this post was submitted on 01 Jun 2024
398 points (98.3% liked)
Technology
59589 readers
2910 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related content.
- Be excellent to each another!
- 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, to ask if your bot can be added please contact us.
- Check for duplicates before posting, duplicates may be removed
Approved Bots
founded 1 year ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
In what ways are you benefiting from a bevy of factually dubious query responses?
Can absolutely never blindly trust the hallucinating plagiarism machine.
It's great where either facts don't matter or you're personally in a position to vet all of its “factual” output 100%. Text revision, prompting for additional perspectives, prompting to challenge beliefs and identify gaps. Reformatting, quick and easy data extraction, outlining, brainstorming.
Reformatting and outlining as long as you go over and revise it again anyway, seemingly making that moot.
Data extraction as long as you don't care if the data is mangled.
Brainstorming is a good one, since off-the-wall ideas can be useful in that context.
In most cases I've seen AI used, the person spends as much time correcting it than they would if they just did the work without AI. So maybe it makes you feel more productive because a bunch of stuff happens all at once, but at least for text generation, I think it's more of a placebo.
It can at least get one unstuck, past an indecision paralysis, or give an outline of an idea. It can also be useful for searching though data.
If all I want is something blatantly false or legible yet nonsensical, like a modern lorem ipsum, it's a real time-saver.
Why not just use lorem ipsum? It's just a copy/paste, and without the liability of having false information if you forget to proofread it.
I guess ChatGPT is just completely useless, then.
You cannot in all seriousness use a LLM as a research tool. That is explicitly not what it is useful for. A LLM's latent space is like a person's memory : sure there is some accurate data in there, but also a lot of "misremembered" or "misinterpreted" facts, and some bullshit.
Think of it like a reasoning engine. Provide it some data which you have researched yourself, and ask it to aggregate it, or summarize it, you'll get some great results. But asking it to "do the research for you" is plain stupid. If you're going to query a probabilistic machine for accurate information, you'd be better off rolling dice.
Exactly my point - except that the word "reasoning" is far too generous, as it implies that there would be some way for it to guarantee that its logic is sound, not just highly resembling legible text.
I don't understand. Have you ever worked an office job? Most humans have no way to guarantee their logic is sound yet they are the ones who do all of the reasoning on earth. Why would you have higher standards for a machine?
I have higher expectations for machines than humans, yes.
Sounds like a recipe for disappointment tbh. But on the other hand, sounds like you trust techno marketing a bit too much.
No, I just know how to spot the lies in a datasheet.
I"m not sure what lie and what datasheet you're referring to ?
Just in general.
I don't really query, but it's good enough at code generation to be occasionally useful. If it can spit out 100 lines of code that is generally reasonable, it's faster to adjust the generated code than to write it all from scratch. More generally, it's good for generating responses whose content and structure are easy to verify (like a question you already know the answer to), with the value being in the time saved rather than the content itself.
It's good at regurgitating boilerplate, from what I've gathered.
This question betrays either your non-use or misuse of the products available. You're either just reading the headlines of the screw-ups or you're just bad at using the tool.
To directly answer your question:
But were you actually looking for a real response to your question?
It's worse at all programming tasks except boilerplate, especially with its tendency to inject booby traps. Not knowing how to use the programming language it emits becomes a significant problem.
Comparing a language model to an idiot is unfair to the idiot.
A normal search engine works for everything else.
Any well-defined query I've ever made of an LLM has resulted in hilariously bad results, but I suppose I was expecting it to do something that I couldn't already do better myself.
I'm a systems administrator, not a programmer. Like I said, quick scripts. An LLM could probably parse my comment better than you, evidently.
Oof.. Was this in reply to my bit about better grammar and ESL individuals?
Fuck no. Especially the python visualization point.
I suppose you're just a god among men then. For the rest of us, it's useful and you've been given plenty of good answers to your disingenuous question.
Someone doesn’t know how to use ChatGPT
Oh, is there an arcane invocation that magically imbues it with reason?
Nope, just gotta know what it IS, what it ISN’T, and how to correctly write prompts for it to return data that you can use to formulate your own conclusion.
When using AI, it’s only as smart as the operator.
Well, it's not AI, for starters.
As much as I hate to do this, it is AI, as ML is a part of Artificial Intelligence.
It isn't AGI, some might say it may be, but they are wrong. But the model is learning.
An LLM is not capable of learning. It won't hallucinate less with additional training input.
Just the notion of a computer having hallucinations should suggest that it's doing more than just basic code.
It's not 'intelligent', but it has 'learned' enough beyond standard CPU instructions.
That's why it's not a General AI, but it's still an AI.
I also talk about gremlins inside CPUs, but that doesn't mean I think there are magical critters turning a crank inside them.
It's called a metaphor, brother.
Regardless, it's all code that's eventually run on a CPU, so there isn't any step where magic is injected.
Sigh.
There is no code for language processing, it's just math approximating results from weights. The whole weight set-up is what's called 'artificial intelligence', because nobody wrote
if prompt like 'python' return ['large snake', 'programming language', 'australian car company']the model 'learned' how to mimic human speech using training, not by 1000s of software engineers adding more branches to the code.
That technique is part of 'artificial intelligence', when computers solve problems they were not programmed to do. The neural network learns its knowledge by the code, but the code has no idea what is going on.
How do you think math is implemented on a computer?
I am now properly confused as to what are you arguing for.
So let me go to the basics.
Computers follow instructions to the letter. Take input, process it, produce output.
There are specific instructions that computer can carry out, we can build on top of them to make them more complex. We write code to do that.
True/false gates can become numbers, which can become text, audio, video.
But everything 'programmed' or 'digitally created' is using the same instructions and only ever does what we tell the computer to do.
Cutting video will require video input, and then user has to do specific actions to produce a specific result.
Almost everything in existence is built like that - someone wrote specific code for technology to behave.
Now, this is very primitive way of solving tasks, specifically for real-world parameters. Computers have gigabytes (10^9) of memory, but just the earth has 10^50 atoms, so we can't put eveything into a computers (which is why we can't 100% predict the weather), and checking for every input parameter is not only futile, but also meaningless.
Enter 'artificial intelligence', approximated way of solving problems. Suddenly we don't code the tasks themselves, we only specify the neural network - weights and connections between them, and code the 'learning' algorhitm that adjusts the weights based on inputs during 'training'. Training is the expensive part, where we put huge amounts of input into the network, and if the answer we get is incorrect, we adjust the weights and try again with another sample.
It's very expensive in every way, but the code involved doesn't care about anything other than adjusting those weights. The network can be fed images and determining whether it's a dog or a cat. It can be fed audio samples and expect to write down the lyrics. The code doesn't know or care, apart from distinguishing between correct and not correct answers and adjusting those weights.
After those weights are set to our satisfaction, we can release them for others to use. We expect the network to have 'reliable' outputs for our inputs, so we just calculate the neuron activations based on those weights for every input, nothing else is necessary.
Therefore you do have code in the machine that learns, but only during training, and you have code that actually 'runs' the algorhitm for calculating output. But the actual solution to the problem is not inside the code, it can't even be coded by humans in any way. The neural network is a statistical model generated by the training set and according to our learning algo. The bigger the network, the bigger the training set, the better should those outputs be (in theory).
To take the cutting video example further, you can train network to cut trailers from movies.
Or you can let editors do that.
They both will use computers, but one is using deteministically coded software that just follows specific orders one by one, and the other just computes the neuron activations based on the inputs and produces an output based on what it had available in the training data with some probability.
So yes, machines can learn, and it's a subset of the 'Artificial Intelligence' field.
It won't hallucinate less with additional training input.
An LLM is good at making sentences that seem convincing, but has no ability to reason.
Thanks for ignoring the same argument over and over again, it makes you look very stuck-up.
Intelligence does not require perfection (you are an example). You also hallucinate random output, but you can learn to stop specific hallucinations - like reading a Wiki page.
LLM aren't different in that regard - they were trained on inputs, and if you extend their training sets, they will be more exact in those areas.
Ability to reason is a very hard concept to specify, and we don't have any foolproof test (that I know of) that would definitely say if LLMs can reach that stage.
I will fight you if you try to tell me that humans are smarter than any current AI - because there are some real dumb people walking this earth and mindlessly reproducing, unable to process basic concepts that they depend their lives on.
Nothing of this changes the fact that there is an intelligence - natural language is an incredibly hard thing to code deterministically - and as such deserves the 'AI' label without a doubt.
There is a complete lack of intelligence, just a passable facade that crumbles under scrutiny.
Keep going…
No you don't understand. The word AI, which was invented to describe this kind of technology, should not be used to describe this technology. It should instead be reserved for some imaginary magical technology that may exist in the future.
From what I've seen online, most people differentiate between AI and AGI, which is cool.
So then don’t call it AI.
I thought the sarcasm in my comment was self evident 🤔
Ahh.
Can't blame you when some people non-ironically use that argument all the time
New version of people who know how to search the web vs those who don't. Currently shit search results broken by search companies notwithstanding.