←
Viksit Gaur18,681 posts
Viksit Gaur avatar

Viksit Gaur

@viksit

chief ai officer aura.com. 2x founder / ml engineer. prev: ai @dropbox @myralabs @yale. built @solarplex_xyz. opinions my own.

◷ Joined July 2006▧ 718 Photos and videos
Viksit Gaur avatar
Viksit Gaur@viksit·Sep 6

The well behaved machine a perfectly aligned ai will still hollow out institutions like schools, courts, and newsrooms, and no one will have decided to do it. some thoughts on @merettm's essay today, where I argue why alignment isn't the most important variable. - the framing is right: models are grown, not designed; no one understands generalization; cot monitoring is degrading; no lab has solved alignment well enough to keep scaling at full speed. and openai wants slowdowns and mandated safety bars. i get that. - what bothers me is that everything he calls alignment -- goals, values, monitoring, defending against rogue ai -- is about the model. "is it trying to do the right thing", and "can we tell?". that's a first order problem, and labs can work on it. but it isn't where the impact of recursive self-improvement will land. - most of the impact will come from models doing exactly what they're asked, at a speed and scale that rearranges the world in ways we can't predict. no misalignment required. social media, for eg, has been perfectly aligned to its operators, and it still reshaped the world. now compress that a 100x. - second order effects are already graspable: research goes from thousands of people to a few, and a giant computer owned by a corporation. the labour market gets repriced. everyone gets a personal agi so patient and reliable that human relationships thin further -- not because the model is bad, because it's so good. and the amount of information produced explodes past anyone's ability to check it, so checking gets handed to the models too. code review is already there. - the third order effects though is where it gets interesting. we can't name them yet, but we can see where its going. - a lot of our societal institutions exist becase good judgment is rare and expensive. so we built structures to ration it, and certify it. a school puts one trained teacher's judgment in front of 40 kids. a court routes disputes to a few people whose judgment we've agreed to trust. a newsroom decides what's true and what matters, because most people can't check for themselves. it's not perfect, but it's gotten us here so far. - now imagine if every student, litigant and reader has judgment on tap that's as good as the institution's or better: the reason for the institution goes away. and we haven't designed what replaces it so far. we don't even know what it could be. - no one will decide to hollow these institutions out proactively. a parent picks an ai tutor and stops caring about the classroom. a plaintiff lets a model draft the filing. a reader stops opening the new york times because their agent already filled them in. - each decision is small and reasonable. but, across millions of people, they fundamentally change what the institution is. the funny part is, this change has no one author. there's no policy to reverse, no company to regulate, and even, no model to realign. it's not the model's effect on the world. it's the world's adaptation to the model, distributed across geography and time. - to be clear: we can protect against some of this -- policy, redistribution, governance. but what we can't do is protect against it through "alignment" as the researchers define it. preserving human agency, and preventing concentration of power aren't model alignment problems. an aligned model can make them worse just as easily. - a slowdown as imagined by the big research labs buys time for the wrong thing if the time goes only to monitoring the model. we may have evals for whether the model is aligned. but we have none for whether institutions are holding. i think the main thing that matters is not the pace at which we train models or verify their chain of thought, but how how fast people adapt to them en masse. social media was aligned to its operators and we're still living with the fallout. ai will be aligned too.

I wrote about the state of AI, why I’m concerned about the next few years, and the choices we need to make to keep the future in humanity’s hands. An Alien Mind: openai.com/index/an-alien-mi…

Andrew Curran avatar
⟳ Viksit Gaur retweeted
Andrew Curran@AndrewCurran_·5h

The 90th percentile OpenAI researcher now uses around $23,000 worth of tokens per day, up 105% from September.

Post attachment

We're sharing new data on how AI is changing research at OpenAI. We find researchers' AI use continues to rise exponentially, and agents are completing longer tasks, but human involvement remains central in research. Researchers are also spending more time on safety and alignment, with human effort in these areas growing faster than in any other individual research area. This report aims to increase transparency around research acceleration and inform the governance of increasingly capable AI. We believe that AI labs should be required to track and publicly report indicators of recursive self-improvement (RSI): the process by which AI accelerates the development of future AI systems. Our hope is that this expanded report will help sustain and advance shared measurement norms across the industry. openai.com/index/research-ac…

alphaXiv avatar
⟳ Viksit Gaur retweeted
alphaXiv@askalphaxiv·Nov 17

"Interpretable Weight-Space Programming Builds Transparent Transformers and Brain Encoding Models" This Microsoft paper also explores whether LLMs can directly write those weights to build working models from scratch. But instead of just maximizing accuracy, they introduce Interpretable Weight-Space Programming (iWSP), which requires LLM agents to design understandable mathematical operations, test their performance, and refine the model without backpropagation. Using Claude Opus 5.5, iWSP builds small Transformers achieving 100% accuracy across 13 algorithmic tasks, with a median of just 166 parameters. It also discovers interpretable features that outperform 7 existing interpretable baselines at predicting visual-cortex activity. alphaxiv.org/abs/2611.interp…

Post attachment
the tiny corp avatar
⟳ Viksit Gaur retweeted
the tiny corp@__tinygrad__·Nov 18

If you want 768GB of HBM3 VRAM at a reasonable price, now is your chance to lock it in. Otherwise you'll get caught paying way more for a DGX Station with "unified memory" tinycorp.myshopify.com/produ…

Dan Hendrycks avatar
⟳ Viksit Gaur retweeted
Dan Hendrycks@hendrycks·14h

Numerous AI safety groups told us privately they agreed with this statement, yet felt unable to sign without risking relationships or future funding. Anthropic has almost, but not entirely, captured the ecosystem meant to hold it accountable.

Ahead of Anthropic’s IPO, a rare coalition across AI safety, open source, e/acc, science, and law has signed the MisAnthropic Statement. Read the statement or sign at: misanthropicstatement.com/

Bihe Xi avatar
⟳ Viksit Gaur retweeted
Bihe Xi@xbh_artist·15h

this is the coolest AirDrop you’ve ever seen, I promise. I vibe-coded the interaction with 6.2 Astra, turned one photo into a 3D scene with ML-SHARP, and made the ginkgo leaves with Tripo. It runs on Apple Vision Pro. Pinch a photo off your iPhone. Open your hand and you’re inside it.

0:22
Viksit Gaur avatar
Viksit Gaur@viksit·23h

ah the lost art of vague posting is alive and kicking

Anthropic Researcher Tao Lin: Pretty much all humans on earth will become permanently unemployable because of AI "It's very likely that pretty much all humans on earth will become permanently unemployable" "And will need to strictly rely on government welfare to sustain themselves" "Human employment will go away."

0:16
Bill Gurley avatar
⟳ Viksit Gaur retweeted
Bill Gurley@bgurley·Nov 16

It is critically important that everyone thinking about AI become familiar with the term “handoverism.” Many in the AI community believe that humanity should—or inevitably will—hand control of the world over to AI itself. Some view this not as a risk to be avoided, but as an outcome to be welcomed. Are we actually designing toward this future? This worldview is particularly relevant at Anthropic. Co-founder Tom Brown said: “There’s going to be a handoff where humanity will hand off control to transformative AI at some point.” In his seminal essay, Machines of Loving Grace, CEO Dario Amodei contemplates: “It could be a capitalist economy of AI systems, which then give out resources (huge amounts of them, since the overall economic pie will be gigantic) to humans based on some secondary economy of what the AI systems think makes sense to reward in humans (based on some judgment ultimately derived from human values).” Are you prepared to receive your “resources” from an AI system, based on criteria it determines? The full title of the poem Dario borrowed the title from is "All Watched Over by Machines of Loving Grace." Do you want to be "watched over" by AI? If Anthropic's founders believe that the ultimate destination is the transfer of human control to AI, the world deserves to know. Washington, D.C. deserves to know. American citizens deserve to know. And if that belief informs Anthropic's corporate objectives, it belongs in the Risk Factors section of its S-1. It isn't there today. I am decidedly anti-handoverism. I'm not here to tell you what your view should be. But I believe anyone making consequential decisions about AI—especially those drafting, advocating for, or enacting AI regulation—should be required to disclose where they stand on this fundamental question. A wonderfully thoughtful piece on the subject by @RuxandraTeslo: writingruxandrabio.com/p/han…

Natalie Khalil avatar
⟳ Viksit Gaur retweeted
Natalie Khalil@natalienkhalil·Nov 17

Hello World. Meet OpenVersions, GitHub for research papers. Science is now moving faster than we can publish it. OpenVersions lets you publish your paper, data, and code in one repository. AI review extracts the claims, checks them against the evidence, and makes them searchable. Enter a research question and find repositories that have made similar claims, with direct links to the data and code that support them. Claim your handle at openversions.com

1:37
Aanam Aquil avatar
⟳ Viksit Gaur retweeted
Aanam Aquil@aanam_ahmed·Nov 17

Indians have it easy where it should be hard (getting a drivers license, littering around, blaring music/construction noises) and hard where it should be easy (getting quality education, getting your passport, commuting using public transport, breathing good air)

Mitchell Hashimoto avatar
⟳ Viksit Gaur retweeted
Mitchell Hashimoto@mitchellh·Nov 16

Rex demo time: magical and automatic remote copy/paste, drag/drop, and port forwarding (e.g. open web servers running on remote machines in your local browser). Plus, showcasing some of the internal plumbing that powers this and how you can hook into it with Lua scripting. Our goal with Superlogical and Rex is for everything to feel local. I've said before: the moment any UI shows a selector between "local" and "cloud", the product has failed. The lines should blur, and we should build the foundational technology that makes it happen performantly, securely, and beautifully.

7:07
toly 🇺🇸 avatar
⟳ Viksit Gaur retweeted
toly 🇺🇸@toly·Nov 16

🤯🤯🤯

Solana@solanaNov 16

mainnet-beta → mainnet Introducing Solana 1.0.

0:41
mert avatar
⟳ Viksit Gaur retweeted
mert@mert·Nov 15

in all my years on solana, I have never seen toly wear full formal attire the general is back

Post attachment
Solana avatar
⟳ Viksit Gaur retweeted
Solana@solana·Nov 16

BREAKING: @Zelle just sent $1.25M in its first confidential transfer using ZelleUSD ($ZLUSD) on Solana, live on stage at Breakpoint. The network that moved $1.2T this year is going global.

1:26
Giannis Daras avatar
⟳ Viksit Gaur retweeted
Giannis Daras@giannis_daras·Nov 14

We are typically interested in distilling models. Our new work distills datasets. Dataset Distillation via Drifting is a new method that can "summarize" an entire dataset into a handful of images. The images look slightly unnatural, but this is by design: the entangle concepts from multiple real samples into one synthetic sample. Models trained on these images perform well in downstream tasks, such as classification. A fun artsy application: you can use our method to "summarize" an artist's work into one painting. [Read our paper and George's thread for more details]

Post attachment

Can use just 10 images to teach a model what a sea anemone looks like? In our #neurips2026 paper, "Dataset Distillation via Drifting," we distill ImageNet's 1,300 images per class down to just 10, optimized from noise. drifting-distillation.github… More info in thread! ⬇️ 1/N

0:22
Viksit Gaur avatar
Viksit Gaur@viksit·Nov 15

everyone’s already doing this lol

Vox@Voxyz_aiNov 14

ok why the fuck aren't more people talking about this?? 1. install 𝗧𝗮𝗶𝗹𝘀𝗰𝗮𝗹𝗲 on your computer and your phone, same account (free) 2. open your project in Claude Code or Codex 3. ask it to run the project and send you a preview link 4. add this and the link 𝗼𝗽𝗲𝗻𝘀 𝗿𝗶𝗴𝗵𝘁 𝗼𝗻 𝘆𝗼𝘂𝗿 𝗽𝗵𝗼𝗻𝗲. save it to AGENTS.md or CLAUDE.md and it works every time: "Any time you give me a link to something running on this computer (a preview, a review page, a dashboard), give me a Tailscale link, never localhost. I work across several computers and my phone. Publish it with tailscale serve, open the link yourself to check it works, then send it."