this post was submitted on 29 Apr 2025
1 points (100.0% liked)
Technology
85282 readers
5385 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related news or articles.
- Be excellent to each other!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, this includes using AI responses and summaries. To ask if your bot can be added please contact a mod.
- Check for duplicates before posting, duplicates may be removed
- Accounts 7 days and younger will have their posts automatically removed.
Approved Bots
founded 3 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
The irony is Gemini is really good (like significantly better than ChatGPT), and cheap for them (no GPUs needed), yet somehow they made it utterly unbearable in search.
Gemini is really good at confidently talking nonsense but other than that I don't really see where you get the idea that it is good. Mind you, that isn't much better with the other LLMs.
So it's really good at the thing LLMs are good at. Don't judge a fish by it's ability to climb a tree etc...
No, it is mediocre at best compared to other models but LLMs in general have a very minimal usefulness.
I get the desire to say this, but I find them extremely helpful in my line of work. Literally everything they say needs to be validated, but so does Wikipedia and we all know that Wikipedia is extremely useful. It's just another tool. But its a very useful tool if you know how to apply it.
But Wikipedia is basically correct 99% of the time on basic facts if you look at non-controversial topics where nobody has an incentive to manipulate it. LLMs meanwhile are lucky if 20% of what they see even has any relationship to reality. Not just complex facts either, if an LLM got wrong how many hands a human being has I wouldn't be surprised.
LLMs with access to the internet are usually about as factually correct as their search results. If it searches someone's blog, you're right, the results will suck. But if you tell it to use higher quality resources, it returns better information. They're good if you know how to use them. And they aren't good enough to be replacing as many jobs as all these companies are hoping. LLMs are just going to speed up productivity. They need babysitting and validating. But they're still an extremely useful tool that's only going to get better and LLMs are here to stay.
That is the thing, they are not "only going to get better" because the training has hit a wall and the compute used will have to be reduced since they are losing money with every request currently.
Technology these days works in that they always lose money at the start. Its a really stupid feature of modern startups IMO. Get people dependent and they make money later. I don't agree with it. I don't really think oir entire economic system is viable though and that's another conversation.
But LLMs have been improving exponentially. I was on board with everything you're saying just a year ago about how they suck and they're going to hit a wall even. But the don't need more training data or the processing power. They have those and now they're refining the LLMs. I have a local LLM on my computer that performs better than chat GPT did a year ago and it's only a few GB. I run it on a shitty laptop.
I experimented with quite a few local LLMs too and granted, some perform a lot better than others, but they all have the same major issues. They don't get smarter, they just produce the same nonsense faster (or rather often it feels like they are just more verbose about the same nonsense).
I don't know what to tell you. I have them successfully compiling tables of search outputs to compare different things for method development and generating code, saving me hours of work each week. It all needs to be checked, but the comparison comes with links and the code is proofread and benchmarked. For most of what I do it's really just a jacked up search engine, but it's able to scan webpages faster than me and that saves a lot of time.
As a hobby, I also have it reading old documents that are almost illegible and transcribing them pretty well.
I really don't know what you're doing that you're just getting nonsense. I'm not.
One other comment pointed me at one issue that might be a major difference. Is the code you generate in one of those ultra-verbose languages like Java where we had basically IDEs generating code from much shorter descriptions already 20 years ago? I could see LLMs doing well with those.
I tend to try to generate code mostly in Rust or sometimes shell or config files or DSL for various programs and 99% of the time the code does not even come close to what I wanted it to do, mainly because it just hallucinates itself some library interfaces that do not exist.
It can be grounded in facts. It's great at RAG. But even alone, Gemini 2.5 is kinda shockingly smart.
...But the bigger point is how Google presents it. It shouldn't be the top result of every search just thrown into your face, it should be a opt-in, transparent, conditional feature with clear warnings, and only if it can source a set of whitelisted, reliable websites.
After just trying it again a few times today for a few practical problems that it not only misunderstood at first completely and then gave me a completely hallucinated answer to every single one I am sorry, but the only thing shocking about it is how stupid it is despite Google's vast resources. Not that stupid/smart really apply to statistical analysis of language.
Gemini 2.5? Low temperature, like 0.2?
The one they use in search is awful, and not the same thing. Also, it's not all knowing, you gotta treat it like it has no internet access (because generally it doesn't).
The one they use on gemini.google.com (which is 2.5 right now but was awful in earlier versions too).
Try it here instead, set the temperature to like 0.1 or 0.2, and be sure to set 2.5 Pro:
https://aistudio.google.com/
It is indeed still awful for many things. It's a text prediction tool, not a magic box, even though everyone advertises it kinda like the later.
"Significantly better than ChatGPT" and "Good" aren't the same. Like ipecac is significantly better to drink than sewage water.