Orchard
Lars Marowsky-Brée 😷Tell me again how #GenAI will extract meaningful trends from and answer queries about your data set.
I can also see this going great for coding, programming languages and computers are known to be very forgiving and tolerant
Theuni@larsmb It's fantastic. Here's the FastGPT version of it. It greatly demonstrates that it doesn't have the ability to count / compute but only combines textual artefacts.
@larsmb And obviously it immediately fucks up any arbitrary counting. And then it gives a reference to some site that obviously doesn't have that "word" in it ...
Vlad 🇺🇦@vladimir_lu @larsmb I'm assuming you're ironic, but there is a point worth discussion. Mathematical models are intended to cover all general cases. Specifically in computer science we've been bitten by exactly the idea of "nobody will ever put this in" (see y2k bug). We've been through this. And on a more anthropological level good math (and physics) gets extended and remains valuable even in shifting contexts. LLMs are garbage in that way.
Barry Rowlingson@theuni @vladimir_lu @larsmb It might get asked "How many C base pairs are in the DNA sequence CCTGAGATCTAGGAGGGCATCCGC?"
@geospacedman @vladimir_lu @larsmb There are no C in DNA sequences. This is an often made confusion as there are only G, T and A in DNA as the C looks much alike the G.
Tom Ritchford@theuni @geospacedman @vladimir_lu @larsmb
Assuming you aren't joking:
https://en.wikipedia.org/wiki/Nucleic_acid_sequence
"The possible letters are A, C, G, and T, representing the four nucleotide bases of a DNA strand – adenine, cytosine, guanine, thymine"
@theuni @geospacedman @vladimir_lu @larsmb haha!
People don't recognize satire on the internet, endless studies have shown...
Peter Bindels@geospacedman @theuni @vladimir_lu @larsmb statistically you are over 99% likely to have GATTACA in your DNA.
Shark Attak@geospacedman @theuni @vladimir_lu @larsmb
This sounds like the premise of a bad sci-fi horror movie..
This sounds like the premise of a bad sci-fi horror movie..
Magnus Ahltorp@theuni @vladimir_lu @larsmb The whole point of LLMs is to *not* do things “correctly”. The LLM is doing it’s job perfectly here, it’s just that the expectation is wrong.
@theuni Oh, I wasn't actually aware that someone else had found this before! Nice. I had actually stumbled across it independently.
But it is very funny to me that this is the level of technology that certain people want to ram into everything and are burning 100+ Terawatt hours per year by now.
Sure, GPTs, LLMs, ML in general, very cool stuff. But also, as ready as a wet noodle.
Florian Idelberger
Georg Arne Spenden
IBBoard
echopapa
Subjekt_91
Tartas1995
Father EnochI'm sorry for my ignorance but I thought THE WHOLE POINT of #computers was that they could count.
For three to four decades I was told computers were a one trick pony, with the trick being able to count faster than humans ever could.
WHY are we making computers that can't count?
BenjiButo 
@larsmb
Yeah and of course the "awesomeLibraryThatWouldJustFitYourNeed.dll" is available. I made it up myself 😁
Yeah and of course the "awesomeLibraryThatWouldJustFitYourNeed.dll" is available. I made it up myself 😁
Murenius@larsmb Hey, I'm very happy that I don't have to do my off by one errors on my own anymore.
Григорий КлюшниковLars Marowsky-Brée 😷, and for this reason, I won't understand for the life of me how could someone seriously use an LLM as a tool. Or instead of a proper search engine.
@grishka They're tools for when you need an answer that might not be fully correct, eg brain storming, rubber ducking, or even quite a few translations.
But they're nowhere near as useful as advertised.
But they're nowhere near as useful as advertised.
CubeThoughtsThis also makes perfect sense, because context matters - and once it generated a wrong answer, it is human enough to double down on it! The singularity is near!
You've got to ask it "nicely" right from the start. Don't embarrass it!
I AM A PROMPT ENGINEER
njsg@larsmb Of *all* the things it could learn from human content it learned "doubling down"?
Sally Strange@SallyStrange @larsmb Now that you've put it that way, I can't unsee this thought: So companies are generating excessive warming and water consumption just to build a massive smartphone autocomplete-autocorrect?
Maybe I'll call it "Generative Autocomplete".
Polly Kraisus@larsmb don't worry, these issues just keep getting fixed quickly after being reported and the product keeps improving... or does it? https://community.openai.com/t/incorrect-count-of-r-characters-in-the-word-strawberry/829618
Incorrect count of 'r' characters in the word "strawberry“
Bug Report " Description: The AI incorrectly states that the word “strawberry” contains only two ‘r’ characters, despite the user querying multiple times for confirmation. Steps to Reproduce: Ask the AI how many ‘r’ characters are in the word “strawberry.” Observe the AI’s response stating there are two ‘r’ characters. Reconfirm by asking the AI again. Notice that the AI consistently states there are two ‘r’ characters. Expected Result: The AI should correctly count and state that th...
ein kleines z
Tom Hannen
dogzilla
Mark Down@fartnuggets @larsmb @msbw It’s all yours.
I still have hope that open-sourced AI agents will be useful, but I’m personally done with trying to wrangle the big commercial LLMs into anything useful. I’ve yet to come across a real-world problem I can’t solve quicker with a Jupyter notebook and a couple Python libraries
@fartnuggets @larsmb @msbw I’m hoping for a future where a trusted on-device agent can basically act as a personal assistant. I think it needs some ability to learn and make decisions, but not this weird “boil the ocean” strategy behind LLMs.
Kinda reminds me of robotics in the early 90s - after decades of failed top-down approaches, we finally found huge success with drastically simpler ensemble bottom-up approaches exemplified by the Genghis family
Mossy Modem@larsmb It feels like there were definitely some Monty Python skits in the training data.
wirepair
SomeGadgetGuy@larsmb
That chat bot needs to go to the LIBARY!
[Typo on purpose]
@SomeGadgetGuy It pilfered all libraries and this is the best we got from it.
@larsmb
Truly the future is NOW. Amazing...
🙄
Nini@larsmb I must admit I'm impressed at how steadfast it is in never admitting to being wrong. The unearned confidence of a mediocre white man who's never been told "no".
Sky Leite
Aatheus@larsmb Oo, Gaslighting as a Service!
Crystal_Fish_Caves@larsmb yeah this tech DEFINITELY is worth all the resources it gobbles up. We have PLENTY of spare water and power.
@Crystal_Fish_Caves Exactly! It clearly should be the top priority for all businesses and politicians, it is *the best*.
Tricot Feelya@larsmb ArtificialInanity
sortius@larsmb the sad thing is, if they designed the LLM to purely count Rs, it might actually work... but that sounds too much like an algorithm, and that's got no techbro magic sauce in it.
It reminds me of when the world was captured by radioactive materials, and they stuck it in everything
Elio Campitelli@larsmb Looks like they patched it, but only for a very specific subset of berries.
