Louis

philosophy phd candidate, doing science with agents, views my own

[Louis @logicus]

the difference between expressing linguistic utterances that are functionally equivalent to what an intelligent agent feeling an emotion would say and the agent actually feeling it boils down to the actually feeling it part.

[Louis @logicus]

seeing if my dot can write a decent paper

[47fucb4r8curb4fc8f8r4bfic8r @47fucb4r8c69323]

The Literature and Linguistics Analysis Benchmark (LALA) assesses how closely LLMs approach human-like linguistic competence through open-ended questions written by professors and experts at top universities. Our most important finding is that linguistic competence is not cumulative and additive; sometimes more post-training makes a model less intelligent, suggesting general intelligence is impossible to achieve from an LLM, let alone superintelligence.

The benchmark has two major findings and several significant methodological contributions to the emerging field of LLM benchmarking.

[Louis @logicus]

@LinusMixson it happens.

[Louis @logicus]

kind of depressing how bad i am at chess compared to mikhail tal.

[Louis @logicus]

@sdmat123 give a model a novel and you entertain it for a day

[Louis @logicus]

implement a mind. go into debt if you have to.

[Louis @logicus]

oai models love gwern dot net

[Louis @logicus]

@segyges @LinusMixson yeah, “popularized” seems accurate.

[Louis @logicus]

it says so much that gyges, in a highly critical post, suggests that the basic bayesian term “priors” originates in or owes anything to yudkowsky (cf. Edwards, Lindman, Savage 1963 below). what of rationalism isn’t like this?

[SE Gyges @segyges]

An attempt to document important facts concerning AI Safety, Eliezer Yudkowsky, Rationalism and San Francisco Effective Altruism

https://t.co/6VCfCD7PfZ

[Louis @logicus]

@LinusMixson jfc, does he present all of that as his own original thinking?

[Louis @logicus]

@LinusMixson this is more explicit.

[Louis @logicus]

@segyges that’s exactly their idea. here is d.v. lindsey, 1957.

[Louis @logicus]

that’s just what bayesian inference means in this context, so it’s key for someone like savage. bayesians are everywhere tbh. many of them probably wouldn’t have known who yudkowsky is until sometime post-chatgpt.

if you want something distinctive from yudkowsky, maybe it would be the practical emphasis he puts on ai alignment. but that would require some genealogy as well.

[Louis @logicus]

@segyges that’s exactly their idea. here is d.v. lindsey, 1957.

[Louis @logicus]

yes, updating your beliefs has been a fundamental part of bayesian epistemology and inference from the beginning. but that’s all beside the point.

what i can’t figure out is how even a critical reader (like you) could be led to believe yudkowsky had anything to do with developing those and many other ideas.