Abhinav SB retweetledi
Abhinav SB
864 posts
Abhinav SB
@Abhhinav_SB
Redefining the future of scientific discovery with AI systems engineered for accelerated innovation. | Cofounder and CTO at https://t.co/mpr5GZfcJk | AI4Science
Hidden dimensions Katılım Ekim 2018
749 Takip Edilen291 Takipçiler
What we’re seeing recently is fascinating.
Several decades old mathematical problems and conjectures are being solved or disproved by frontier AI models often by discovering rare constructions, counterexamples or structural patterns that humans struggled to find for decades.
Where does this matter beyond mathematics? Material science, chemistry, biology and protein engineering. Many breakthroughs come down to finding the right combination of atoms, molecules, materials, or protein structures within an huge search space.
What models need now to perform better in this space isn’t just more capability it’s better problem framing. Frame the real world problems in a way that they can understand them and design the right search, reasoning and verification loops around them.
steve hsu@hsu_steve
GPT 5.6 solved an open problem in quantum information theory, related to distillation of mixed states. Refining, combining, and testing NN architecture ideas that already exist in the AI/ML literature is less difficult than obtaining this result. RSI seems not far off...
English
Welcome a new warrior to the battlefield
I’ve been refreshing Jensen Huang’s profile for a minute and every refresh adds a few thousand followers.
Jensen Huang@JensenHuang
For my first post, I’m sharing a letter @NVIDIA signed on why open models matter. AI will transform every industry, power every company, and be built by every country. Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty. The world needs both frontier closed models and frontier open models. images.nvidia.com/pdf/Open-Weigh…
English
Feels like an AI tsunami on maths 😂
Jake Brukhman@jbrukh
I have tracked 16 math problems recently solved by AI. - 9 solved by counterexample: 56.3% - 7 solved by proof: 43.8% - 10 of 16 have formal proofs: 62.5% - 8 of 16 have exact finite witnesses or certificates: 50% - Among the 11 with known posing dates, they had been open for an average of 47 years
Filipino
@skdh It’s either a game over situation or a straight jump to the final boss level once it step into the area of theoretical physics.
English
At this point…I don’t think Anthropic is trying to become just the best AI company anymore. It feels like they r aiming for something much bigger. And honestly I have a feeling Dario Amodei’s real vision is something many people wouldn’t like if he ever openly shared it
Grigori Karapetyan@GregKara6
bro anthropic tweaked their classifier to trigger on math research now. these kneejerk reactions are hilarious man. someone at anthropic: quick quick turn off the intelligence faucet for the common folk so we can be the first to solve math! desperate and hilarious. me and claude have been working on this research for months, including with fable since day one of release (even when classifier was very strict) and now, all of a sudden, that people are solving math problems and its trending on X. fable classifier starts refusing on every single prompt
English
Imagine a scaled up version of these with full self driving capabilities…delivering parcels across any terrain.
And imagine swarm of these…we’re way closer to scifi than we realize. The future is going to be unrecognizable…man we’re cooked
Unitree@UnitreeRobotics
Unitree Super Athlete AS2-W 😉 Fluid movement, compact yet exceptional. Supports a continuous 16kg payload, with 30+ km unloaded driving range
English
@pmddomingos May be now a days ‘Stanford dropout’ tag giving them more value iife than ‘graduated from Stanford’ tag 😂
English
There are no students left at Stanford. They all left to do startups.
English
Everyone’s excited about frontier models solving insanely hard math problems.
It’s cool…but what I would actually love to see is someone take a cheap model that’s been out for more than 6 months (like DeepseekV3pro) and get it to solve genuinely interesting reasoning or open math problems using clever feedback loops, self improvement or with smart orchestrations.
To me that’s what a real AI breakthroughs in problem solving looks like not just throwing more compute at bigger models but finding better ways to make existing ones think. So that those orchestrations or feedback loops could make much bigger results using the frontier models
Matt Shumer@mattshumer_
We are now in a world where anyone can just ask AI to “do a breakthrough” in science and the AI… will. Update your priors.
English
Abhinav SB retweetledi
@mkratsios47 there is no such thing as a "distillation" attack.
if we're going to play that game, then Anthropic's gonna need to answer for one of the largest IP and copyright thefts of all time.
English
Abhinav SB retweetledi
Introducing SensorFM, a large-scale Sensor Foundation Model that learns from 1 trillion-minutes of unlabeled wearable data drawn from five million consented participants.
SensorFM learns a single, reusable representation of sensed human physiology that transfers across cardiovascular, metabolic, sleep, and mental health, as well as lifestyle and demographic factors.
More →goo.gle/4ycJvot
English
Abhinav SB retweetledi
Introducing NEO’s 25 Degrees of Freedom, tendon-driven hands — nearing or surpassing human-level dexterity, strength, speed, and reliability.
For seventy years, robotics worked around the hand problem. The humanoid bet is the reverse: it lives or dies at the fingertips.
English
I don’t fully agree with the pause and play until 2040 approach, but one idea in it is genuinely important for protecting humanity is we need full transparency and openness in all AI research and development. Secret, private R&D racing toward superintelligence should be restricted globally especially automated loops of research.
Frontier labs must slow down so that research automation happens in an open source way, not behind closed doors.
The US and Chinese governments should actively encourage as many labs as possible to catch up to the frontier.
Otherwise there are two possibilities is there either a terrifying concentration of power in a few hands or an existential crisis level threat for humanity.
English
AI 2040 has been released! What a day.
AI 2027 was probably the most forward-looking blog about the future, and much of it has come to pass with remarkably accurate timing.
Alongside Aschenbrenner’s “Situational Awareness,” it is probably the most important blog for theoretically preparing ourselves for the future. That makes me all the more excited about AI 2040! I’m going to read it immediately, and I can only recommend that you do the same.
English
Abhinav SB retweetledi
Announcing our $130M Series A to build the Open Superintelligence Stack
Led by Radical Ventures, with NVIDIA, Intel Capital, Dell Capital, and existing investors
Train, deploy, and continuously improve your own models using our stack.
Own your intelligence.
English
Abhinav SB retweetledi
there will soon come a time, perhaps next year, when voice interaction + multimodal reasoning models are good enough and fast enough that you can actually just program like this
English
Abhinav SB retweetledi
