No, you're misunderstanding the findings. It does show that LLMs do not explain their reasoning when asked, which makes sense and is expected. They do not have access to their inner-workings and generate a response that "sounds" right, but tracing their internal logic shows they operate differently than what they claim, when asked. You can't ask an LLM to explain its own reasoning. But the article shows how they've made progress with tracing under-the-hood, and the surprising results they found about how it is able to do things like plan ahead, which defeats the misconception that it is just "autocomplete"
It's true that LLMs aren't "aware" of what internal steps they are taking, so asking an LLM how they reasoned out an answer will just output text that statistically sounds right based on its training set, but to say something like "they can never reason" is provably false.
Its obvious that you have a bias and desperately want reality to confirm it, but there's been significant research and progress in tracing internals of LLMs, that show logic, planning, and reasoning.
EDIT: lol you can downvote me but it doesn't change evidence based research
It’d be impressive if the environmental toll making the matrices and using them wasn’t critically bad.
Developing a AAA video game has a higher carbon footprint than training an LLM, and running inference uses significantly less power than playing that same video game.
only the non critical to function parts can be plastic
This just isn't reality. For .22lr, nearly 100% of the gun can be plastic, including the barrel. Here are two real examples where the only required metal component is the firing pin:
For more reliable and more powerful guns, some critical components must be metal, of course, like the barrel might at least require a liner, but the majority of the gun and internal mechanism can still be plastic, not only non-functional or cosmetic parts.
And you can use literally anything else other than 3D printed stuff. So why is 3D printed even an issue.
Idk, I didn't say it is. I'm just informing you that 3D printed guns are real, not a "fallacy", some function very well and reliably, and can actually be made with almost no or minimal metal parts
Uhhhh my friend you can print a 9mm CZ scorpion that shoots as well as the real thing. Same with an MP5. You can find videos of them being tested and shot on the internet. A lot has changed in the past few years with 3D printed guns.
Of course, they are not 100% plastic, but that's irrelevant. They can be made at home with little effort using a 3D printer and from simple materials anyone can buy at a hardware store, without any registration or serial numbers.
A lot of "real" guns are made from plastic, too, btw
Factually, they illegally searched his bag without a warrant at the mcdonald's, repacked the bag, put the bag in a police vehicle and drove to the police station without bodycam, and then turned bodycam back on to search the bag again and instantly "find" the ghost gun in his bag, which, without a serial number, is conveniently impossible to prove it was not planted.
The motion goes on the state that once that officer’s body cam footage resumes, it shows her immediately re-opening and closing the backpack compartments she already searched and then opening the front compartment of the backpack “as if she was specifically looking for something. Instantly, she ‘found’ a handgun in the front compartment.”
No, you're misunderstanding the findings. It does show that LLMs do not explain their reasoning when asked, which makes sense and is expected. They do not have access to their inner-workings and generate a response that "sounds" right, but tracing their internal logic shows they operate differently than what they claim, when asked. You can't ask an LLM to explain its own reasoning. But the article shows how they've made progress with tracing under-the-hood, and the surprising results they found about how it is able to do things like plan ahead, which defeats the misconception that it is just "autocomplete"