Clemens Helmut Sageder
486 posts
Clemens Helmut Sageder
@clemens1
@eth zurich, @stripe (patrick is awesome), soon @princeton
Zurich, Switzerland Katılım Temmuz 2021
1.5K Takip Edilen249 Takipçiler
Stripe is in discussions to acquire OpenRouter. Possibly for as high as $10 billion.
English
@_sholtodouglas @Mascobot we will also finally be able to execute feynman’s nanoscale robotic “arms”.
calteches.library.caltech.edu/1976/2/1960Bot…
English
@Mascobot Yeah pairs of arms probably a few k at most, ABB arms way too expensive - lots of scope all the way down the stack for rare earths/motor manufacturing/cheap arms
English
Sometime next year pairs of arms become reliable enough that they can do work on an assembly line or fulfillment center I think?
In the same way that ~the most important trendline to have a view on is how many GW of accelerator capacity are deployed in 26/27/28/29, you should also have a view on how many pairs of arms are built and deployed
Generalist@GeneralistAI
Why create robot intelligence for just one hand, when we could have it learn from many? GEN-1, our latest embodied foundation model, now supports a broad range of end effectors from 5-finger hands, to specialized tools, and everything in between.
English
@cremieuxrecueil von neumann was right game theoretically
English
“If you say, ‘Why not bomb Kimi tomorrow?’ I say, ‘Why not today?’ If you say, ‘Today at five o’clock,’ I say, ‘Why not one o’clock?’”
Director Michael Kratsios@mkratsios47
We have information that Moonshot AI distilled Anthropic’s Fable for the development of its K3 model. To do this they developed a sophisticated internal platform to conduct large scale distillation against U.S. models, allowing them to quickly switch between multiple methods of access to avoid detection. Moonshot AI has also acquired GB300-equipped servers and has accessed GB300s in Thailand, likely to train its AI models. The United States strongly supports the free and fair development of AI, including a thriving competitive ecosystem that spans frontier models, specialized systems, open-source frameworks, and open-weight models. Legitimate AI distillation used to create smaller, more efficient models plays a vital role in this open innovation ecosystem. However, large-scale, covert industrial distillation aimed at stealing proprietary U.S. technology and undermining American research is unacceptable.
English
@marty188586 i still hope magic generalization will kick in once we’ve rl’ed in enough domains
English
I now truly believe if you can't RL something, agents will behaive at GPT-3.5 level
Tried very hard to get Sol to design interpablity experiments on niche arch then Terra to code. Sol's idea are always bad and treated things black and white. Never it considered we know nothing
English
@CJHandmer it’s going to be the singularity in the truest sense
English
@scaling01 yeah, he literally claims to have invented everything
English
@max_paperclips @cremieuxrecueil respect this reference, this is literally sooooo accurate
English
@cremieuxrecueil sophon locking potential competition. they claim they got rid of that (they mentioned it also involved nerfing with PEFT adaptors, steering vectors, and yeah, silent downgrading). But do I believe they did away with that entirely? Ha. Use OpenAI for serious work
English
we know why Fable can't do it
Crémieux@cremieuxrecueil
GPT 5.6 Sol is actually able to do CUDA acceleration now. Fable still can't do it, but Sol managed to give me a 1.4x speedup on something I was able to speed up 22.5x by hand. I don't get why this is still so hard for LLMs to do, but whatever, at least there's progress.
English
@scaling01 it needs to be way better than glm 5.2. especially as 1.5T is double glm's parameter count
English
The 1.5T model is coming tomorrow
I have high expectations with all that cursor data and it's much bigger than all the Grok 4.X versions
It should be at least on par with GLM-5.2
English
@jxnlco it’s annoying if you want a quick answer between two longer tasks but you set reasoning to xhigh because it’s just too much friction to change it. something like “\low [prompt]” would be nice (ie only set reason for one prompt.
English
@MollySOShea i’ve set my claw up to buy the merch and let my gf palantirmax
English
Palantirmaxxing 🎾 🇺🇸
Eliano A Younes@eliano
The Palantir Tennis Club 🎾 > women’s tennis skirt > women’s tennis polo [July 9 • 9:30am est]
Indonesia
@BillAckman have you been to morcote on lake lugano; it’s stunning
English
If you find yourself on Lake Como, our family’s favorite restaurant is Navedano, ristorantenavedano.it/en/ a family-owned, beautiful restaurant set in spectacular gardens with superb food at reasonable prices.
You will love it.
English
@markpinc if they actually just focused on profitability they should be valued way higher
English
This chart is amazing! Also note that $SNAP ranks so high but is considered worthless by wall st.
Deedy@deedydas
WhatsApp may be the statistically best app we've ever made. It's a product manager's dream with its unparalleled addictiveness (DAU / MAU) of 87% and stickiness (M1 retention) of 86%, both #1 in the world while having the #1 most monthly active users for any non preinstalled app of 2.7B users. Here are the top 25 most used apps in the world by MAU on both these metrics. Some surprising observations: — There are now 15 1B+ user apps in the world, 8 by Google, 4 by Meta. — The 3 that aren't are TikTok, Telegram (!) and ChatGPT — Telegram has more users than Spotify, Pinterest, Netflix, Amazon, Snap! — ChatGPT's one month retention is #5 after WA, Instagram, Chrome and Youtube. 2yrs ago, it was been #20 by M1 retention — Shopee, a shopping app in southeast Asia, is huge and retains users better than Amazon! Useful way to break down consumer businesses especially within certain categories (X vs Reddit vs Threads is a good one). It's shocking how few new apps have been able to break through in the past 10yrs.
English
@ai_for_success @elonmusk it’s in the ballpark of opus so clearly not
English
@elonmusk Will it also be blocked by Govt just like Fable 5 and GPT 5.6 ?
English
Grok 4.5, based on our 1.5T V9 foundation model, with Cursor data added in supplemental training, is now in private beta at SpaceX & Tesla. Early evals show performance close to, perhaps exceeding Opus.
RL is continuing to significantly improve the model, and the Grok Build harness gets better every day.
Nice work by all those involved!
Completely trained from scratch new models will be released by @SpaceX every month this year.
English
@aidan_mclau i need at least a movie of the fall of gondolin
English
if i became a billionaire i wouldn’t tell anyone but there’d be signs
Phelan MacKenzie@PhelanMackenzie
They say The Silmarillion is unfilmable. But what if we were to focus on one part of it, say The War of Wrath. Yes it lasted decades, but events like the fall of Morgoth & the slaying of Ancalagon the Black, it could make for an epic trilogy. #Tolkien #Fantasy
English
@PhelanMackenzie omg yesss, there are so many parts that would in itself be awesome movies, especially the fall of gondolin
English
