This is just a draft, best refrain from linking. (I hope we'll get this up tomorrow or Monday. edit: probably this week? edit 2: it's up!!) The [bracketed] stuff is links to cites.
Please critique!
A vision came to us in a dream — and certainly not from any nameable person — on the current state of the venture capital fueled AI and machine learning industry. We asked around and several in the field concurred.
AIs are famous for “hallucinating” made-up answers with wrong facts. The hallucinations are not decreasing. In fact, the hallucinations are getting worse.
If you know how large language models work, you will understand that all output from a LLM is a “hallucination” — it’s generated from the latent space and the training data. But if your input contains mostly facts, then the output has a better chance of not being nonsense.
Unfortunately, the VC-funded AI industry runs on the promise of replacing humans with a very large shell script. If the output is just generated nonsense, that’s a problem. There is a slight panic among AI company leadership about this.
Even more unfortunately, the AI industry has run out of untainted training data. So they’re seriously considering doing the stupidest thing possible: training AIs on the output of other AIs. This is already known to make the models collapse into gibberish. [WSJ, archive]
There is enough money floating around in tech VC to fuel this nonsense for another couple of years — there are hundreds of billions of dollars (family offices, sovereign wealth funds) desperate to find an investment. If ever there was an argument for swingeing taxation followed by massive government spending programs, this would be it.
Ed Zitron gives it three more quarters (nine months). The gossip concurs with Ed on this being likely to last for another three quarters. There should be at least one more wave of massive overhiring. [Ed Zitron]
The current workaround is to hire fresh Ph.Ds to fix the hallucinations and try to underpay them on the promise of future wealth. If you have a degree with machine learning in it, gouge them for every penny you can while the gouging is good.
AI is holding up the S&P 500. This means that when the AI VC bubble pops, tech will drop. Whenever the NASDAQ catches a cold, bitcoin catches COVID — so expect crypto to go through the floor in turn.
It is a more nuanced issue than this and a bit naive as well, if I can be blunt and still say; friend, I cared enough to read and comment.
The next most probable token is not related to hallucinations as described. This is like saying statistics is worthless because it doesn't give absolute answers.
AI can mean a great many different architectures and all have different strengths, weaknesses, and issues. The best way I can relate current LLM's is the early days of the microprocessor. There were a lot of issues and limitations to work through. Eventually companies like Sun made some really capable and powerful machines after the first few generations of microprocessors. These devices became much more complex with time; integrating many peripheral devices. Many systems used several microprocessors in a single machine when it wasn't cost prohibitive. If you have a computer in the last few generations, you have around twenty very similar microprocessors all working together on the same chip.
AI is presently at that early stage. It is a useful fundamental tool, but by itself it is not very remarkable. The innovations in the peripheral space are where things get interesting. The way these innovations get integrated and the way the complexity multiplies over time will follow a similar curve as the microprocessor.
There is and will be lots of misuse and failed companies over time, but the technology will continue. This is an inevitable future and it will never go away.
When someone talks about hallucinations, it means the model is outside of alignment. The complexity of the model and its bias is the largest factor. In many ways this is why the the proprietary AI models will fail eventually. Open Source, offline models are the future of AI for text. An 8×7B unfiltered research model hallucinates far less than others and does not involve the massive amount of data that can be collected an inferred by proprietary AI. The majority of hallucinations are due to user input errors that are not accounted for in the model tokenizer and loader code. This is just standard code errors. Processing every possible spelling, punctuation, and grammar error is a difficult task. The next probable token is not simply a matter of the probable vector in the dataset. If the input contains a rare error the output will be in the style of a foolish error. This is not a hallucination. It is responding in the style it was addressed. If the user does not have control of the entire text inside the present context, aka previous questions and answers, the style of "stupid error" will likely remain persistent. It is still not an error, it is just a mirror. Indeed this is the greatest analogy. AI is like a mirror of yourself upon the dataset. It can only reflect what is present in the dataset and only in a simulacrum of yourself through the prompts you generate. It will show you what you want to see. It is unrivaled access to information if you have the character to find yourself and what you are looking for in that reflection.
I have no idea what the future holds but I hope one day to obtain the title of "confusion monk".
spoiler
idk I've been playing Larn a lot and this is where my mind went.Suddenly everything in my life looks less pathetic by comparison.
i promise we did it! we made iphone 2! this is just like iphone 2! of course it doesn't work yet but it will work eventually! we made iphone 2 please believe us!!
he's already banned but i love how every time this argument comes up there's absolutely no substance to the metaphor. "ai is like the internet/microprocessors/the industrial revolution/the Renaissance", but there's no connective tissue or actual relation between the things being compared, just some hand-waving around the general idea of progress and pointing to other popular/revolutionary things and going "see! it's just like that!"
"i'm sorry, but you used the wrong form of 'their' in your prompt, that's why it inexplicably included half a review of Click in your meeting summary."
s-tier. no notes. does lemmy have user flairs? because if so i'm calling dibs
It doesn’t have flairs yet (I also thought of that a while back), but I guess now we can do it in philthy!
user flairs, even in a non-federated form, are a feature I’ve wanted since the beginning. they’d be a great early feature for philthy!
Early microprocessors could do arithmetic correctly.
if you think about it hard enough, it's very like a blockchain
No they could not. They could not do floating point at all
Sooo... it turns out Bing Chat and Gemini can't do floating point math right now :D
And don't tell me the question is vague or misleading until the chatbots can actually recognize that or return an error code, this isn't exactly the sort of thing people are feeding into integration tests since they're indeterministic as heck and generally computed by some company losing tons of money somewhere off-site.
(Not sharing the result, because I still stubbornly refuse to spread AI output whatsoever)
The really fantastic part about this is that it’s long been possible to get reliable performance out of irc chatbots for this sort of thing, with people pulling all kinds of nasty extractor shit
“It’s just people badly providing input” was the most big-brain take I’ve seen in a while. Honestly reads like someone in the industry and hates their users, of the “my code is great, it’s these damn idiots that don’t know how to use the system!” variety
"you're holding it wrong" worked for iphones, so maybe it'll work for llms too…
the iphone did already have other utility alongside it though, so people were making use of it regardless. not excusing how apple handled that, mind, that was bullshit, but I meant that people were still motivated users
openai's particular flavours of this shit are still failing to find viable footholds and there's nothing that is a so-called "killer app", which is the other thing that really weakens its case
but that dynamic, of programmer disregard for how people use their products.... oof. less pls. return to sender. unsubscribe.
holy fuck. computer science in shambles
I’ve never said this before, but please tell me you used an LLM to generate this horseshit. no part of what you said is correct and it doesn’t take much knowledge of the tech to realize you’re either bullshitting or regurgitating marketing materials
I use the tech every day. Good luck with your echo chamber. You are a statistical inevitability. Time will teach you far more than I care to.
Buddy, I am a statistical inevitability.
wait... i'm sure it sounded cool and all but what does it mean for a person existing here and now to be inevitable in a statistical sense.....
that’s why I wish they’d given us more before they went I said good day sir and fucked off. I wanted more fractally wrong shit from the mind that gave us “the only issue with LLMs is user input, you poor naive soul” and “early computers couldn’t do arithmetic, ever heard of floating point? you fools” and that last one keeps being wrong in exciting new ways every time I think about it
not least in that calling floating point "arithmetic" is being far too generous to floating point...
oh i assure you the hits just keep on coming
a beautiful mind, with just a soupcon of LLM assistance
Aw shucks, I bet you say that to all the girls
ahahaha ok bye
I know your idiot ass has already been banned and I’ll likely never get an answer, so I ask this in this in full recognition of saying hi to the void, but
You were the kind of person to excitedly evangelise about bitcoin (and other shitcoins) with people at bars in the 2016~2020 years, weren’t you?