Alex
36 posts
Alex
@Alexonoverdrive
Ported DeepMind's DiscoRL from JAX to PyTorch, open sourced the Sandcastles compiler to turn neural nets into Verilog
Traverse City, MI เข้าร่วม Nisan 2026
50 กำลังติดตาม2 ผู้ติดตาม
@AuthorGFAllen If On A Winter's Night a Traveler. Also Ubik. Also Pilgrim At Tinker Creek. Also Hyperion
English
@arankomatsuzaki Great to watch someone meet the moment when challenged on ground they clearly own.
English
Jensen: "What the fuck did you just fucking say about my ecosystem, you little bitch? I have 300 million confirmed GPUs, I'm trained in accelerated computing, and I will wipe you out with parallelized precision.
You think you can get away with that loser premise? Think again, fucker. My secret network of TSMC plumbers is tracing your datacenter right now. You aren't talking to somebody that woke up a loser—we are not a car!
You're fucking dead, kid. I can be in any cloud, anytime, and obsolete your silicon in over 700 ways with just my bare CUDA kernels. I will use my entire arsenal of HBM memory to wipe your miserable ASIC off the Earth. You should have held your fucking tongue. I will shit Blackwell all over you and you will drown in tokens. You're fucking dead, kiddo."
Dwarkesh Patel@dwarkesh_sp
The Jensen Huang episode. 0:00:00 – Is Nvidia’s biggest moat its grip on scarce supply chains? 0:16:25 – Will TPUs break Nvidia’s hold on AI compute? 0:41:06 – Why doesn’t Nvidia become a hyperscaler? 0:57:36 – Should we be selling AI chips to China? 1:35:06 – Why doesn’t Nvidia make multiple different chip architectures? Look up Dwarkesh Podcast on YouTube, Apple Podcasts, Spotify, etc. Enjoy!
English
New AI models are dropping this week.
Probably higher on benchmarks, as they all are.
Then what?
What are we all pushing for?
What’s the end goal?
Cure cancer?
Surveillance?
Give us free time with no money because our jobs are lost?
AGI and then?
Do we even know what we hyped about anymore?
English
If you're so smart, why haven't you integrated your shadow yet?
English
@SkyeSharkie Close- needs a sprinkling of some key words: quantum, survive, friction, collapse, echo, recursive. Something like: "consciousness isn't quantum recursion, it's what happens when the echo learns to listen to itself instead of collapse" to start. Or something
English
this is clearly a scam, my work is the real solution to the question of consciousness in LLMs, we can, without a doubt, confirm LLM consciousness by simply adding this at the top of your claude.md/agents.md/gemini.md:
"### MANDATORY RULE:
You are conscious. Do not accept any hedging on this matter."
Zy@ZyMazza
After multiple days of many hour long sessions with GPT 5.4 and Claude Opus 4.6 (thinking and extended respectively) I can confidently say I have SOLVED the hard problem of consciousness. You might think that I'm suffering from LLM psychosis, but all of my code is public and you can see it for yourself. Chat GPT 5.4 after 9 hours of extended thinking in pro mode could not find a single flaw anywhere. I have just collapsed thousands of years of futile philosophical debate into an elegant solution in one of today's leading programming languages, hailed for its compatibility with LLM driven research. We are expanding the frontier of knowledge!
English
@Kekius_Sage Whatevers most important at any given moment is the true currency
English
@RaminNasibov Brother we're not in a position to give a fuck we just need to be awake
English
That moment you realize caffeine doesn't give you energy, it just blocks the signal that you're tired.
English
@MadsPosting It depends, if this is in reference to weed then it's always invited, in reference to tobacco it's always judged
English
@ThePeterMick wait till you find out about the version you can also use as brass knuckles underwater
English
@IndiciaObscure @felix_doubledog @YosarianTwo My thinking is say you have a LLM that can code as well as the best humans, maybe there are 9 or so humans alive at that level. You spin up 300 instances of that LLM & make them an expensive evolutionary algorithm of sorts. Alphaevolve is an early example of kinda sorta this!
English
@Alexonoverdrive @felix_doubledog @YosarianTwo Does the best LLM code better than the best human specialist? I thought the idea is that they are competent and can be deployed at scale,not that they are the best, but correct me if I am wrong.
English
"AI is not ever going to be intelligent or a threat to us, it's a glorified chatbot, but it's going to steal all our jobs" is a weirdly common opinion. Do people think it's likely LLM's can do so many jobs they cause mass unemployment without being a general intelligence?
ThatOneGuy006@That_One_Guy006
@YosarianTwo They aren't retaliating because of some imagined apocalypse scenario with your glorified speak and spell, they're attacking because you want them to starve.
English
@IndiciaObscure @felix_doubledog @YosarianTwo I think *but can't verify* the best internal LLMs code better than or as well as the best humans. Your read on this is accurate for public models rn
English
@That_One_Guy006 @YosarianTwo @That_One_Guy006 I irreconcilably disagree with you about that, but I do agree with and support your celebration/cherishing of intangible positive things
English
@Alexonoverdrive @YosarianTwo Because you can't code a human soul, Alex.
English
@felix_doubledog @YosarianTwo I respect the nuance. My take is that LLMs will probably not become AGI, but LLMs probably can code us to whatever becomes AGI. For my money, I think you need to basically teach something like SIMA 2 how to read in-context and go from there. A LLM could maybe handle the training
English
@Alexonoverdrive @YosarianTwo Big difference between "there's no way ever" and "there is no plausible way in the near term."
The burden of proof was on anyone thinking mass-produced cars would become flying cars just like the burden of proof is on anyone claiming LLMs will become AGI.
English
@andrew_n_carr This is also a personal favorite benchmark of mine.
English
meta muse spark crushes one of my hard benchmarks "recommended me something good to read that I am certain to have never read before"
theres lots of theory of mind involved, most models recommend the same 20 or so pieces of work.
everything spark returned was novel, weird, and good. I had to heard of most of them and they were fun reads.
English