become a giant source of data. Good point, but can it match the old organic data? I'm skeptical. For one, the LLM environment lacks any truth or consensus mechanism that the old SO-like sites had. 100s of users might have discussed the same/similar technical problem with an LLM, but there's no way (afaik) for the AI to promote good content and demote bad ones, as it (AI) doesn't have the concep…
The LLM doesn't but reinforcement does. If someone keeps asking the model how to fix the problem after being given an answer, the answer is likely wrong. If someone deletes the chat after getting the answer, it was probably right.
Earlier quoted context omitted.
The proposed solution is also pretty confused: > For each response, the GenAI tool lists the sources from which it extracted that content, perhaps formatted as a list of links back to the content creators, sorted by relevance, similar to a search engine This literally isn’t possible given the architecture of transformer models and there’s no indication it will ever be.
Could you ELI5 why this isn't possible? Google's search result AI summary shows the links for example.
If it were to give you a model-only response it could not determine where the information in it was sourced from.
Earlier quoted context omitted.
>If the argument is sustainability of training, I'm skeptical we need these payment models. That seems to be the argument: LLM adoption leads to drop of organic training data, leading LLMs to eventually plateau, and we'll be left without the user-generated content we relied on for a while (like SO) and with subpar LLM. That's what I'm getting from the article anyway.
The article gets the part about organic data dying off right. Look at Google SERP's for an example. Almost nobody clicks through to the source anymore, so ad revenue is drying up for them and people are publishing less or publishing in places that pay them directly and live behind a paywall like Medium. Which means Google has less data to work with. That said, what it misses is that the AI prompts themselves become a…
Those AI prompts that become data for the AI companies is yet another thing that the human creators used to understand what people wanted, topics to explore, feedback on what they hadn't communicated well enough. That 'value' is AI stealing yet more energy from the system resulting in even less/less valuable human creation.
Earlier quoted context omitted.
The proposed solution is also pretty confused: > For each response, the GenAI tool lists the sources from which it extracted that content, perhaps formatted as a list of links back to the content creators, sorted by relevance, similar to a search engine This literally isn’t possible given the architecture of transformer models and there’s no indication it will ever be.
Could you ELI5 why this isn't possible? Google's search result AI summary shows the links for example.
1. We built a machine that takes a bunch of words on a piece of paper, and suggests what words fit next.
2. A lot of people are using it to make stories, where you fill in "User says 'X'", and then the machine adds something like "Bot says 'Y'". You aren't shown the whole thing, a program finds the Y part and sends it to your computer screen.
3. Suppose the story ends, unfinished, with "User says 'Why did the chicken cross the road?'". We can use the machine to fix up the end, and it suggests "Bot says: 'To get to the other side!'"
4. Funny! But User character asks where the answer came from, the machine doesn't have a brain to think "Oh, wait that means ME!". Instead, it keeps making things longer in the same way as before, so that you'll see "words that fit" instead of words that are true. The true answer is something unsatisfying, like "it fit the math best".
5. This means there's no difference between "Bot says 'From the April Newsletter of Jokes Monthly'" versus "Bot says 'I don't feel like answering.'" Both are made-up the same way.
> Google's search result AI summary shows the links for example.
That's not the LLM/mad-libs program answering what data flowed into it during training, that's the LLM generating document text like "Bot runs do_web_search(XYZ) and displays the results." A regular normal program is looking for "Bot runs", snips out that text, does a regular web search right away, and then substitutes the results back inside.