Reminder that EVERYTHING Anthropic or OpenAI announce in public is propaganda designed to boost their market cap when they hit the IPO they're aiming for.
It's marketing, folks. There is no intelligence behind the artifice, it's just spicy autocomplete and hucksters in $5000 suits trying to pick your boss's pocket.
https://mastodon.social/@arstechnica/115548040527129225
@menelion @cstross "An autocomplete would never suggest new names, new implementation and architectural ideas, and so on, and so forth."
Why? Autocompletes often make changes I do not want and that are out of context with bad parameters.
LLMs are basically large scale autocomplete, running on GPU, with a grain of stochastic variablity put on top.
@menelion @cstross The fact that a word is made up and that it sounds fitted for a given context are two different things, none of which validate your claim.
One the hand you say that your word is made up so it’s unlikely that ChatGPT gives you its context. Yet it is fairly easy to guess the context of some made up words such as FartLand or Poppilympulla. The first one can be a parody country or planet and the second a fake medica or botanical term. Yet, they’re made up. Because made up words are also often chosen so that we make sense of their use.
In addition to that, you say that you've chosen your word to a given use case that ChatGPT guessed. By the mere fact that you have chosen this word for a fantasy land context is an indication that you found it fit for the purpose, based on your own knowledge of fantasy literature or art. This is exactly what ChatGPT is also trained with. It is fooling you.
On the other hand, for the use of such tools, my brain does a better job most of the time withour relying on a venture industry that is likely to dramatically raise its costs at some point and that is polluting the planet, essentially for mediocre outputs that nobody requires.
@menelion @Enthalpiste @cstross
In the past few months AI has wasted far more of my time than it could ever possibly save.
One of the things I do in my day job is proof-read, re-word and fact-check matters of local history. I spent days trying to unpick a batch of information that simply didn't add up and couldn't be referenced. In the end it became apparent that someone had searched the internet for some background and believed the first things they read on those wretched AI overview search engine results.
@Enthalpiste @menelion @cstross "...a bit of stochasticity on top" this is so far the best description of the "intelligence" in LLMs. Randomness. It does not "invent" -- it picks at random. Similarily it does not "guess" and certainly does not "know".
I find it disturbing when people rely on LLMs to act as teachers for them or ask for advice in legal or medical aspects. Worst of all is that children learn to rely on these services.
A trillion dollar stochastic autocomplete.
@jluukka @menelion @cstross I really had a full "WTF?" moment when I discovered that what some models call "temperature" really is a Boltzmann like "exp(-1/kbT)" sampling over the output probability distribution.
Like... why? This makes sense in statistical physics because of canonical distribution. But I'm not sure there is a solid theoretical ground for this in ML.
@cstross
Don't look behind the curtain.
Ignore that guy.
The Wonderful Wizard of Oz, 1900
https://en.wikipedia.org/wiki/The_Diamond_Age
A Young Lady's Illustrated Primer, 1995.
The AI was fake.
It is pretty amazing to think they would go down this track but I guess they figure Congress is so dysfunctional that they can totally dismiss any regulatory risk that might carry otherwise.
While it absolutely is marketing - in some situations it also is an effective tool. When used by a skilled operator, a *good* AI (something very large and recent) can increase coding speed significantly - I can do in a day what would have taken me a week. The hackers are using it too, and it's effective, and that's a real problem in computer security.
But for human/creative stuff, especially with an unskilled user? Yeah it's a joke. And for unskilled technical folks it can actually be a negative, leading you down unproductive paths. And yes they're built on theft. A bazillion problems.
But: AI is not entirely useless. Dismiss it at your peril.
@tbortels @cstross When I was young, AI was also big (but not tech bubble big). At the time, it was based on search, and some useful search strategies came out of it. But no one calls those AI any more.
Similarly, I expect some useful (but expensive in power/cooling) machine learning will still be used when the dust settles, but no one will call it AI either.
@FeralRobots @cstross They frame it like this because it makes LLM seem capable of more than just statistical reproduction of training material.
Heck, even calling LLMs AI trying to do the same thing.
Since the 1990s every tech wave has been a pump and dump bubble laying the groundwork for the next pump and dump bubble. The staggering inefficiency of such a Ponzi economy, channelling capital into mega-billionaires pockets instead of productive growth is why China is rolling out 400 kph mag lev trains while we are shoveling brain dead software into toasters.
@michaelgemar @cstross @nicopap
Brilliant 👏
you don't need to do P&L statements when you're never going to turn a P
@mastodonmigration @nicopap @cstross
Could be a covid metric. .
Inter pupil airy distance. > 6’
@mastodonmigration @cstross I've been a fan of the Austrian Business Cycle model. I believe real finance people have debunked it, but we sure seem to see the cycles anyway. Easy money is "invested" into worse and worse businesses, which -- surprisingly -- go bust and take the economy with it.
This AI bust coupled with current USA policies is going to bring (IMO) a world-wide Depression.
Not said better anywhere. Its not just theft and unfairness, its massively destructive.
Like supply side economics slows the economy to a standstill, while also starving people with no healthcare.
Its not just not worth protecting supply side accumulationist billionairism, its VITAL to get rid of it with fair wages and fair taxation.
Grand Theft Autocorrect would have been a better term that LLM or GenAI 😉
@simonzerafa @cstross I've been calling it "grand theft autocomplete" for a while now (fairly much ever since I read someone else calling it that...)
And I'm still waiting for an anti-malware program to set its sights on de-AI-fying computers...
@cstross Plus most people in the world are literally evaluating this thing based on vibes and what the rich & powerful are saying. They don’t know enough and can’t learn fast enough.
Then there’s the issue of job insecurity as a threat that’s skewing their response towards anxiety about their jobs. Silicon Valley exploits this anxiety 250%
@cstross I've been part of an IPO. It's legalized theft. When my company went IPO, the price was set *far* too high, and immediately started tanking. Employees were prohibited from trading their options, I suppose to make sure none of *us* could take down the kind of money the bankers running the IPO did.
We opened at $20, and I noticed our CEO owned...zero shares. Once her trading program kicked in, I could see that she *knew* it was an $8 stock. Should be criminal fraud, but %100 legal.
@cstross Can you short sell an IPO?
If I had a couple of spare millions lying around, I'd be very, very tempted.
Kimbo
Charlie Stross
synlogic4242
André Polykanine
Pier Hegeman
Drew
G. Clavier
Hurricane Daveosaurus
Joel Luukka
Gary McGraw
Ray McCarthy
Lightfighter
Shafik Yaghmour
Tom Bortels
Analog AI
unattributed 𓂃✍︎
Tim Hergert
FeralRobots
Miron 🇪🇺
Mastodon Migration
norapap 🏳️⚧️
Michael Gemar
paranoia Agent Rosenflower
ideaPDish
Jon
Yora
AI is Woke ™
BibbleCo
Kevin Russell
robloblaw
Grumpy Grimnir 
Simon Zerafa
Toni Aittoniemi
McNeely
Third spruce tree on the left
Maryan Pelland Pen2Profit
papis802
Jürgen Hubert