Serena Ge
11 followers · 18 following · 1328 views
on the atlas — 68
- Gmail110 savers
- Home / X20 savers
- Edward Tufte’s 6 Data Visualization Principles | by Yahia zakaria | Medium1 savers
- International Consortium Builds ‘Google Map’ of Human Metabolism1 savers
- TABLE OF CONTENTS — Almanack of Naval Ravikant3 savers
- Dario Amodei — Machines of Loving Grace13 savers
- World's fair2 savers
- Productivity paradox3 savers
- What will be scarce? - by Alex Imas - Ghosts of Electricity15 savers
- [INT] Anthropic call agenda Thurs April 16 - Google Docs1 savers
- Measuring AI Ability to Complete Long Tasks - METR7 savers
- Scaling long-running autonomous coding · Cursor7 savers
- Transfer ownership of a team – Figma Learn - Help Center1 savers
- YYZ to SFO Porter Tips1 savers
- Why AI Will Save The World - Marc Andreessen Substack1 savers
- What's going on here, with this human? - Graham Duncan Blog49 savers
- Curius / Bookmarks for the extremely curious184 savers
- Graph Enabled Llama Index - siwei.io1 savers
- Ben Eater2 savers
- Welcome – Cloud TPUs – Google Cloud console1 savers
- Welcome – Cloud TPUs – Google Cloud console1 savers
- Quickstart: Run a calculation on a Cloud TPU VM by using TensorFlow | Google Cloud1 savers
- Deep Learning Online Courses | NVIDIA1 savers
- Installation — Dive into Deep Learning 1.0.0-beta0 documentation1 savers
- Paradigms of Parallelism | Colossal-AI1 savers
- Parallelism in Distributed Deep Learning · Better Tomorrow with Computer Science1 savers
- Chatbot UI2 savers
- GPUs on Google Colab with RAPIDS cuDF | Google Colab1 savers
- Introduction to Cloud TPU | Google Cloud1 savers
- Excalidraw | Hand-drawn look & feel • Collaborative • Secure20 savers
- PyTorch documentation — PyTorch 2.0 documentation2 savers
- 6.S898 Deep Learning, Fall 20221 savers
- Learning complex goals with iterated amplification2 savers
- Undergraduate Studies Calendar | University of Waterloo1 savers
- How to build a self-driving car in one month | by Max Deutsch | Medium2 savers
- Curius / Onboarding2667 savers
- sparkly people and how to find them - by Anson Yu54 savers
- Godly — The Best Web Design Inspiration36 savers
- Zoom In: An Introduction to Circuits22 savers
- How to Work Hard22 savers
- How to Think for Yourself20 savers
- GPT-417 savers
- How to Beat Procrastination - LessWrong11 savers
- Anthropic10 savers
- Investigations ← Forensic Architecture10 savers
- Future ML Systems Will Be Qualitatively Different10 savers
- More Is Different for AI9 savers
- Essays8 savers
- The Bus Ticket Theory of Genius8 savers
- Feature Visualization7 savers
- Four Background Claims - Machine Intelligence Research Institute7 savers
- Szymon Kaliski7 savers
- ChatGPT7 savers
- Visualizing the deep learning revolution | by Richard Ngo | Jan, 2023 | Medium7 savers
- Xerox Parc’s Engineers on How They Invented the Future—and How Xerox Lost It - IEEE Spectrum6 savers
- How People Get Rich Now6 savers
- Anthropic | Core Views on AI Safety: When, Why, What, and How6 savers
- Twitter's Recommendation Algorithm5 savers
- There's An AI For That | AI Database5 savers
- How to Be an Expert in a Changing World5 savers
- Urban Camping4 savers
- Stanford Alpaca Model Release4 savers
- The Atlas of Economic Complexity4 savers
- AGI safety from first principles.pdf - Google Drive4 savers
- nayafia/lemonade-stand: A handy guide to financial support for open source4 savers
- Racing through a minefield: the AI deployment problem3 savers
- Deconstructing dualisms to understand the world3 savers
- All Posts - LessWrong2 savers
highlights — 158
but it made the human behind any specific product invisible, and ultimately, replaceable.
What will be scarce? - by Alex Imas - Ghosts of Electricityalienation: severing workers from the product of their labor
What will be scarce? - by Alex Imas - Ghosts of Electricityexploitation: the ability to pay workers less than the value of what they produce
What will be scarce? - by Alex Imas - Ghosts of ElectricityBefore industrialization, it was difficult to separate a product from the person who made it.
What will be scarce? - by Alex Imas - Ghosts of ElectricityAfter trying to streamline the store experience with fewer workers and more automation, the company concluded that this had been a mistake. CEO Brian Niccol said that ``handwritten notes on cups’’, ceramic cups, and ``the return of great seats’’ had led more customers to ``sit and stay in our cafes’’, showing that ``small details and hospitality drive satisfaction.’’ More baristas are being hired per store and automation is being rolled back.
What will be scarce? - by Alex Imas - Ghosts of ElectricityIf advanced AI brings material abundance—if machines can produce many if not all forms of human production at very low marginal cost—does economics become irrelevant? No, we will still have scarcity, but the kind of scarcity that matters will change
What will be scarce? - by Alex Imas - Ghosts of ElectricityCurrent frontier AIs are vastly better than humans at text prediction and knowledge tasks
Measuring AI Ability to Complete Long Tasks - METRAt the end of each cycle
Scaling long-running autonomous coding · CursorPlanners and workers
Scaling long-running autonomous coding · CursorNo agent took responsibility for hard problems or end-to-end implementation.
Scaling long-running autonomous coding · CursorThe system was brittle: agents could fail while holding locks, try to acquire locks they already held, or update the coordination file without acquiring the lock at all.
Scaling long-running autonomous coding · Cursorplanning ahead would be too rigid. The path through a large project is ambiguous, and the right division of work isn't obvious at the start. We began with dynamic coordination, where agents decide what to do based on what others are currently doing
Scaling long-running autonomous coding · Cursorrunning hundreds of concurrent agents on a single project, coordinating their work, and watching them write over a million lines of code and trillions of tokens.
Scaling long-running autonomous coding · CursorAI is quite possibly the most important – and best – thing our civilization has ever created, certainly on par with electricity and microchips, and probably beyond those.
Why AI Will Save The World - Marc Andreessen Substacknatural intelligence today can be done much better with AI
Why AI Will Save The World - Marc Andreessen Substackcreation of new jobs, and wage growth, and resulting in a new era of heightened material prosperity across the planet.
Why AI Will Save The World - Marc Andreessen Substackhuman intelligence makes a very broad range of life outcomes better.
Why AI Will Save The World - Marc Andreessen Substacksimilar to how people do it
Why AI Will Save The World - Marc Andreessen Substackin life, the challenge is not so much to figure out how best to play the game; the challenge is to figure out what game you’re playing.”
What's going on here, with this human? - Graham Duncan BlogWe (humanity) develop a reasonable set of tests for whether an AI system might be dangerous.
Racing through a minefield: the AI deployment problemInformation about what cutting-edge AI systems can do - especially if it is powerful and impressive - could spur incautious actors to race harder toward developing powerful AI of their own (or give them an idea of how to build powerful systems, by giving them an idea of what sorts of abilities to aim for).
Racing through a minefield: the AI deployment problemSelective information sharing (so the incautious don’t catch up). Sharing some information widely (e.g., technical insights about how to reduce misalignment risk), some selectively (e.g., demonstrations of how powerful and dangerous AI systems might be), and some not at all (e.g., the specific code that, if accessed by a hacker, would allow the hacker to deploy potentially dangerous AI systems themselves).
Racing through a minefield: the AI deployment problemcautious actors need to move fast enough that they can’t be overpowered by others’ AI systems, but slowly enough that they don’t cause disaster themselves.
Racing through a minefield: the AI deployment problemA high-stakes race (for advanced AI) can dramatically worsen outcomes by making all parties more willing to cut corners in safety.
Allan Dafoe - AI Governance: Opportunity and Theory of ImpactAt the limit, AI could catalyze (global) robust totalitarianism.
Allan Dafoe - AI Governance: Opportunity and Theory of ImpactFrom this perspective, the priority is to focus on those groups who are most likely to incubate superintelligence, and help them to have the best culture, organization, safety expertise, insights, and infrastructure for the process to go well.
Allan Dafoe - AI Governance: Opportunity and Theory of Impactbecause of a host of structural dynamics
Allan Dafoe - AI Governance: Opportunity and Theory of ImpactThese systems, individually or in collaboration with humans, could give rise to cognitive capabilities in strategically important tasks that exceed what humans are otherwise capable of.
Allan Dafoe - AI Governance: Opportunity and Theory of ImpactGiven how important intelligence is---to the solving of our global problems, to the production and allocation of wealth, and to military power---this perspective makes clear that superintelligent AI would pose profound opportunities and risks.
Allan Dafoe - AI Governance: Opportunity and Theory of Impactfor many kinds of surveillance; by authoritarian governments to shape online discourse; for autonomous weapons systems; for cyber tools and autonomous cyber capabilities; to aid and make consequential decisions such as for employment, loans, and criminal sentencing; in advertising; in education and testing; in self-driving cars and navigation; in social media.
Allan Dafoe - AI Governance: Opportunity and Theory of ImpactThey seem to be one of several heuristics for detecting the boundaries of objects
Zoom In: An Introduction to CircuitsEarly layers contain features like edge or curve detectors, while later layers have features like floppy ear detectors or wheel detectors
Zoom In: An Introduction to Circuitsspend thousands of hours tracing through every neuron and its connections? What kind of picture of neural networks would emerge?
Zoom In: An Introduction to CircuitsWhat if we treated individual neurons, even individual weights, as being worthy of serious investigation?
Zoom In: An Introduction to Circuitssmall research community realizing they can now study their topic in a finer grained level of detail.
Zoom In: An Introduction to Circuitsdiscovery tool is a “knowledge graph,” which consists of nodes and edges representing real-world objects and the relationships between them, presenting your data ecosystem as a visualization.
Data Search and Discovery: What it Is, and Why It Matters | data.worldthere are so many possible inputs that can cause a model to generate harmful text. As a result, it’s hard to find all of the cases where a model fails before it is deployed in the real world.
Red Teaming Language Models with Language Modelsbuild assistants that can be trusted to take on all of the cognitive labor needed for evaluation, so humans can focus on communicating their preferences.
AI-written critiques help humans notice flawswe found that models are better at discriminating than at critiquing their own answers, indicating they know about some problems that they can’t or don’t articulate. Furthermore, the gap between discrimination and critique ability did not appear to decrease for larger models. Reducing this gap is an important priority for our alignment research.
AI-written critiques help humans notice flawstrain AI assistants that help humans provide feedback on hard tasks
AI-written critiques help humans notice flawsModels may then learn to give outputs that look good to humans but have errors we systematically fail to notice.
AI-written critiques help humans notice flawsEmbeddings are a numerical representation of text that can be used to measure the relateness between two pieces of text.
Models - OpenAI APIin more complex reasoning situations, GPT-4 is much more capable than any of our previous models.
Models - OpenAI APIAmerican Psycho gave me chills. LH says that the modern business card scene is just people comparing their personal websites. It’s funny because it’s a little true.
sweet summer child - by Anson YuThis is the thesis that ultimately underlies every one of my interests; well designed cities, software that feels like magic, and media that inspires optimism.
sweet summer child - by Anson YuI want to help people feel like they can dream → I want to help people fight the pessimism for life that hardens into us as we age.
sweet summer child - by Anson Yulogs, updates, and short essays
sweet summer child - by Anson Yufocus pays dividends that brute force can only dream of.
sweet summer child - by Anson YuI’ve never been hyped up like that in my life
sweet summer child - by Anson YuChain of thought prompting is a simple and broadly applicable method for improving the ability of language models to perform various reasoning tasks.
Language Models Perform Reasoning via Chain of Thought – Google AI Blog