RunT1ME

I've been thinking about the Halting Problem on and off lately, and how it relates to the unbounded recursion of PA. Unbeknownst to me until recently, Gödel put it brilliantly in his letter to Von Neumann:

@jdegoes (2).Cup problem to gemini thinking: "Believe it or not, your cup is likely in perfect working order. You just need to flip it over. Once you turn it upside down: The sealed top becomes a solid, leak-proof base. The large hole in the bottom becomes the opening you drink from."

@jdegoes (1) Big and small SoTA LLMs both ingest the entire internet and more. They have enough data. But synthetic reasoning and RL make the SoTA Big ones *reason*. It's not a perfect world model, they make mistakes, but it's different than just text matching

Remarkable example where an AI thinks a chair is 'not suitable' for windy areas. Why? Because the chair is 'very heavy' and 'cannot' get thrown around. The use of 'cannot' generally indicates criticism, which is where this is coming from. World models > language models.

@Bashmohandes @AhmedShaaban3m not just leverage, but leverage with 'no recourse' in many states. A lower bound on the amount you can lose and 5x on any upside.

@AhmedShaaban3m The idea with Real Estate is the low cost leverage, so even if the percentage is lower, the leverage makes a huge difference.

الاسهم ام العقارات؟

@ciphergoth It's brilliant, recursive meta-satire. In reugritgating trite arguments for lack of model intelligence, he shows that most humans in fact are simply stochastic parrots. At least, I hope that's what he was going for.

That Quillette would publish this nonsense is giving me severe Gell-Manns.

@JustDeezGuy But if we have types that specify intent and the feedback loop is as fast as compiler errors, AI will be even better at writing code (with our help).

This AI hype bubble (#4 in my lifetime) will burst, but when it does, it might just leave in the crater an awareness of the Curry-Howard Correspondence those of us in software have been unsuccessful in advancing for 40 years.

@DrRichardPitre @Mathonymics hard disagree here. Logic is about as fundamental as you can get. There are meta axioms one cannot get around if you try to lean into intiuitionism. Just my two cents.

@Mathonymics Mathematics is neither invented nor discovered. Logic is an extension of linguistic communication that augments instinctive behavior. It emerged from an evolutionary process.

was mathematics invented or discovered?

@rakyll Is it hard for them to find project ideas? My issue is that I can't get many off the ground because I need consensus from 8 different UTLs in my org to make big changes. 🥲

Every day I’m talking to 1-2 L7s and L8s who are trying to find a truly hands on project to become a pure IC again. The dynamics of engineering has changed.

@EcZachly I did multi-team interview loops at Google, Netflix and Microsoft (Azure) simultaneously. Two of those three companies impressed me a lot, one was an absolute clown show. Guesses?

When I worked at Netflix a guy on the reliability team got fired for losing the 4th 9. Are we even trying anymore?

@DrRichardPitre @JohnFugelsang This person has no concept or even ability to describe what 'that' person is hoarding. There may be reasons to dislike rich and powerful people, but if someone thinks they're hoarding, they're retarded.

@JohnFugelsang So you think that managing billions of dollars of capital is equivalent to hoarding bananas. A chimp might see it that way because the concepts needed to understand the difference would be beyond him.

Azure.

@Rifty_Bear @PubliusDB @JeremiahDJohns Haha well maybe some have taken too many shots to the head this side or the Atlantic. But a goose or coyote is getting choked out.

@runT1ME @PubliusDB @JeremiahDJohns We do rugby over here, I’m not sure why that make someone feel they can stand up to a charging elephant…

Can't decide which is crazier - the guy who thinks he could beat a grizzly bear, or the guy who thinks he'd lose a fight to a rat.

@PubliusDB @JeremiahDJohns A large amount of high school boys in the US enroll in wrestling or american football. They spend hours a day after school learning and getting used to being tackled, slammed, flipped and more.

@JeremiahDJohns No comment on the fact that Yanks are more confident than Brits?

@jdegoes Some of this gets into speculation/opinion: they were bloated, but it takes a lot of engineering to do an ad network. How much that affected revenue/was good or bad business I dunno. Also, according to the "internet", they use GCP a lot more now. 🤫

@runT1ME They were fat and bloated and now the product works as well as it ever did for a tiny fraction of the annual spend. (The advertising/revenue crash was real and came down to free speech and politics.)

He's missing the culture component here. Look at Tesla and Waymo. Who is winning there? Look at Gemini vs Grok. Elon's companies have a problem retaining talent. They might be good at identifying it, but talent leaves at a much higher rate.

@jdegoes If we are talking fully autonomous robotaxi revenue ,it’s 0 to 350m.

@runT1ME Tesla: 94B, Waymo: 350m. From a revenue perspective, Waymo is a blip of insignificance.

@jdegoes As for Twitter, they had way too many employees supporting a custom stack that was unnecessary. But you can't say laying off the staff was a 'success', their ad network fell apart along with their uptime and their cloud bills went WAY up.

@runT1ME Tesla runs circles around Waymo. Twitter (X) continued to function after nearly all the original engineers were fired. Talent, it would seem--and, at least, with a capital 'T'--is grossly overrated.

@BlakeADriver @tomfgoodwin Hmm, this isn't right. By what metric do you think Meta is better at it than Google? Have you seen Llama benchmarks?

@tomfgoodwin You know Meta is an AI company right? And better at it than Google and Amazon.

It’s hard to underestimate how smug Apple is going to be about this in 3 years time. Why on earth did people with zero right or need to do this, pile in. Google I get Amazon -sure Facebook - you sell ads against stuff people provide you for free.

@DrRichardPitre @elonmusk Do you think the universe is continuous?

I hope that the wine was really good. What is a "simulation"? I'm gonna go with "pieces of software that run on a Turing machine" for 500 Alex. So the computer scientist also happens to be a defender of the Turing faith. He believes, hallelujah, that the essence of our reality can be -completely- (the simulation hypothesis) captured by finite one dimensional strings of symbols and a Turing machine doing a River Dance on those strings; amen brothers and sisters. The physicist, because he must have been a truly "top" physicist, knows how fcking strange the universe has thus far revealed itself to us to be. Being "top" he also knows in his heart of hearts, that the universe is probably infinitely stranger than even that. Hopefully, the physicist was not openly laughing at the Turing faithful's beliefs ( but if the wine was good ...). From the physicist perspective, the probability that linguistic expressions and their logical transformations can capture the essence of our reality at all levels, has a zero (0) percent probability of being true. To him, it is obvious that language and logic (this extremely useful but explicitly limited tool that evolved as an extension of our limbic system) will ultimately fall short of encompassing the essence of all things.

I had dinner once with a top physicist and a top computer scientist and asked what they thought the probability was that we were in a simulation. They answered simultaneously at 0% and 100% respectively. It was like a double-slit experiment, but with humans.