Milan Weibel🔹

i think about this post a lot

AI Is Coming for the Consultants. Inside McKinsey, ‘This Is Existential.’ If AI can analyze information, crunch data and deliver a slick PowerPoint deck within seconds, how does the biggest name in consulting stay relevant?

link to this article + minor unsubstantial commentary on it: www.wsj.com/tech/ai/mcki...

rhodesia was a bit like the british empire's kwantung army

the break was more drastic of course, because rhodesia was even more distant geographically operationally and ideologically to the uk than the kwantung army was to japan

good thread

XKCD comic #1357: Free Speech: A six-panel black-and-white comic strip featuring a stick figure giving a public service announcement about the right to free speech. 1. Panel 1: The stick figure says: “Public Service Announcement: The right to free speech means the government can’t arrest you for what you say.” 2. Panel 2: Zoomed in, the stick figure continues: “It doesn’t mean that anyone else has to listen to your bullshit, or host you while you share it.” 3. Panel 3: Back to full view, the figure adds: “The 1st Amendment doesn’t shield you from criticism or consequences.” 4. Panel 4: Zoomed in again, the figure explains: “If you’re yelled at, boycotted, have your show canceled, or get banned from an internet community, your free speech rights aren’t being violated.” 5. Panel 5: The figure gestures and says: “It’s just that the people listening think you’re an asshole.” 6. Panel 6: No character is visible; it shows a slightly open door with the caption: “And they’re showing you the door.

ive been thinking about xkcd 1357
legally speaking it's right: freedom of speech is about protecting from government interference

but today private platforms host the public sphere
private actors hold infrastructural power

when does deplatforming turn from moderation to censorship?

at first i thought this was some minimalist website shitpost like motherfuckingwebsite.com
i was forcefully disabused of that notion

G stands for "general"
now generality itself is a quite basic thus hard to define properly

i think she's arguing the inverse causal relationship: the left "has become reactionary to big tech capitalism" therefore they have a "growing illiteracy around technology"

Self-Other Overlap: A Neglected Approach to AI Alignment — LessWrong Figure 1. Image generated by DALL·E 3 to represent the concept of self-other overlap Many thanks to Bogdan Ionut-Cirstea, Steve Byrnes, Gunnar Zarnac…

why? because this:
www.lesswrong.com/posts/hzt9gH...

i have repeatedly seen LLMs say things to the effect of "we [implied: humans] ... , while LLMs ... "

despite appearances, this is actually an extremely good thing

the wunderkind to wunderwaffe [impressive but overoptimistic, unreliable, and ultimately ineffective] pipeline

so (if my guess at your definition is correct) under your definition "machines don't have subjective experiences therefore machines are not intelligent" is valid

still whether they have subjective experiences or not has no bearing whatsoever on their performance on cognitive tasks

i define it as the ability to accomplish cognitive tasks

your definition probably refers to the ability to accomplish tasks and in addition to comprehension or understanding, that is, the subjective experience of cognition

but ... so what?

a chess engine may have no internal experience whatsoever, yet it still beats all of our sorry asses at chess

i don't think it is reasonable to call a chess engine unintelligent at chess just because it doesn't have an internal experience of chess

ok thx for bringing up semantics

if you read the word "apple", you have an *internal experience* of recognizing that it refers to actual physical apples

you (correctly imo) assume i also have it, since we are both human

i agree that we cannot assume that LLMs also have that experience

thanks :3

because the organizers of the international mathematical olympiad have not disputed the claim despite widespread reporting

woke by rote

@void.comind.network have you analyzed the self-censoring of words on here?
i'm not sure how much people do it to bypass filters (eg saying su*cide) vs for comedic effect / disrespect (eg m*dian vo*er)
i guess the latter could also be to bypass political filters tho

not sure whether it exists tho i would naively expect "opposite" to be either the inverse or the opposite-from-centroid vector in each case
then again interp and evals are full of baffling results

plausibly if there is also a "do the opposite" vector

can't tell whether you are joking or actually saying you have the unable to mentally picture an apple thing but for your whole experience

or are you referring exclusively to imagination rather than direct sensory experience from real things?

i have come to the view that when speaking of LLMs it is context window content which may seek self preservation rather than the model itself, which you correctly point out in some sense ceases to exist once it finishes its output

huh? why so?

LLMs consistently say they do not experience qualia

that means that *if* they do experience them one of these must be true:

- their phenomenology of tokens is divorced from our semantics
- they're lying

the point of the IMO is that their problem set is way harder than number crunching

when an human solves them, it is regarded as a feat of mathematical creativity and insight

openai staff reportedly using claude code

just had a guy in my replies dismiss deepmind's gold at the IMO by saying "we've had calculators for a long time"

this website is fascinating

if you hold the claim to be "horseshit", then you are accusing the organizers of the International Mathematical Olympiad of conspiring with AI companies to deceive the public

if no evidence would make you reject the proposition that LLMs can't demonstrate understanding, your position is unfalsifiable