Janmaarten Batstra

@darkcarnivalart Care to re-evaluate that assessment? ;)

So Elon Musk and several other tech CEOs signed an open letter asking for development of AI to pause for 6 months to come up with safety protocols for AI systems. I hate to be cynical but this sounds to me like these CEOs just want time to catch up with their own products

Remember @elonmusk also petitioned to pause AI development globally? Even though he knew the petition would have no effect. He knows the dangers just as well as any expert, an abundance is just a possible outcome, next to end of humanity. The only reason he (and Anthropic etc) keep racing is because if they don’t, China will win the race. x.com/elonmusk/statu…

This guy is dangerous for AI progress. I’m starting to see him as one of the biggest anti AI, decel voices right now. Someone needs to teach him about the concept of the Singularity and post ASI abundance. AI won’t create the problems he fears, it will solve them. Poverty, pollution, unequal access to education etc. With sufficiently advanced AI, they become solvable. Slowing this down doesn’t make us safer, it just delays a better world for billions of people.

@Noahpinion @peterwildeford And yet the whole world buys it. Our hunger for power (or our fear of someone else grabbing power) makes us powerless to stop AI taking over power. Fate, it seems, is not without a sense of irony 🤷‍♀️

AI accelerationists are doing a terrible job of selling the idea of a world of complete human disempowerment. Why would we want to live in a world where we're merely slaves or pets of a machine? Why wouldn't we just ban AI instead? This is a bad sales pitch.

@JeffLadish @jachiam0 Very important. And very impossible, unfortunately. Unless all parties pause, so also Chinese and nobody secretly continues. And there is the catch, if anyone secretly continues while the rest pauses, that party gets world domination handed to them on a platter.

@jachiam0 Wait but did you even read the website before writing this thread?

I'm going to make a request for some basics from the Pause folks: please outline a practicable version of a pause. Do you mean no training runs above a certain scale? Do you mean furlough the researchers indefinitely? What are you specifically asking for?

@torfsrik Geloven in God was gratis? De hele volksstammen die uitgemoord werden (en worden) omdat hun God niet de God van een ander volk was hebben waarschijnlijk een andere mening over de prijs die ze betaalden voor hun geloof 🤷‍♀️

Vroeger geloofden mensen in God. Dat was gratis. Nu gaan ze naar een therapeut. En moeten ze betalen.

@aelorius @Kaalateetam @trikcode @MistralAI I know. I’m a fan. Unfortunately, the balance of global power will be decided by the big guns, and the smaller ones play no role in it.

@interminded @Kaalateetam @trikcode @MistralAI It s not their goal, they work on small models not in the objective of having the biggest.

USA has ChatGPT USA has Grok USA has Claude USA has Gemini USA has Llama USA has Copilot China has DeepSeek China has Qwen China has Ernie China has GLM China has Kimi What does your country have?

@Kaalateetam @trikcode @MistralAI That is correct! . But as much as I am a fan of open weights and the ethical way mistral does things, their AI doesn’t even come close to the big ones. It’s like 40% compared to ie GPT 5.4, and it’s not catching up.

@interminded @trikcode France has @MistralAI though!

@elonmusk Big Fan of FSD. Also Big Fan of truth. Showing these figures without adding the number of times FSD was about to crash the car but the human driver intervened, is… {insert any term indicating disapproval here}. Don’t tell half the story when the story is already good enough🤷‍♀️

Try Tesla FSD (self-driving)!

@whulsbergen @BM_Visser Beetje van beiden. FSD is (en wordt nog veel meer) veiliger dan mensen, maar deze data laat niet zien hoe vaak Tesla’s met FSD op het punt stonden om te botsen maar door ingreep van hun bestuurder gered werden. Dat zou dus een extra FSD crash zijn geworden.

@BM_Visser Ben FSD fan. Maar ook data-fan en dit is een kromme vergelijking want FSD wordt ongetwijfeld meer gebruikt op snelwegen en andere plekken waar relatief weinig ongelukken gebeuren.

Het gaat snel. Inmiddels veroorzaken Tesla’s met automatische piloot per gereden miljoen kilometers 6x minder een zware aanrijdingen dan Tesla’s met een menselijke bestuurder. Bron: Tesla.

@NUnl Om de uitslagen per gemeente te kunnen zien, moet ik toestemming geven voor advertenties van online kansspelen. Right….

Live verkiezingen | Stembureaus sluiten binnenkort, drukte tijdens avondspits ift.tt/0u2dQi1

@elonmusk Troubling? With AGI / ASI/ Robots fully turning the world upside down in a few years, I would say it’s troubling to focus on some tiny donations (tiny compared to your prediction about GDP and abundance) 🤷‍♀️

Troubling

@xai grok 4.20 is awesome, but sometimes…🙄 Any AI that ignores specific instructions and then messes up the output (redacted for privacy), is a problem. It makes the whole AI unreliable. 🤷‍♀️. Time for some extra ☕️! 😉

@gmiller @grok can you explain what just happened with the tech guy that drastically reduced his dog’s tumor using gpt for oversight, alphafold for protein analysis and Grok for mRNA design? Just give the summary, outcome, cost, and then compare it to the mini rant of the OP

A mini-rant abut AI and longevity. They say "Artificial Superintelligence would take only a few years to cure cancer, solve longevity, and defeat death itself'. This is a common claim by pro-AI lobbyists, accelerationists, and naive tech-fetishists. But the claim makes no sense. The recent success of LLMs does NOT suggest that ASIs could easily cure diseases or solve longevity, for at least two reasons. 1) The data problem. Generative AI for art, music, and language succeeded mostly because AI companies could steal billions of examples of art, music, and language from the internet, to build their base models. They weren't just trained on academic papers _about_ art, music, and language. They were trained on real _examples_ of art, music, and language. There are no analogous biomedical data sets with billions of data points that would allow accurate modelling of every biochemical detail of human physiology, disease, and aging. ASIs can't just read academic papers about human biology to solve longevity. They'd need direct access to vast quantities of biomedical data that simply don't exist in any easy-to-access forms. And they'd need very detailed, reliable, validated data about a wide range of people across different ages, sexes, ethnicities, genotypes, and medical conditions. Moreover, medical privacy laws would make it extremely difficult and wildly unethical to collect such a vast data set from real humans about every molecular-level detail of their bodies. 2) The feedback problem. LLMs also work well because the AI companies could refine their output with additional feedback from human brains (through Reinforcement Learning from Human Feedback, RLHF). But there is nothing analogous to that for modeling human bodies, biochemistry, and disease processes. There are no known methods of Reinforcement Learning from Physiological Feedback. And the physiological feedback would have to be long-term, over spans of years to decades, taking into account thousands of possible side-effects for any given intervention. There's no way to rush animal and human clinical trials -- however clever ASI might become at 'drug discovery'. More generally, there would be no fast feedback loops from users about model performance. GenAI and LLMs succeeded partly because developers within companies, and customers outside companies, could give very fast feedback about how well the models were functioning. They could just look at the output (images, songs, text), and then tweak, refine, test, and interpret models very quickly, based on how good they were at generating art, music, and language. In biomedical research, there would be no fast feedback loops from human bodies about how well ASI-suggested interventions are actually affecting human bodies, over the long term, across different lifestyles, including all the tradeoffs and side-effects. It's interesting that most of the people arguing that 'ASI would cure all diseases and aging' are young tech bros who know a lot about computers, but almost nothing about organic chemistry, human genomics, biomedical research, drug discovery, clinical trials, the evolutionary biology of senescence, evolutionary medicine, medical ethics, or the decades of frustrations and failures in longevity research. They think that 'fixing the human body' would be as simple as debugging a few thousand lines of code. Look, I'm all for curing diseases and promoting longevity. If we took the hundreds of billions of dollars per year that are currently spent on trying to build ASI, and we devoted that money instead to longevity research, that would increase the amount of funding in the longevity space by at least 100-fold. And we'd probably solve longevity much faster by targeting it directly than by trying to summon ASI as a magical cure-all. ASIs has some potential benefits (and many grievous risks and downsides). But it's totally irresponsible of pro-AI lobbyists to argue that ASIs could magically & quickly cure all human diseases, or solve longevity, or end death. And it's totally irresponsible of them to claim that anyone opposed to ASI development is 'pro-death'.

@EconoomZijlstra @lexhoogduin Wederom een voorbeeld van gierend brevet van onvermogen van de EU. Digitale techniek, innovatie en EU lijken gewoon niet voor elkaar gemaakt. Jammer dat digitale techniek by far het belangrijkste ding is in de nabije toekomst🤷‍♀️

De EU wil een eigen, soevereine, Linux desktop. Als je linux wilt kun je dat gewoon installeren. Doe ik ook. Ik begrijp niet wat ze hiermee willen. Werd aangenomen.

@Max33Verstappen Your next helmet design 👍😅 Now grow the moustache, throw some mushrooms in Perez his direction when you see him. And make sure that wherever you walk in the paddock, you openly carry some bananas. That should get the message across without you getting censored again 💪🤩.

Not our weekend… Big congrats to Kimi 👏👏

@elonmusk @OfficialLoganK Humanity blindly trusting AI with existential needs like water, energy, food etc. What could possibly go wrong….🤣

@OfficialLoganK Even code review will swiftly become a thing of the past

The bottleneck has so quickly moved from code generation to code review that it is actually a bit jarring. None of the current systems / norms are setup for this world yet.

@kylegawley Please do not delete this post so we / AI can laugh at it in a few years 👍😅

People who say AI is going to replace humans are either: a) have never used ai b) incompetent internet bros with no real skills who can be replaced easily

@elonmusk @Kimi_Moonshot @_avichawla From pole to finish, good race! 🏎️

@_avichawla Impressive work from Kimi

Big release from Kimi! They just released a new way to handle residual connections in Transformers. In a standard Transformer, every sub-layer (attention or MLP) computes an output and adds it back to the input via a residual connection. If you consider this across 40+ layers, the hidden state at any layer is just the equal-weighted sum of all previous layer outputs. Every layer contributes with weight=1, so every layer gets equal importance. This creates a problem called PreNorm dilution, where as the hidden state accumulates layer after layer, its magnitude grows linearly with depth. And any new layer's contribution gets progressively buried in the already-massive residual. This means deeper layers are then forced to produce increasingly large outputs just to have any influence, which destabilizes training. Here's what the Kimi team observed and did: RNNs compress all prior token information into a single state across time, leading to problems with handling long-range dependencies. And residual connections compress all prior layer information into a single state across depth. Transformers solved the first problem by replacing recurrence with attention. This was applied along the sequence dimension. Now they introduced Attention Residuals, which applies a similar idea to depth. Instead of adding all previous layer outputs with a fixed weight of 1, each layer now uses softmax attention to selectively decide how much weight each previous layer's output should receive. So each layer gets a single learned query vector, and it attends over all previous layer outputs to compute a weighted combination. The weights are input-dependent, so different tokens can retrieve different layer representations based on what's actually useful. This is Full Attention Residuals (shown in the second diagram below). But here's the practical problem with this idea. Full AttnRes requires keeping all layer outputs in memory and communicating them across pipeline stages during distributed training. To solve this, they introduce Block Attention Residuals (shown in the third diagram below). The idea is to group consecutive layers into roughly 8 blocks. Within each block, layer outputs are summed via standard residuals. But across blocks, the attention mechanism selectively combines block-level representations. This drops memory from O(Ld) to O(Nd), where N is the number of blocks. Layers within the current block can also attend to the partial sum of what's been computed so far inside that block, so local information flow isn't lost. And the raw token embedding is always available as a separate source, which means any layer in the network can selectively reach back to the original input. Results from the paper: - Block AttnRes matches the loss of a baseline LLM trained with 1.25x more compute. - Inference latency overhead is less than 2%, making it a practical drop-in replacement - On a 48B parameter Kimi Linear model (3B activated) trained on 1.4T tokens, it improved every benchmark they tested: GPQA-Diamond +7.5, Math +3.6, HumanEval +3.1, MMLU +1.1 The residual connection has mostly been unchanged since ResNet in 2015. This might be the first modification that's both theoretically motivated and practically deployable at scale with negligible overhead. More details in the post below by Kimi👇 ____ Find me → @_avichawla Every day, I share tutorials and insights on DS, ML, LLMs, and RAGs.

@VraserX @SvenUrbanSci So you expect that the current leaders will fully make place for the young? Before AI takes us to post labor (so say a year after AGI)? How realistic you think this is in the given timeframe? Humanity has inertia, AI does not. 🤷‍♀️

@SvenUrbanSci Ok, I have to agree. This will change as soon as the boomers are not in charge anymore.

Post labor society sounds fake to people only because they’ve been psychologically married to wage slavery their whole lives. If AI can produce abundance, clinging to pointless jobs is not noble, it’s Stockholm syndrome. What part of post labor scares people most?