Too many bangers in this paper to list, but as a whole this is a comprehensive rebuke to those who claim anything resembling cognition of large language models.
-
Too many bangers in this paper to list, but as a whole this is a comprehensive rebuke to those who claim anything resembling cognition of large language models.
-
Too many bangers in this paper to list, but as a whole this is a comprehensive rebuke to those who claim anything resembling cognition of large language models.
@mttaggart Just half-way through the abstract, but this is *brutal*:
"we argue that LLMs are not epistemic agents but stochastic pattern-completion systems, formally describable as walks on high-dimensional graphs of linguistic transitions rather than as systems that form beliefs or models of the world"
-
@mttaggart Just half-way through the abstract, but this is *brutal*:
"we argue that LLMs are not epistemic agents but stochastic pattern-completion systems, formally describable as walks on high-dimensional graphs of linguistic transitions rather than as systems that form beliefs or models of the world"RE: https://infosec.exchange/@mttaggart/116235185090539701
@mttaggart I love it when a sentence says more than a thousand words: "A central open question is how judgment itself is instantiated and operationalized in LLMs"
-
RE: https://infosec.exchange/@mttaggart/116235185090539701
@mttaggart I love it when a sentence says more than a thousand words: "A central open question is how judgment itself is instantiated and operationalized in LLMs"
So long to all this bullshit cognitive metaphors they use for LLMs!
This paper burns them all ! -
Too many bangers in this paper to list, but as a whole this is a comprehensive rebuke to those who claim anything resembling cognition of large language models.
@mttaggart well, they ain't pulling any punches.
-
@mttaggart Just half-way through the abstract, but this is *brutal*:
"we argue that LLMs are not epistemic agents but stochastic pattern-completion systems, formally describable as walks on high-dimensional graphs of linguistic transitions rather than as systems that form beliefs or models of the world"@hakona @mttaggart I take it that you mean 'brutal' here in the sense of 'brutally fantastic'? It is an extremely cool language to hear LLM described for the statistics it is rather than according to what CEOs like to present it as
-
@hakona @mttaggart I take it that you mean 'brutal' here in the sense of 'brutally fantastic'? It is an extremely cool language to hear LLM described for the statistics it is rather than according to what CEOs like to present it as
@i_dont_like_ai Brutally savaging all the AI cultists. @mttaggart
-
@i_dont_like_ai Brutally savaging all the AI cultists. @mttaggart
@hakona @mttaggart You mean that there are any AI cultists left anywhere?
-
@hakona @mttaggart You mean that there are any AI cultists left anywhere?
@i_dont_like_ai I don't think they know themselves, that they no longer have any kind of credibility left and should go hide under a rock. They are the walking discredited. (what passes for walking dead in academia) @mttaggart
-
@i_dont_like_ai I don't think they know themselves, that they no longer have any kind of credibility left and should go hide under a rock. They are the walking discredited. (what passes for walking dead in academia) @mttaggart
@hakona @mttaggart true. But insofar as these 'AI cultings' people read statistics, they will see that there is only one item viewed less positively than their petty tacky el-stealing "AI" at least in the US public view, and that's the topic "Iran".
-
M mttaggart@infosec.exchange shared this topic
