432
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
this post was submitted on 17 Nov 2026
432 points (97.8% liked)
Technology
89881 readers
3801 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
The quote is from 2018. Back then most neural networks were still mostly single purpose - far from today’s all-in-one package. Text generation was barely working at all.
The underlying problems are still the same. Transformer models do not have an inkling of intelligence, and they still don't have an internal model of the world around them. Creativity is also non-existent, which is a hard requirement if you want these things to actually create something new instead of rehashing what already exists.
There was no rehashing in 2018. What you did was you trained your own (local) model on some data to help you work. For example, a grad student would painstakingly mark cells in em-scans for weeks and even months, now they could do it for a day, use that result to train a neural network and only supervise the job done. And the job wiuld take minutes instaid of weeks or months.
There were no llms in 2018, IIRC the google/deepmind paper was from 2019 (can be wrong here).
Source: built AI training app in 2018 for science company.
This has been shown repeatedly to not be correct since 2023.
That's literally how transformers turned out to have been working so well. They build world models and bags of heuristics fitting the training data to extrapolate from.
It's almost surprising you managed to use the correct term for the tech but then managed to misrepresent it so egregiously.
You really think that an Othello board could represent the world? Talk about being delusional. This is pathetic. The subheadline: "OTHELLO-GPT USUALLY PREDICTS LEGAL MOVE" says it all - it cannot have a representative internal world model, because in that case the "USUALLY" would go away, as anyone who has played othello for longer than 10 minutes will tell you - and we are talking about a game with very simple rules here, the complexity does not even reach chess - and people expect these things to spit out coherent and valid solutions from target selection in wars to the color their bedroom walls should be painted in.
Edit: I really have to add here the following excerpt (percentages are the rate of invalid moves as the top choice):
So even if you just feed them real world data without completely synthesizing it, 1 in 20 moves are invalid in a very simple (rule-side) game. If you do not give them a whole lot of played games as a basis, it fails in over 93% of cases (not at making a GOOD move, just at making any valid move!), which is probably very close to a random guess for a move. Something that has even an inkling of an internal world view would be able to adapt after being told that their move is invalid. Also, an internal world view would allow for simply dropping the rulebook onto ChatGPT as a basis, as it works with humans that can imagine.
Spoken with the confidence of a non user.
I use Claude every day for actual production software used in a real product and I can say with confidence that even the most advanced publicly available models totally lack the higher order thinking required to build good software. They're pretty good at writing code, when directed well. They are not good at architectural or design decisions and they're even worse at actually following decisions made outside of their context. Those are the actually challenging parts of software engineering and you have to beat them into the model every step of the way or your software quickly becomes an unmaintainable mess.
Use Spec driven development and a good verification loop to overcome the "beat it into the model at every step".
Most human programmers are bad at architecture. They’re making lots of mistakes too. Perfection is not a criterion for intelligence. In that case no human is intelligent.
I am not a non-user. I run local models, because i am a tech enthusiast, and am watching the space closely. I will never spend a cent or use the offerings of OpenAI and Anthropic tho - i will not finance the people who are providing the tools to bind us all.
Open weight models are not far behind and good enough to get software developers jobless, but there is a huge difference in what you can do with 1M context and 500token/sec or 60k context and 40token/sec.
Sure! End the AI Bubble and i'll go ahead and buy a bunch of H200's.
But for real, if you structure your tasks well, you can get by for most things with local models, and for a small business in IT, investing in dedicated hardware you own is a smart move. The employer of an acquaintance of mine bought some inference hardware for his 5 engineer business, and they rely mostly on that. Around 80% of all tasks are done by a GLM5.3 model, and while they currently still use Claude for some things, the percentage of external inference has dropped dramatically in the last months, because people have adapted on how to structure tasks for the local model.
Noone can explain to me how Claude, OpenAI and the NeoClouds plan to become profitable businesses when you can reduce inference costs by 90% by getting your own inference - and it's not even that hard to set up. if you already have IT staff for your inhouse stuff, they will probably manage.
Someone will come up with a decent accelerator eventually. Right now we're in the pre-soundblaster days where audio sythesis is being done by the CPU but for neural networks.
We're repurposing graphics hardware, these are primitive times.
Also, there, I explained it for you.
I have to disagree here. We are on the absolute edge of current physics regarding miniaturization - without a massive breakthrough in quantum physics on how we can use subatomic particles for transistors we are nearing the maximum we can shrink these circuits. These are already optimized for exactly this kind of math - a Vera Rubin will never be able to render a game. Also, the mathematical basis (a ton of matrix multiplication) does not give much hope that we can replace all this math with something simpler. Maybe a radical redesign taking a completely different approach can help, but i wouldn't count on that - it would mean reinventing the whole damn thing on a completely new base.
I can't really see any explanation when i take the above into account.
In any case, confidence without knowledge about the matter, that is something we laughed at LLMs just 2 years ago. Those became much much better at it.
Now suddenly it's en vogue with humans.
Hey that's a good zinger, good work. I found an article to help you take down that AI Hater. If you want me to jazz it up, or reword future zingers for better impact, just let me know. I'm down for anything.
There's been no qualitative change ever. Computers are computing today just like in 2018 and like 1960.
Just like a slug neuron and a human brain neuron? No difference at all?