Adso @scasey.bsky.social · 10hThere is no evidence for AI being conscience, if there were, Anthropic would be demonstrating it, they do research on their models constantly. This also is a weird way to put the burden of proof on skeptics. 2280
Adso @scasey.bsky.social · 15hThere was pushback on national guard deployment though which is why the Supreme Court blocked it outside of DC, unless requested by the governor. Which is why Trump keeps asking Chicago if he can send them there. 110
Adso @scasey.bsky.social · 26/09/2026Lol, you also are forced to have 5 clocks on your screen at all times. 020
Adso @scasey.bsky.social · 24/09/2026The AP isn't some left leaning anti-AI monolith, that's just silly and wishful thinking. If you are going to make this claim, you need to show some evidence, not just throw out some dumb conspiracy. 001
Adso @scasey.bsky.social · 24/09/2026I was reading their research paper on sniffing out how it is making decisions, and even they are putting things like thinking in quotes to convey it's a computer model, not a human. I have no issue with using the word thinking for computers as long as we know the context. 100
Adso @scasey.bsky.social · 24/09/2026Most journalists are on X and Threads lmao. Anthropic doesn't even use your " correct" language. Hilarious. 100
Adso @scasey.bsky.social · 24/09/2026That isn't thinking though, even Anthropic's paper on how LLMs make decisions isn't using these terms. 100
Adso @scasey.bsky.social · 24/09/2026It doesn't have to be. If you want to know the extremely complex nature of an LLM, you aren't going to find it in an AP story, and using terms like "thinking" isn't going to help. I don't even think Anthropic uses these terms and when they do, they uses quotes. 000
Adso @scasey.bsky.social · 24/09/2026That isn't what they are doing, no one is obfuscating anything by not using your preferred language in a story. 100
Adso @scasey.bsky.social · 24/09/2026Or they just think writing about machines should use clear language to describe what it does and not confuse the person in to thinking the machine is making its own decisions. I just find the sudden leap to a conspiracy hilarious. 200
Adso @scasey.bsky.social · 24/09/2026First of all, reporting on tech has always been pretty poor by news outlets that don't specialize in tech reporting. Second, I just don't get the connection to the words they choose to use and the vast conspiracy by the AP to cover up LLMs. 100
Adso @scasey.bsky.social · 24/09/2026They aren't using human words to describe machines so they are covering up what LLMs can do it a ridiculous statement. 200
Adso @scasey.bsky.social · 24/09/2026I don't think the AP is involved in some conspiracy to cover up the capabilities of LLMs. 370
Adso @scasey.bsky.social · 24/09/2026It's a paper by Anthropic on Claude using circuit tracing. www.anthropic.com/research/tra...anthropic.comTracing the thoughts of a large language modelAnthropic's latest interpretability research: a new microscope to understand Claude's internal mechanisms 010
Adso @scasey.bsky.social · 24/09/2026It doesn't even matter the details. If a person is going to claim a machine is doing the same thing as the human brain, they need to show why they think that is. Too many people see amazing results and draw the wrong conclusions. 200
Adso @scasey.bsky.social · 24/09/2026Researchers can definitely see nodes and how an LLM made a decision. One of the papers on this showed this with math problems and how it approaches it and how the answer was detached from the actual process used. 100
Adso @scasey.bsky.social · 23/09/2026Because we know how these models work, we can peer inside their processing and look at each node and see why they make that decision. We may not know how the human mind works, but it isn't token prediction. I think the burden of proof here is backward anyway, this is an extra-ordinary claim. 770
Adso @scasey.bsky.social · 22/09/2026The idea that this things know about the world is an ext-ordinary claim and has zero evidence or science behind it. 100
Adso @scasey.bsky.social · 22/09/2026It doesn't, but your claim requires evidence that does not exist. 100
Adso @scasey.bsky.social · 21/09/2026Lmao, come on. It's using token based perdition. We know how LLMs work. We can literally see what each node is doing. 100
Adso @scasey.bsky.social · 21/09/2026Yes, except they cost you 25 million and several datacenters of compute. Maybe the model itself can not be said to be brute forcing it, but models themselves run on an extreme amount of resources. 100
Adso @scasey.bsky.social · 21/09/2026It's fancy in that he knows math and how to prompt an LLM and has unlimited compute. This isn't something someone can just type out in ChatGPT as you implied lol 100
Adso @scasey.bsky.social · 21/09/2026The prompt was written by a mathematician who worked there. As far as creativity goes, it's not an easy thing to define, but I would be skeptical of calling this that. aaronlou.com/jacobian_cou...aaronlou.com 100
Adso @scasey.bsky.social · 21/09/2026It's okay though if uses that exact same power to demolish a building or hold it hostage. 100
Adso @scasey.bsky.social · 21/09/2026It's the way people dismiss people getting together and having fun for a cause. People singing about voting and having fun made this thread really mad and I do think that's pretty sad. 100
Adso @scasey.bsky.social · 21/09/2026You have no rebuttal to the president and Musk literally killing 100s of thousands of people using the same power, pathetic really. 100
Adso @scasey.bsky.social · 21/09/2026I think the main issue is people on social media are jealous that there are others who are passionate about something. 100
Adso @scasey.bsky.social · 21/09/2026It hasn't hurt anyone, and is more good than bad yes. I support bringing issues like Trump's bypassing Congress to steal money for concessions to the forefront. People have been hurt, hundreds of thousands probably died through canceled congressional funding. Weird to hand wave this away. 200
Adso @scasey.bsky.social · 21/09/2026Yes, AI is good at complex problems given enough compute, which is what I literally said. 100
Adso @scasey.bsky.social · 21/09/2026I agree, which is why I think keeping it grounded is important here. An LLM does some crazy things you would not think a computer can do which is why it's prone to so much hyperbole and anthropomorphism. 100
Adso @scasey.bsky.social · 21/09/2026Yes, he'll only use the power of taking congressional money and holding it hostage in exchange for concession for just property. He would never use it for against a University to jail or deport critics. 100
Adso @scasey.bsky.social · 21/09/2026Humans will one day create a brain and it will be 1:1 with a real brain given enough time and progress, physics is physics, brains are made out of known particles. That doesn't mean LLMs are alive lol. 000
Adso @scasey.bsky.social · 21/09/2026Whether or not you consider it brute force or not, the amount of compute used was insane They used 300 billion output tokens, or 22 million dollars worth of compute. 110
Adso @scasey.bsky.social · 21/09/2026I don't think this makes it creative. It makes it a powerful tool that is really good at solving complex problems. We have know this about ML for a long time, LLMs take that to a new level now doubt about it 100
Adso @scasey.bsky.social · 21/09/2026If you read the paper though, most of the work being done here was coordinating agents, training agents, etc. They didn't just had over a question and 90 minutes later the result came out, it was lots of prep work. 220
