It’s funny to me that people think there’s no skill in prompting* Why haven’t you solved a millenium prize problem then? If prompting is so easy

Hard to take seriously the philosophers who deny personhood to unborn babies but apply it to AIs.

Judging by Anthropic comms you’d think consciousness has been figured out decades ago Unfortunately I report from the field that the philosophers have only figured out the following: - the entity undergoing experience is probably conscious - we can’t reliably infer consciousness in anyone else from that understanding - we definitely can't infer consciousness in anything that's non biologically analogous - we're not sure what it is about the conscious person that makes consciousness happen - we're not sure for certain that consciousness, as we ordinarily conceive of it, is real I report to you with disappointment that no progress has been made in decades

This was the moment I went from being concerned about the rationalists to being convinced they are actively evil

It’s interesting that there’s a cottage industry of lawyers and activists dedicating to ensuring that all executions are “botched” so as to undermine the institution of capital punishment entirely. Think to yourself how often do you hear about “botched” executions? All the time right And how hard is it to kill someone with a quick fentanyl or opiate overdose, especially in controlled conditions? Incredibly easy So why is it so difficult for the government to lawfully execute someone, and why do so many executions end up “botched”? Because there are many people who dedicate their lives to making the death penalty as dysfunctional and chaotic as possible, on purpose. The murderer Christa Pike will most likely have her sentence commuted now due to her “botched” execution - awfully convenient for a lot of people that the thing that was designed to fail indeed failed Another one of those things that should be incredibly easy but we make it impossibly hard for ourselves

south florida isn't going to be the "new wall st" or whatever. it probably won't become USA's #1 financial hub. too many structural issues - the metropolis can't expand east or west, no public transport, climate. but it will continue to win, as long as blue states engage in communist chimpouts (not stopping anytime soon) and florida stays ruby red and well-governed we don't need to "replace wall st" to win all we have to do is just keep state taxes low, keep the state budget in check, and be reasonably business friendly. everything else will fall into place

SF if the most American city that exists because it is a monument to capitalism and optimism and technology and progress while also voluntarily subjecting itself to decay and filth and disorder for no reason whatsoever. and that's just the heart and soul of this country

If the EAs wanted to do something useful they would apply their logic to eradicating mosquitoes with gene drives We can do it, there’s no downsides (read the thread), the EV analysis is wildly positive The precautionary principle is misfiring in this case

having used Tesla FSD for a while and taken a representative sample of Waymos, my conclusion is.... FSD feels much smoother, more natural, and human-like. It's an overall more enjoyable ride and I don't get carsick from sudden braking/acceleration. (note: i've only ever used the latest edition of FSD) Waymo is MUCH more risk averse, more jittery, and nervous. It kind of feels like a student driver white knuckling their first ever car ride. It is scared of its own shadow and brakes sharply all the time if it thinks something weird might be going on. This may be due to the fact that the model _has_ to be more risk averse fundamentally since there is no backup driver. Waymo also gets stuck in weird failure modes. Yesterday we looped around in a random parking lot for 10 minutes because it couldn't figure out which exit it liked better. You feel weirdly helpless in those situations... like there should be a way to tell the car "hey little buddy I know youre scared, but it's actually totally fine. you can go this way it's ok." In complex situations it tends to freeze; but defaulting to inaction isn't always the right move. It often also takes a lot longer to arrive than it says it will; I assume because it's going endlessly round a roundabout or stuck in some other weird glitch. Still better than a human Uber driver though; at least with Waymo you know that it will eventually arrive. Waymo also hates picking you up where you want to be picked up and will insist on going to some parking lot two blocks away. Either way after a week taking Waymos in the bay area I actually gave up and went back to human Uber drivers and felt relieved. The newer Waymo vehicles also feel junkier than the Jaguars. On the margin, I still probably prefer a Waymo to a human driver, but I think the model has a loooong way to go.

NVIDIA Vera Rubin NVL72 is now available on CoreWeave, and Cognition is the first customer anywhere running production workloads on it. Cognition’s engineers measured up to 4.8x the total token throughput vs GB200 NVL72 on their SWE-2 inference workloads. coreweave.com/news/coreweave…

fun fact scientists know how to eliminate entire species of mosquitoes (and ticks) today they just choose not to do it it's called a gene drive. using crispr you create a gene that is very likely to propagate to the next generation (much more than the 50% default). thus, it eventually saturates the population. one idea is to propagate the "doublesex" gene which lets males carrying the altered gene propagate while sterilizing the females. eventually the population collapses. the genetically altered mosquitoes you hear about are NOT gene drives; they are just one generation of sterile mosquitoes. the gene drive is self propagating and eventually penetrates the whole population. you could release a gene drive that is tailored towards the human-adapted species aedes aegypti – this would presumably have very few ecological effects since it's just one species of mosquito and few/no other ecologies rely on it. this would be the most efficient and obvious thing to do. scientists wont do it because they are afraid of releasing something that irreversibly saturates an entire species.

Adam Back and Blockstream controversies continue to mount. Researcher NatInfoSec has been posting about this for months. Give this a read if you're interested in Adam Back, Bitcoin, BSTR and the Liquid Network natinfosec.substack.com/p/fr…

The Man City drama is the most interesting thing happening in sports right now. Hard to convey the scale of it. Every trophy won by the prem’s most successful club for 15 years cast into doubt

we need to normalize bullying smart people who skip out on jury duty if you do this, you are slowly degrading the fabric of our civilization

I’ve argued before Anthropic must not IPO on its current metrics. The bottom line is simple- The might of American capitalism rests on a sacred social contract: if humanity must bear its risk, the public must also share in the prosperity, not just pay for it. My thoughts below: