All I ask is in what way are LLMs progress. Ability to generate a lot of slop is pretty much only thing LLMs are good for. Even that is not really cheap, especially factoring the environmental costs.
How much do you know about transformers?
Have you ever programmed an interpreter for interactive fiction / MUDs? It’s a great example of the power that even super tiny models can accomplish.
Also consider that Firefox or Electron apps require more RAM and CPU and waste more energy than small language models. A Gemma slm can translate things into English using less energy than it requires to open a modern browser.
I am not implying that transformers-based models have to be huge to be useful. I am only talking about LLMs. I am questioning the purported goal of LLMs, i.e., to replace all humans in as many creative fields as possible, in the context of it’s cost, both environmental and social.
Sure everything starts with meager beginnings. The AI you’re upset about existing may find the cure to many diseases. It may save the planet one day.
The type of AI that researchers are building to try cure diseases are not LLMs. So not the stuff that is running behind these kind of tech for games.
or a silly, halfwit race to build out the infrastructure because they’re smoking their own product could crash the economy.
You’re only seeing the upsides - make nifty pictures, ai music, whatever - because the entire shitshow is a free or exceptionally underpriced preview of what’s to come. while everyone from google to grok to your mom fails to find a way to actually profit off of it all when they have to figure the costs of the water, power, training data, lawsuits and other shit into the actual equation it blows up.
These aren’t my ideas - please, take a break from your preconceptions and read:
futurism.com/data-centers-financial-bubble
zdnet.com/…/todays-ai-ecosystem-is-unsustainable-…
Where is the idea that LLMs will ever to curing diseases coming from? What is the possible mechanism? LLMs generate text from probability distributions. There is no reason to trust their output because they don’t have built-in concept of true or false. When one cannot judge the quality of the output, how can one reliably use it as a tool for any purpose, let alone scientific research?
There are other general AIs that can look over imaging like cat scans and in some situations catch things a doctor can’t.
There are also ones that can simulate drug interactions with the body and can be used to model creating novel drugs for treatments.
These are not LLMs though.
NuXCOM_90Percent@lemmy.zip 1 day ago
LLMs are actually spectacular for indexing large amounts of text data and pulling out the answer to a query. Combine that with natural language processing and it is literally what we all thought Ask Jeeves was back in the day. If you ever spent time sifting through stack overflow pages or parsing discussion threads, that is what it is good at. And many models actually provide ways to get a readout of the “thought process” and links to pages that support the answer which drastically reduces the impact of hallucinations.
And many of those don’t necessarily require significant power usage… relative to what is already running in data centers.
The problem is that people use it and decide it is “like magic” and then insist on using it for EVERYTHING. And you go from “Write me a simple function to interface with this specific API” to “Write me an application to do my taxes”
Of course, there is also the issue of where training data comes from. Which is why so much of the “generative AI” stuff is so disgusting because it is just stealing copyrighted data left and right. Rather than the search engine style LLMs that mostly just ignore the proverbial
README_FBI.txtfile.And the “this is magic” is on both sides. The evangelists are demonstrably morons. But the rabid anti-AI/“AI” crowd are just as bad with “it gave you a wrong answer, it is worthless”. Think of it less like a magic box and more like asking a question on a message board. You are gonna get a LOT of FUD and it is on you to do additional searches to corroborate when it actually matters.