251–260 of 387 posts
Since then the only difference between an Onion fiction and things actually sucking that much is a decade or less in almost all cases.
If we blacklisted content seen in the Onion, we'd automatically wipe out most news.
Earlier quoted context omitted.
Not only is the current internet 80% spam, it's rapidly approaching 99% thanks in large part to LLMs. At this point I would be shocked if Google had a solid plan for how to handle this going forward as the problem space gets more difficult.
that's the part that scares me. I railed on someone's comment the other day about "indexes will come back into fashion" but the more I think about how much garbage has increased in just the past 2 to 3 years, I think I was wrong. Indexes and forums may be the only way to have a sane net where you can find things. Perhaps communities linking together in a ring like format, a "web ring" of sorts.
https://groupsy.applicationfitness.com/post/healthymeals/664...
Earlier quoted context omitted.
LLMs are completely deterministic even if that's kind of weird to state because they output things in terms of probabilities. But if you simply took the highest probability next word, you'd always yield the exact same output given the exact same input. Randomness is intentionally injected to make them seem less robotic through the 'temperature' parameter. Why it's not just called the rng factor is beyond me.
Maybe some models can be deterministic at a point in time, but train it for another epoch with slight parameter changes and a revised corpus and determinism goes out the proverbial (sliding) window real quick. This is not unwanted per se, and the exact feedback loop that needs improving to better integrate new knowledge or revise knowledge artefacts incrementally/post-hoc.
It is still "deterministic" in that training on exactly the same data and asking exactly the same questions should (unless someone manually adds randomness) lead to the same results.
Another example of the distinction might be a pseudo-random number generator: For any given seed, it is entirely deterministic, while at the same time being very deliberately hard to predict without actually running it to see what happens.
Earlier quoted context omitted.
> And i still think it is a spurious complaint, given that we have exactly the same limitations when dealing with any human agent We're not talking about an LLM that is trying to do the job of a human, here We're talking about an LLM that is trying to give authoritative answers to any question typed into the Google search bar It's already well past the scale that humans could handle Talking about human shortcomings w…
Nothing of the sort. I'm trying to understand why anyone cares about formal verifiability in this context, since it's not something we rely on when asking humans to answer questions for us. We evaluate any answer we get without such mathematical proofs, and instead simply judge the answer we're given on its fit and usefulness. Anyone who doubts the usefulness of even these nascent LLMs is fooling themselves. The proo…
An LLM does not have such real world limitations. It will hallucinate nonstop and then create layers of gaslighting explanations to its hallucinations. The problem is that you absolutely must be a domain expert at the LLM’s topic or always go find the facts elsewhere to verify (then why use an LLM?).
So a company like Google using an LLM, is not providing information, it’s doing the opposite. It is making it more difficult and time consuming to find information. But it is then hiding their responsibility behind the model. “We didn’t present bad info, our model did, we’re sorry it told you to turn your recipe into poison…models amirite?”
A human doing that could likely face some consequences.
Earlier quoted context omitted.
It could be because they actually released something. If you look back, the Google Research blog posts always have grandiose claims, but you can often never use them.
AlphaGo, AlphaFold, and Waymo FSD are all released in the sense that you can see them actually working in the real world. Those all took much longer to put together than whatever rushed features were released to catch up with OpenAI, however.
This approach to remove bad search suggestions manually reminded of a different approach Google once took, where they weren’t satisfied with manually tweaking search results but rather wanted to tweak the algorithm that produces these results when there were bad results. 'Around 2002, a team was testing a subset of search limited to products, called Froogle. But one problem was so glaring that the team wasn't comfort…
I got a couple of answers that are based on SEO spam produced by an ecommerce with a lot of reputation and of course the answers don't make any sense
This approach to remove bad search suggestions manually reminded of a different approach Google once took, where they weren’t satisfied with manually tweaking search results but rather wanted to tweak the algorithm that produces these results when there were bad results. 'Around 2002, a team was testing a subset of search limited to products, called Froogle. But one problem was so glaring that the team wasn't comfort…
Sounds rather like how Google photos does not identify anything as a Gorilla.
Earlier quoted context omitted.
You choose such an edge case question - how about this sort of thing: Which is the best political party? Are the side effects to X medical treatment? I bet there are even cases when eating rocks is ok! PS It has been written about: https://www.atharjaber.com/works/writings/the-art-of-eating-... > Lithophagia is a subset of geophagia and is a habit of eating pebbles or rocks. In the setting of famine and poverty, cons…
Are you really going to start eating rocks just to convince yourself that Google's AI isn't shit and objective truth is not real?
My point is that I object to the idea that a result can be 100% right! Even in the case of eating rocks, it seems there are times that it can be beneficial.
To think '100% correct' is achievable is to misunderstand the nature of reality.
Earlier quoted context omitted.
> LLMs so far seem to be entirely unverifiable. I don't understand this complaint. Are they any less verifiable than a human?
I can ask a human to explain the steps they took to answer a question. I can ask a human a question 100 times and I don't get back 100 different answers. None of those applies to an LLM.
If the answer has any length at all, I imagine the answer can vary every single time the person answers, unless they prepared for it, memorized it word by word.