981–990 of 1001 posts
The best solution to hallucination and inaccuracy is to give the LLM mechanisms for looking up the information it lacks. Tools, MCP, RAG, etc are crucial for use cases where you are looking for factual responses.
I get so confused on this. I play around, test, and mess with LLMs all the time and they are miraculous. Just amazing, doing things we dreamed about for decades. I mean, I can ask for obscure things with subtle nuance where I misspell words and mess up my question and it figures it out. It talks to me like a person. It generates really cool images. It helps me write code. And just tons of other stuff that astounds me…
Truly amazing technology which is very good at generating and correcting texts is marketed as senior developer, talented artist, and black box that has solution to all your problems. This impression shatters on the first blatant mistake, e.g. counting elephant legs: https://news.ycombinator.com/item?id=38766512
Earlier quoted context omitted.
A municipality in Norway used LLM to create a report about the school structure in the municipality (how many schools are there, how many should there be, where should they be, how big should they be, pros and cons of different size schools and classes etc etc). Turns out the LLM invented scientific papers to use as references and the whole report is complete and utter garbage based on hallucinations.
And that says… what? The entire LLM technology is worthless for all applications, from all implementations? A company I worked for spent millions on a customer service solution that never worked. I wouldn’t say that contracted software is useless.
You're the first in the thread to have brought that up; there are far more charitable ways to have interpreted the post you're replying to.
You're using them wrong. Everyone is though I can't fault you specifically. Chatbot is like the worst possible application of these technologies. Of late, deaf tech forums are taken over by language model debates over which works best for speech transcription. (Multimodal language models are the the state of the art in machine transcription. Everyone seems to forget that when complaining they can't cite sources for s…
> You're using them wrong. Everyone is though I can't fault you specifically. If everyone is using them wrong, I would argue that says something more about them than the users. Chat-based interfaces are the thing that kicked LLMs into the mainstream consciousness and started the cycle/trajectory we’re on now. If this is the wrong use case, everything the author said is still true. There are still applications made be…
I’ve been using Claude a lot lately, and I must say I very much disagree. For example, the other day I was chatting with it about the health risks associated with my high consumption of grown salmon. It then generated a small program to simulate the accumulation of PCB in my body. I could review the program, ask questions about the assumptions, etc. It all seemed very reasonable. A toxicokinetic analysis it called it…
"Trust but verify" With an LLM it should be "Don't trust, verify" - but it isn't that hard to verify LLM claims, just ask it for original sources. Compare to ye olde scientific calculators (90s), they were allowed in tests because even though they could solve equations, they couldn't show the work. And showing the work was 90% of the score. At best you could use one to verify your solution. But then tech progressed a…
Have you tried actually doing this? Most of the time it makes up urls that don't exist or contradict the answer it just gave.
Earlier quoted context omitted.
> They push the narrative that they’ve created something akin to human cognition This is your interpretation of what these companies are saying. I'd love to see if some company specifically anything like that? Out of the last 100 years how many inventions have been made that could make any human awe like llms do right now? How many things from today when brought back into 2010 would make the person using it make it f…
No major company directly states "We have created human-like intelligence," they intentionally use suggestive language that leads people to think AI is approaching human cognition. This helps with hype, investment, and PR. >I'd love to see if some company specifically anything like that? 1. DeepMind researchers: Sparks of Artificial General Intelligence: Early experiments with GPT-4 - https://arxiv.org/abs/2303.12712…
Earlier quoted context omitted.
I am so confused too. I hold these beliefs at the same time, and I don't feel they don't contradict each other, but apparently for many people some of these do: - LLMs are a miraculous technology that are capable of tasks far beyond what we believed would be achievable with AI/ML in the near future. Playing with them makes me constantly feel like "this is like sci-fi, this shouldn't be possible with 2025's technology…
> but even just the tought of becoming... irrelevant is depressing In my opinion, there can exist no AI, person, tool, ultra-sentient omniscient being, etc. that would ever render you irrelevant. Your existence, experiences, and perception of reality are all literally irreplaceable, and (again, just my opinion) inherently meaningful. I don't think anyone's value comes from their ability to perform any particular feat…
I get so confused on this. I play around, test, and mess with LLMs all the time and they are miraculous. Just amazing, doing things we dreamed about for decades. I mean, I can ask for obscure things with subtle nuance where I misspell words and mess up my question and it figures it out. It talks to me like a person. It generates really cool images. It helps me write code. And just tons of other stuff that astounds me…
> Wah, it can't write code like a Senior engineer with 20 years of experience! No, that's not my problem with it. My problem with it is that inbuilt into the models of all LLMs is that they'll fabricate a lot. What's worse, people are treating them as authoritative. Sure, sometimes it produces useful code. And often, it'll simply call the "doTheHardPart()" method. I've even caught it literally writing the wrong algor…