That includes a nontrivial number of chefs.
web: http://maxim.ece.illinois.edu substack: https://realizable.substack.com
6,497 followers 928 following 2,239 posts
view profile on Bluesky
That includes a nontrivial number of chefs.
Dear god, that entire subreddit …
Maybe, but he’s not good at jokes either. bsky.app/profile/bart...
Vesper should be stirred, not shaken.
The whole "clanker" thing is ridiculous.
Come on, AI nerds are too inept for that. Certain Girard-misunderstanding billionaires, on the other hand ...
Cosma has a good take (as usual): bactra.org/notebooks/ev...
The best one so far!
Farmers’ market haul.
what we do in the shadows
www.youtube.com/watch?v=IGNI...
Where do consistent histories go?
But now it also includes culture war slop.
We can have a bit of state capitalism, as a treat.
Seven decades of Soviet rule led to a deeply ingrained cynicism which, somehow, transitioned to willingness to believe any charismatic charlatan on the teevee. What @hannahgais.bsky.social described in this article I witnessed firsthand as a teen in Russia. thebaffler.com/latest/psych...
Where, it should be noted, their definition of communism has lost any anchor in reality and can mean whatever they want it to mean.
Relatedly, what blackpilled me regarding the purported immunity of American system against cults of personality is that, even though the fall of the USSR was in part precipitated by so many powerful exposés of the horrors of Stalinism, they decided, nah, they still want personalist dictatorship.
Many such cases.
A lot of the ex-Soviets also learned to love the culture war slop, alas.
You'd be surprised how many of them still yearn for the *right* kind of authoritarianism and at their mental acrobatics skills to justify the current moment.
Accidental circumcision by falling window sash ranks high up there.
Theodore Porter, The Rise of Statistical Thinking Gigerenzer et al., The Empire of Chance Bernstein, Against the Gods
This has nothing to do with whatever beliefs Weizenbaum may have had regarding humans' direct psychic access to the Akashic records or whatever.
The problem of where to draw the boundary between judgment and calculation is a political problem, not an engineering one. bsky.app/profile/mrag...
Sure, writing a preface to Steiner is crank central, but what bearing does this have on Weizenbaum's distinction between judgment and calculation? This guilt-by-association criticism is every bit as ideological as the one put forward by Bender and co.
And deeply ironic as well, because Scott is a true believer in the AI safety cause and an adept of vulgar rationality a la Nate Silver.
Yep.
Lunch of champions: p’tit basque, roasted almonds, and chilled Crenshaw melon.
Also, it is worthy of note that, for the AI safety people, the ends (preventing the emergence of skynet and bioweapons) justify the means (among other things, immiserating huge quantities of people by normalizing runaway gambling in the guise of "predictive rationality").
An example of this grand guignol: an ML researcher who pivoted to AI safety and fully embraced all of the x-risk doomsday narratives was demanding quantitative evidence to the claims that gambling at scale is harmful. What is that saying by Scott Alexander? "Beware solated demands for rigor?"
Also, it is worthy of note that, for the AI safety people, the ends (preventing the emergence of skynet and bioweapons) justify the means (among other things, immiserating huge quantities of people by normalizing runaway gambling in the guise of "predictive rationality").
I had a somewhat technical thread on Twitter about compression algorithms and underlying conceptual ideas right after Ted Chiang's "blurry JPEG" New Yorker article came out. I have deleted my Twitter account, but perhaps I should resurrect that thread from the downloaded archive.
bsky.app/profile/mrag...
realizable.substack.com/p/how-to-do-...
With some minor modifications, that’s Husserlian phenomenology (you don’t have to be a Platonist to be a realist).
Ah. Got it.
Traditional (Cartesian) dualism is substance dualism.
Keynes thought so :)
Substance monism and property dualism. This is the way.
As for Newton, he had enough epistemic humility to refrain from making hypotheses about the mechanism behind action at a distance.
I personally prefer Philip Agre’s _Computation and Human Experience_. I don’t think he could be accused of being a crank.
That may be the case, but so are the opinions of the AI safety crank club.
Which takes us straight back to the intentional stance.
I’d love to do engineering science of universal next-token predictors, but instead we’re stuck here arguing about everyone’s brand of woo. bsky.app/profile/mrag...
Jack Parsons was a Crowleyite, yet he also founded the JPL. This is a weak argument.
So instead of taking the time to understand transformers’ capabilities as powerful architectures for data-driven approximation of control policies, we’re consigned to this.
Yes, and it’s in large part because chatbots sell, whereas training transformer architectures on tokenized natural signs for applications in science would take much longer but wouldn’t enrich Sam Altman.
One compelling reason why it may be different this time is that, unlike the past examples, here we are dealing with early-generation intentional artifacts (in the sense of Dennett), which novels and rock lyrics are not.
realizable.substack.com/p/learning-f...
Sample-efficiency of human intelligence was paid for in aeons of evolution (“what’s a priori for an individual is a posteriori for the species,” like Konrad Lorenz said).
And this very recent thing as well: eschwitz.substack.com/p/minimal-au...
As I keep saying, everyone should read this: dspace.mit.edu/handle/1721....
Yes, that’s my idiosyncratic brand of constructivism too (that was Herman Weyl’s perspective).
To be fair, his Society of Mind framework was fairly sophisticated. But he did have a bit of a fanatic’s zeal in him.
Constructivism stays winning.
Yes, among others. It’s described in some detail in Pamela McCorduck’s _Machines Who Think_.
let’s go
Not that they hadn’t tried, though. When the book came out, many of the bigshot AI researchers accused Weizenbaum of sour grapes on account of no longer being able to contribute anything original as a scientist.
Also, just like Ben says, you should all read @jessica.bsky.social's essay on the artificiality of alignment. joinreboot.org/p/alignment
Ben is right. The "AI Safety" cult is all bent out of shape about the hypothetical harm from AGI shoggoths, but the real, tangible harm is happening right here and now. I used to think the late Dan Dennett was too alarmist about the danger of "counterfeit people," but he was spot on.
Blackalicious Cannibal Ox Deltron 3030 Antipop Consortium Open Mike Eagle Shabazz Palaces
On the other hand, it’s more realistic than sandpiles or the Ising model.
Pourover of the Beast
Of all people, Chomsky had it right: why is everyone so obsessed with consciousness? Did anyone say anything new about it that wasn't already said in the 18th century?
mitpress.mit.edu/978093675610...
No, no one owes anyone an article on Searle. Searle is irrelevant.
This is primo content.
Helen Andrews. Of course.
Next step: know the difference between y’all and all y’all.
Frankfurt School remains undefeated
Xunzi did it before Cicero (but after Xenophon).
Vapnik-Chervonenkis candy bar
It’s like if Asteriskmag and Vox had a child.
As well as a great Emerson Lake & Palmer song.
_Legitimation Crisis_ and _ Communication and the Evolution of Society_ are genuinely good analyses.
constructed
We don't know how the brain *works* for more or less the same reason and with the same implications.
(With the caveat that "modern AI" is machine learning/pattern recognition using very large, flexible model architectures. But the book does get into that.)
This is indeed a good book.
Everything is control theory, but not everything is optimal control theory.
There's nothing platonic about it, it's pragmatism all the way down.
Everything is control theory. Always has been.
I just subscribed to yours as well.
realizable.substack.com/p/supertzar-...
realizable.substack.com/p/how-to-do-...
realizable.substack.com/p/aristoteli...
realizable.substack.com/p/the-daoist...
realizable.substack.com/p/the-daoist...
realizable.substack.com/p/the-daoist...
Here are some of my posts that touch upon the same ideas you were referring to (connections between intelligence, language, knowledge, cognition, systems, etc.) realizable.substack.com/p/c-west-chu...
I tend to write long posts as well, which is why can only put them out once a month or so.
What S&Z were saying is that either one doesn't need catastrophe theory to arrive at the stated conclusion or Zeeman's invocation of Thom's classification of elementary catastrophes goes beyond what the result says rigorously (e.g., as regards the meaning of terms like "generic," "local," etc.).
There were articles by Hector Sussmann and Raphael Zahler in Synthese critiquing Zeeman's ideas (basically arguing that the results he obtained using catastrophe theory are self-evident and that Thom's theorems don't support Zeeman's interpretation). Some of it even spilled out into popular press.
your daily dose of prog weirdness, enjoy www.youtube.com/watch?v=zflp...
Two different days, two different places to go for breakfast, and both are playing “Come on Eileen”? Reset the simulation please!
for the trollish terminology alone -- casino vs casinoe (rhymes with 'canoe')
read theory
