Adi Mehta
17 followers · 24 following · 2029 views
on the atlas — 83
- Why open-weight models without guardrails are a AI safety risk : NPR1 savers
- Cognition's Scott Wu says AI coding agents shouldn't replace humans | TechCrunch1 savers
- What is a news peg, again? | Journalism 1011 savers
- Chapter 7: Basics of Writing – Broccoli and Chocolate: A Beginner’s Guide to Journalism News Writing1 savers
- Nut graf and lead duos that point readers in the right direction : NPR Training : NPR1 savers
- We Don’t Really Know How A.I. Works. That’s a Problem. - The New York Times1 savers
- AI Deterrence by Betrayal4 savers
- The internet is being rebuilt for machines | TechCrunch1 savers
- What is news? How journalists assess news value1 savers
- Xi-Trump summit showcases Beijing, China's futuristic capital - The Washington Post1 savers
- Bridge or Border: Why Scientific Securitization Won't Make Nations Safe - The Wire China2 savers
- The third wave of American philanthropy - by Nan Ransohoff17 savers
- Four Lessons From $4 Billion in Impact-focused Giving1 savers
- The lies I used to tell myself - by Cate Hall1 savers
- Do Less. - by Cate Hall - Useful Fictions4 savers
- Can You Just Do Things? - by Cate Hall - Asterisk Magazine1 savers
- What happens when AI starts building itself? | TechCrunch2 savers
- AI infrastructure in the "Era of experience"4 savers
- Radical Optionality — Governing Transformative AI Under Uncertainty6 savers
- Mandate of AI - by afra - Concurrent2 savers
- Xi-Trump to talk AI Safety, Huh?2 savers
- We Spent 10 Days Touring Chinese AI Labs. Here’s What We Saw.2 savers
- AI isn’t smart, but neither are we - by Julian Zhang1 savers
- Is xAI a neocloud now? | TechCrunch1 savers
- Say Hi To People You Antisocial Bastards - by Aaron1 savers
- My six stages of learning to be a socially normal person3 savers
- Why Some Men Don't Open Up - by Leo Zaroff - Monismos1 savers
- Secretly Loyal AIs: Threat Vectors and Mitigation Strategies — LessWrong1 savers
- AI Integrity: Defending Against Backdoors and Secret Loyalties — Institute for AI Policy and Strategy2 savers
- AIs Should Have Proactive Prosocial Drives1 savers
- The importance of AI character3 savers
- AI for Decision Advice1 savers
- GPT-5.5 and the broken state of government evals2 savers
- Stop being a pussy - by Natalie Cargill1 savers
- Rebuilding our world, with reference to strong AI - Marginal REVOLUTION3 savers
- AI Tools for Existential Security | Forethought5 savers
- Post 50: Good Research Takes are Not Sufficient for Good Strategic Takes — Neel Nanda3 savers
- Ivan Zhao on X: "Steam, Steel, and Infinite Minds" / X10 savers
- Sarvasv Kulpati3 savers
- How Should AI Liability Work? (Part I) - by Dean W. Ball3 savers
- Give your friends a chance to abandon you - by Kasra2 savers
- Most AI value will come from broad automation, not from R&D | Epoch AI11 savers
- About Me - philkrav2 savers
- How to Make the Most of Your 20s - Catherine Shannon2 savers
- The Math Behind DeepSeek: A Deep Dive into Group Relative Policy Optimization (GRPO) | by Sahin Ahmed, Data Scientist | Jan, 2025 | Medium2 savers
- Dario Amodei — Machines of Loving Grace50 savers
- Introduction - SITUATIONAL AWARENESS: The Decade Ahead50 savers
- no good alone - by rayne fisher-quann - internet princess45 savers
- Learning By Writing42 savers
- A Mathematical Framework for Transformer Circuits39 savers
- Laws of Tech: Commoditize Your Complement · Gwern.net26 savers
- An app can be a home-cooked meal26 savers
- Post 38: On Slack - Having room to be excited — Neel Nanda23 savers
- Why Tool AIs Want to Be Agent AIs · Gwern.net22 savers
- Mini Blog Post 19: On systems - Living a life of zero willpower — Neel Nanda16 savers
- Guessing the Teacher's Password12 savers
- i think about it all the time - by Claire Wang11 savers
- Two kinds of introspection - by Henrik Karlsson11 savers
- Intelligence Arbitrage10 savers
- Beware the Casual Polymath | Applied Divinity Studies9 savers
- Turning 20 in the probable pre-apocalypse — LessWrong8 savers
- An Intuitive Explanation of Sparse Autoencoders for LLM Interpretability | Adam Karvonen8 savers
- Detecting misbehavior in frontier reasoning models | OpenAI7 savers
- The Artificiality of Alignment - by jessica dai - Reboot7 savers
- Measuring AI Ability to Complete Long Tasks - METR7 savers
- 🌱 My blog is a digital garden, not a blog7 savers
- Building Machine Learning Systems for a Trillion Trillion Floating Point Operations - YouTube6 savers
- Post 33: Seek Upside Risk — Neel Nanda6 savers
- AI progress is about to speed up | Epoch AI6 savers
- The Technology of Liberalism - by Rudolf Laine6 savers
- [Guest post] How Substack Became Milquetoast5 savers
- On having more interesting ideas - by Henrik Karlsson4 savers
- Cognitive Security as an AI Safety Cause Area — LessWrong4 savers
- How Did Solar Power Get Cheap? Part II - by Brian Potter3 savers
- Mimetic Desire 101 - Anti-Mimetic—A Field Guide to Mimetic Desire3 savers
- The Problem with OpenTelemetry3 savers
- Open Questions | cguti.xyz3 savers
- Grandmother cell3 savers
- A defence of slowness at the end of the world2 savers
- [UPDATE] From Comfort Zone to Frontiers of Impact: Pursuing A Late-Career Shift to Existential Risk Reduction — EA Forum2 savers
- How Developers Steer Language Model Outputs: Large Language Models Explained, Part 2 | Center for Security and Emerging Technology2 savers
- AI populism's safety problem2 savers
- AGI could drive wages below subsistence level | Epoch AI2 savers
highlights — 1100
One such tool is Heretic, which automates the abliteration process
Why open-weight models without guardrails are a AI safety risk : NPRHugging Face, which hosts open-source AI models, currently lists over 6,000 abliterated models, compared to about 600 in 2024
Why open-weight models without guardrails are a AI safety risk : NPRUnlike with ChatGPT, Claude or Gemini, it's easier to permanently remove their built-in safety guardrails – and the companies behind them have no idea how they're being used
Why open-weight models without guardrails are a AI safety risk : NPRIn fact, Wu has been called one of the most accomplished child competitive programmers of all time, according to a recent profile in Colossus. As a second-grader, Wu won a nationwide math competition for seventh-graders, which launched a childhood filled with math and programming tournaments. It also introduced him to other wunderkinds who went on to launch other AI tech startups, like Scale AI founder Alexandr Wang.
Cognition's Scott Wu says AI coding agents shouldn't replace humans | TechCrunchyou ALWAYS want to have a news peg for your story
What is a news peg, again? | Journalism 101It would be the element of the story that makes it timely and important now
What is a news peg, again? | Journalism 101The point of news writing is not to impress the reader with your vast vocabulary, but to provide information succinctly and creatively
Chapter 7: Basics of Writing – Broccoli and Chocolate: A Beginner’s Guide to Journalism News WritingIt tells the impatient reader where they're headed, which is essential to keeping them as engaged as possible for as long as possible
Nut graf and lead duos that point readers in the right direction : NPR Training : NPRIBM’s chess-playing supercomputer
We Don’t Really Know How A.I. Works. That’s a Problem. - The New York Timesthose stories with which many people can identify are stronger than those which only apply to a few
What is news? How journalists assess news valueThe more companies deploy AI agents, the more pressure there will be to redesign
The internet is being rebuilt for machines | TechCrunchThe shift is emerging across the cloud industry
The internet is being rebuilt for machines | TechCrunchOpenSearch Serverless will integrate natively with AI development platforms like Vercel and Kiro
The internet is being rebuilt for machines | TechCrunchThink of it like always paying for a parking space, even when you’re not using it. With AWS’s upgraded Serverless, it’s more like paying for a metered parking spot
The internet is being rebuilt for machines | TechCrunchThe key technical change with this new generation is that it decouples compute from storage, allowing compute to scale up in seconds to accommodate agent traffic bursts and to scale down to zero, so customers pay $0 when agents are idle.
The internet is being rebuilt for machines | TechCrunchand enterprise needs search that keeps up without paying for empty or idle compute.”
The internet is being rebuilt for machines | TechCrunchThat’s where AWS’s new OpenSearch Serverless comes in.
The internet is being rebuilt for machines | TechCrunchBut the buck doesn’t stop at consumer-focused AI agents. Enterprises are increasingly deploying agents internally and for their customers, creating new kinds of machine-generated traffic behind the scenes. As a result, cloud providers and infrastructure companies have been
The internet is being rebuilt for machines | TechCrunch“Non-human traffic will exceed human traffic sometime in the first half of 2027,” said Lai Yi Ohlsen, senior product manager at Cloudflare, to TechCrunch.
The internet is being rebuilt for machines | TechCrunchCloudflare says bots accounted for 31% of overall HTTP traffic over the last six months.
The internet is being rebuilt for machines | TechCrunchThe launch reflects a growing realization across the tech industry
The internet is being rebuilt for machines | TechCrunchscale back down to zero
The internet is being rebuilt for machines | TechCrunch— essentially a system for storing and retrieving information at scale —
The internet is being rebuilt for machines | TechCruncharound humans who search, click, scroll, and stream in a steady and predictable fashion. AI agents behave differently
The internet is being rebuilt for machines | TechCrunchWhenever you have a story which tells of how something has happened which affects both people and property, always put the people first
What is news? How journalists assess news valueIt is when reporting these stories that it is important to make sure that the story is centred on people
What is news? How journalists assess news valueIf an archbishop says it, it is news, because what he says on the subject is significant
What is news? How journalists assess news valueit becomes a matter of concern to them
What is news? How journalists assess news valueEvents which happened days or even weeks earlier can still be news, as long as they have not been reported before
What is news? How journalists assess news valueThese elements make up what we call the “news value” of information. The stronger the elements are, the higher the news value
What is news? How journalists assess news valueThe way in which the news is judged, though, is the same everywhere.
What is news? How journalists assess news valueThe same event can have different levels of interest in different societies, and will be talked about in different ways
What is news? How journalists assess news valuehelp the public make informed decisions
What is news? How journalists assess news valuenew, unusual, significant, and people-centred
What is news? How journalists assess news value“Technology that should liberate human labor is instead trapping delivery workers in longer working hours without significantly improving their income,” the magazine wrote.
Xi-Trump summit showcases Beijing, China's futuristic capital - The Washington Postthe U.S. Chamber of Commerce warned that China’s industrial strategy was becoming “more systemic and pervasive, extending across all layers of production from upstream inputs and industrial equipment to downstream applications, services, and frontier technologies.”
Xi-Trump summit showcases Beijing, China's futuristic capital - The Washington Post“We have achieved a mutually assured deterrence of some kind.”
Xi-Trump summit showcases Beijing, China's futuristic capital - The Washington Postand a threat, according to U.S. lawmakers, to U.S. national security.
Xi-Trump summit showcases Beijing, China's futuristic capital - The Washington Postcarefully arranged set pieces will conceal the stubborn challenges dogging China’s economy, such as youth unemployment, as well as the country’s lingering reliance on certain U.S. technology like advanced computer chips
Xi-Trump summit showcases Beijing, China's futuristic capital - The Washington PostGovernments and corporations should deploy different kinds of AI systems simultaneously rather than relying on a single one
AI Deterrence by Betrayalstrengthen security around model weights
AI Deterrence by BetrayalTo deter rivals from high-stakes AI deployment, states may work to improve the stealth of poisoned data and triggers, the potency of backdoor payloads, and to develop secret loyalty attacks
AI Deterrence by Betrayalbetrayal may be more covert and less escalatory; actors might thus attempt betrayal attacks before other forms of sabotage.
AI Deterrence by Betrayaldeterrence by betrayal mainly disincentivizes high-stakes AI deployment
AI Deterrence by Betrayalthere is no reliable method to detect backdoors
AI Deterrence by BetrayalDuring automated research processes too rapid for humans to meaningfully oversee, compromised AI systems may propagate their disloyalty to the successor systems they develop.
AI Deterrence by BetrayalAI corporations may be subject to surveillance and eventual co-option by the state as a result. Meanwhile, the public may react harshly to automation and economic disempowerment, placing additional pressure on governments to co-opt the AI industry and creating new hazards for AI developers in the form of rudimentary attempts at backdooring and data poisoning.
AI Deterrence by Betrayalhave the ability to embed event-driven backdoors that would cause AIs to betray legal chains of command
AI Deterrence by Betrayalmiddle powers may become increasingly dependent on superpowers for protection and economic support, while recognizing that their goodwill can be unreliable
AI Deterrence by Betrayala long tail of smaller ones; some are generalists, and others focus on a specific industry or thesis
The third wave of American philanthropy - by Nan Ransohoff