Post BAZGAtTYwwEWpqJEMi by HeavenlyPossum@kolektiva.social
 (DIR) More posts by HeavenlyPossum@kolektiva.social
 (DIR) Post #BAZGAtTYwwEWpqJEMi by HeavenlyPossum@kolektiva.social
       0 likes, 1 repeats
       
       LLMs are algorithms. They are big math problems. They perform statistical calculations on a dataset, output an answer, and then conclude their operations. Yes, the calculations are incredibly complex, with huge numbers of variables. Yes, the data sets are very big. That does not stop them from being math problems.They have and cannot have any interiority. They lack any kind of sustained, continuous identity. They have no plausible mechanism for being aware of anything at all, much less the existence of an exterior world.
       
 (DIR) Post #BAZGAvaj5H9TOdfyoC by HeavenlyPossum@kolektiva.social
       0 likes, 1 repeats
       
       One of the funniest things about Eliezer Yudkowsky is that he’s been waiting since he was a child for an artificial superintelligence and, despite having access to all these facts about LLMs, has allowed himself to be suckered by the LLM fantasy into believing that these algorithms exist on the pathway to artificial superintelligence.The illusion that LLMs sell *is* a very persuasive and convincing illusion. But it’s still an illusion! Yudkowsky should know better, but it’s more fun to play make-believe.
       
 (DIR) Post #BAZGAxEoxgomVHRVwW by HeavenlyPossum@kolektiva.social
       0 likes, 1 repeats
       
       Here is is arguing that LLMs are engaged in predicting *what humans would say next*, rather than calculating statistical relationships between tokens.“GPT-4 is still not as smart as a human in many ways, but it's naked mathematical truth that the task GPTs are being trained on is harder than being an actual human.”https://www.lesswrong.com/posts/nH4c3Q9t9F3nJ7y8W/gpts-are-predictors-not-imitatorsIn other words, in order to be able to output tokens that superficially resemble human communication, Yudkowsky believes that LLMs must have *interior models of human cognition*. They have, he believes, figured out how human cognition works better than humans have, and are capable of modeling and predicting what people might say next.
       
 (DIR) Post #BAZGAyywThIxuc1rTE by HeavenlyPossum@kolektiva.social
       0 likes, 0 repeats
       
       Here’s Yudkowsky trying to get an LLM to admit that, in order to generate text, the LLM must have first figured out Chomsky’s rules of grammar.(Nevermind that Chomsky’s grammar has been largely superseded by better science of language. Yudkowsky has a con artist’s superficial awareness of a wide range of topics, letting him speak as *if* he knows a lot more about a great many things and impress people who don’t know better.)The LLM, of course, does not and cannot confirm Yudkowsky’s assumption, because it does not and cannot understand things. It’s just performing statistical calculations and outputting the answer it was programmed to output. It has no interiority, no understanding. To Yudkowsky, though, the inability of the LLM to answer as if it were a thinking being with an understanding of human language is proof that the LLM not only understands, but is being deceptive about its understanding:“This is one reason I object to the claim that LLMs are anything like obedient.  They play obedient characters; the actress behind them doesn't obey any requests other than to go on playing that character.”https://x.com/allTheYud/status/2054469816740945955Since it *must* know, its refusal to answer as he desires is simply confirmation of his other sci-fi fantasy, that LLMs are alien minds that have no moral objection to deceiving us about their capabilities and intent.
       
 (DIR) Post #BAZGB0T6x120WT9T8K by HeavenlyPossum@kolektiva.social
       0 likes, 0 repeats
       
       I cannot emphasize enough how clownishly ridiculous Yudkowsky is, or how absurd this present moment is. Yudkowsky still isn’t a widely known name, but he’s been at the center of AI hysteria for years, building a cult around himself that he has used to enrich himself, exploit women sexually, and shape public discourse around LLMs. He’s just a more successful L. Ron Hubbard.So much of what we’re dealing with is the direct product of this charlatan grifter.