Frank’s Ramblings
Vision Transformers are Overrated
Vision transformers (ViTs) have seen an incredible rise in the past four years. They have an obvious...
a year ago
Vision transformers (ViTs) have seen an incredible rise in the past four years. They have an obvious upside: in a visual recognition setting, the receptive field of a pure ViT is effectively the entire image 1. In particular, vanilla ViTs maintain the quadratic time complexity...
The Berkeley...
How to Evaluate Jailbreak Methods: A Case Study with the StrongREJECT Benchmark
When we began studying jailbreak evaluations, we found a fascinating paper claiming that you could...
4 months ago
When we began studying jailbreak evaluations, we found a fascinating paper claiming that you could jailbreak frontier LLMs simply by translating forbidden prompts into obscure languages. Excited by this result, we attempted to reproduce it and found something unexpected.
The...
AI Snake Oil
OpenAI’s policies hinder reproducible research on language models
LLMs have become privately-controlled research infrastructure
a year ago
LLMs have become privately-controlled research infrastructure
AI Snake Oil
What the executive order means for openness in AI
Good news on paper, but the devil is in the details
a year ago
Good news on paper, but the devil is in the details
Artificial Ignorance
What President Biden's AI executive order actually means
I read all 111 pages so you don't have to.
a year ago
I read all 111 pages so you don't have to.
Strange Loop Canon
LLMs have special intelligence, not general, and that's plenty.
“Life can only be understood backwards, but it must be lived forwards.”
9 months ago
“Life can only be understood backwards, but it must be lived forwards.”
Artificial Ignorance
AI Roundup 043: Happy birthday, ChatGPT
December 1, 2023.
a year ago
Matt Mazur
Is the ChatGPT API Refusing to Summarize Academic Papers? Not so fast.
Yesterday on X, I shared a post about some responses I was getting from the ChatGPT 3.5 API...
a year ago
Yesterday on X, I shared a post about some responses I was getting from the ChatGPT 3.5 API indicating that it was refusing to summarize arXiv papers: There has been a lot of discussion recently about the perceived decrease in the quality of ChatGPT’s responses and seeing...
Weighty Thoughts
When the AI Bubble Bursts
It’s when, not if, for these kinds of new technologies
6 months ago
It’s when, not if, for these kinds of new technologies
IEEE Spectrum
Video Friday: Multiple MagicBots
Video Friday is your weekly selection of awesome robotics videos, collected by your friends at IEEE...
4 weeks ago
Video Friday is your weekly selection of awesome robotics videos, collected by your friends at IEEE Spectrum robotics. We also post a weekly calendar of upcoming robotics events for the next few months. Please send us your events for inclusion.
Humanoids Summit: 11–12 December...
One Useful Thing
Confronting Impossible Futures
We shouldn't be certain about what is next, but we should plan for it
5 months ago
We shouldn't be certain about what is next, but we should plan for it
Sam Altman
Hard Startups
The most counterintuitive secret about startups is that it’s often easier to succeed with a hard...
over a year ago
The most counterintuitive secret about startups is that it’s often easier to succeed with a hard startup than an easy one. A hard startup requires a lot more money, time, coordination, or technological development than most startups. A good hard startup is one that will be...
Society's Backend
OpenAI's Strawberry May Enhance AI Reasoning, Optimization of Vision Language Models, Info on...
Weekly updates and resources 7/15/24
5 months ago
Weekly updates and resources 7/15/24
Artificial Ignorance
AI Roundup 036: The money pit
October 13, 2023.
a year ago
AI Snake Oil
AI existential risk probabilities are too unreliable to inform policy
How speculation gets laundered through pseudo-quantification
5 months ago
How speculation gets laundered through pseudo-quantification
Sam Altman
What I Heard From Trump Supporters
After the election, I decided to talk to 100 Trump voters from around
the country. I went to the...
over a year ago
After the election, I decided to talk to 100 Trump voters from around
the country. I went to the middle of the
country, the middle of the state, and talked to many online.
This was a surprisingly interesting and helpful experience—I highly
recommend it. With three exceptions,...
Sam Altman
What I Wish Someone Had Told Me
Optimism, obsession, self-belief, raw horsepower and personal connections are how things get...
a year ago
Optimism, obsession, self-belief, raw horsepower and personal connections are how things get started.
Cohesive teams, the right combination of calmness and urgency, and unreasonable commitment are how things get finished. Long-term orientation is in short supply; try not to worry...
fast.ai
Deep Learning Foundations Signup, Open Source Scholarships, & More
Signups are now open for Practical Deep Learning for Coders Part 2, 2022. Scholarships are available...
over a year ago
Signups are now open for Practical Deep Learning for Coders Part 2, 2022. Scholarships are available for fast.ai community contributors, open source developers, and diversity scholars.
Weighty Thoughts
I'm writing a book!
Why I've been gone over a month, and how you can help
a month ago
Why I've been gone over a month, and how you can help
Society's Backend
An AI Revolution Isn't Coming - We're Already In It
A small showcase of companies using AI in ways that will change your life
a year ago
A small showcase of companies using AI in ways that will change your life
Society's Backend
Backend Biweekly #1: Important AI Developments and ML Learning Resources
Updates on Apple, Mistral, and Microsoft and resources to build the GPT tokenizer, use MLX to train...
10 months ago
Updates on Apple, Mistral, and Microsoft and resources to build the GPT tokenizer, use MLX to train models, and benchmark LLMs
Sam Altman
US Digital Currency
I am pretty sure cryptocurrency is here to stay in some form (at least as a store of value, which is...
over a year ago
I am pretty sure cryptocurrency is here to stay in some form (at least as a store of value, which is the only use case we have seen work at scale so far). There was possibly a time when governments could have totally stopped it, but it feels like that’s in the rearview mirror.
...
Artificial Ignorance
AI Roundup 032: The Insight Forum
September 15, 2023.
a year ago
Artificial Ignorance
Jevons Paradox and the future of programming
Supply and demand in the Age of AI.
9 months ago
Supply and demand in the Age of AI.
Society's Backend
Weekly Backend #5: 55 Resources
The inner workings of transformers, Machine Learning Q and AI, machine learning classifies harmful...
8 months ago
The inner workings of transformers, Machine Learning Q and AI, machine learning classifies harmful viruses, and more
Artificial Ignorance
AI Roundup 040: GPTs
November 10, 2023.
a year ago
Society's Backend
Mastering the Art of Documentation
Documentation is really just glorified dog sitting
a year ago
Documentation is really just glorified dog sitting
Society's Backend
MASSIVE New Model Releases from Google, OpenAI, and Meta Change the AI Landscape
Society's Backend Reading List 12-16-2024
2 weeks ago
Society's Backend Reading List 12-16-2024
Artificial Ignorance
How to fine-tune ChatGPT
No GPU cluster required.
a year ago
One Useful Thing
A guide to prompting AI (for what it is worth)
A little bit of magic, but mostly just practice
a year ago
A little bit of magic, but mostly just practice
Society's Backend
Taking the AI to Consumers
The Google Pixel 8 Pro is a much bigger deal than you think
a year ago
The Google Pixel 8 Pro is a much bigger deal than you think
Marcus on AI
Generative AI’s Continuing Copyright Problems, an Essay in Memory of Suchir Balaji, 1998 - 2024
In early November, I had a stimulating Zoom call with a former OpenAI employee and Berkeley graduate...
3 weeks ago
In early November, I had a stimulating Zoom call with a former OpenAI employee and Berkeley graduate named Suchir Balaji, who had just left OpenAI.
Weighty Thoughts
Let’s Talk About AI Compute
What really matters and why AI computing is exactly the same as all other large-scale computing
a year ago
What really matters and why AI computing is exactly the same as all other large-scale computing
Made by Ollin
Maple Diffusion
I ported Stable Diffusion to my phone
over a year ago
I ported Stable Diffusion to my phone
The Berkeley...
Designing Societally Beneficial Reinforcement Learning Systems
Deep reinforcement learning (DRL) is transitioning from a research field focused on game playing to...
over a year ago
Deep reinforcement learning (DRL) is transitioning from a research field focused on game playing to a technology with real-world applications. Notable examples include DeepMind’s work on controlling a nuclear reactor or on improving Youtube video compression, or Tesla attempting...
Andrej Karpathy blog
Deep Reinforcement Learning: Pong from Pixels
-->
This is a long overdue blog post on Reinforcement Learning (RL). RL is hot! You may have...
over a year ago
-->
This is a long overdue blog post on Reinforcement Learning (RL). RL is hot! You may have noticed that computers can now automatically learn to play ATARI games (from raw game pixels!), they are beating world champions at Go, simulated quadrupeds are learning to run and leap,...
AI Snake Oil
FAQ about the book and our writing process
What's in the book and how we wrote it
3 months ago
What's in the book and how we wrote it
The Berkeley...
Why do Policy Gradient Methods work so well in Cooperative MARL? Evidence from Policy Representation
In cooperative multi-agent reinforcement learning (MARL), due to its on-policy nature, policy...
over a year ago
In cooperative multi-agent reinforcement learning (MARL), due to its on-policy nature, policy gradient (PG) methods are typically believed to be less sample efficient than value decomposition (VD) methods, which are off-policy. However, some recent empirical studies demonstrate...
Matt Mazur
It’s Time to Build
It’s been a few months so I wanted to say hey to the 7 of you who follow this blog and share a few...
8 months ago
It’s been a few months so I wanted to say hey to the 7 of you who follow this blog and share a few updates about what I’ve been up to. Quick recap At the start of 2023 I quit consulting to go full time on Preceden, my SaaS timeline maker, after growing it on … Continue reading...
Daniel Miessler
Scott Kuffer of Nucleus Security | SPONSORED INTERVIEW SERIES
In this standalone episode we’re doing a sponsored interview with Scott Kuffer, co-founder and COO...
over a year ago
In this standalone episode we’re doing a sponsored interview with Scott Kuffer, co-founder and COO of Nucleus Security. I was already excited by this vendor just based on the research I did to allow them to be a sponsor, but the conversation with them really made me think they’re...
One Useful Thing
Gradually, then Suddenly: Upon the Threshold
Small improvements can lead to big changes
6 months ago
Small improvements can lead to big changes
Marcus on AI
Could 2025 see the largest cyberattack in history?
In a just-published series of very brief essays called “The Incredible, World-Altering ‘Black Swan’...
2 days ago
In a just-published series of very brief essays called “The Incredible, World-Altering ‘Black Swan’ Events That Could Upend Life in 2025”, Politico asked “15 futurists, foreign policy analysts and other prognosticators”, including me, “to provide some explosive potential...
Strange Loop Canon
Progress
when benefits are worth tallying the human cost of technological leaps
a year ago
when benefits are worth tallying the human cost of technological leaps
The Berkeley...
Koala: A Dialogue Model for Academic Research
In this post, we introduce Koala, a chatbot trained by fine-tuning Meta’s LLaMA on dialogue data...
a year ago
In this post, we introduce Koala, a chatbot trained by fine-tuning Meta’s LLaMA on dialogue data gathered from the web. We describe the dataset curation and training process of our model, and also present the results of a user study that compares our model to ChatGPT and...
Daniel Miessler
News, Analysis, and Discovery | NO. 356
SECURITY NEWS TikTok has now admitted, after denying last week, that Chinese staff can in fact read...
over a year ago
SECURITY NEWS TikTok has now admitted, after denying last week, that Chinese staff can in fact read European TikTok data. Pressure is increasing across the US government to outright ban the app, but it's quickly becoming national infrastructure so many young people. MORE |...
Artificial Ignorance
AI Roundup 073: Music make you lose control
June 28, 2024.
6 months ago
Artificial Ignorance
How Shopify is making AI Magic
A case study on turning competitive advantages into useful AI.
a year ago
A case study on turning competitive advantages into useful AI.
Marcus on AI
AI influencer hype gone wild
Reality is quite the opposite
2 months ago
Reality is quite the opposite
Artificial Ignorance
AI's massive cash needs are Big Tech's chance to own the future
Over the past year, AI startups have raised some impressive amounts of money. OpenAI raised $10...
11 months ago
Over the past year, AI startups have raised some impressive amounts of money. OpenAI raised $10 billion, Anthropic did $6 billion, Inflection AI raised $1.3 billion, and dozens of companies closed rounds in the hundreds of millions.
Society's Backend
Meta's New Segmentation Model, A New Open-Source Image Generation Model, Apple Intelligence Model...
Machine learning resources and updates 8/5/2024
5 months ago
Machine learning resources and updates 8/5/2024
Artificial Ignorance
Tutorial: How to build an AI agent (Part 2)
Part 2: Giving our agent internet access.
a year ago
Part 2: Giving our agent internet access.
Society's Backend
What it's Like to Work in AI and Advice from 10 AI Professionals
I asked 10 AI professionals 4 questions about their work in AI and the qualifications required for...
a month ago
I asked 10 AI professionals 4 questions about their work in AI and the qualifications required for their job
Society's Backend
Why Software Engineers Need to Understand Machine Learning
And how ML helps software engineers in their daily work
7 months ago
And how ML helps software engineers in their daily work
Society's Backend
All Machine Learning Resources and Updates 06/21/2024
Below are the machine learning resources and updates from the past few days you don't want to miss....
6 months ago
Below are the machine learning resources and updates from the past few days you don't want to miss. If you want all the ML updates from X, follow me there. Mamba + Sliding Window Attention = SAMBA with Efficient Unlimited... China runs to be one of top global players in AI model...
One Useful Thing
What OpenAI did
A new model opens up new possibilities
7 months ago
A new model opens up new possibilities
Society's Backend
AI and Software Reading List 4: State of the Job Market, Apple's Private Cloud Compute Released for...
Society's Backend Reading List 10-28-2024
2 months ago
Society's Backend Reading List 10-28-2024
Society's Backend
One Year of Society's Backend
The lessons I've learned along the way
4 months ago
The lessons I've learned along the way
Artificial Ignorance
I fell for a deepfake Elon Musk’s cryptocurrency scam bait
A lesson in deepfakes, Ponzi schemes, and YouTube algorithm manipulation.
7 months ago
A lesson in deepfakes, Ponzi schemes, and YouTube algorithm manipulation.
Artificial Ignorance
AI Roundup 090: Computer Use
October 25, 2024.
2 months ago
Strange Loop Canon
When we become cogs
on motivation and technology
a month ago
on motivation and technology
One Useful Thing
Assigning AI: Seven Ways of Using AI in Class
Also prompts! And things to watch out for!
a year ago
Also prompts! And things to watch out for!
IEEE Spectrum
Video Friday: Unitree Talent Awakening
Video Friday is your weekly selection of awesome robotics videos, collected by your friends at IEEE...
a week ago
Video Friday is your weekly selection of awesome robotics videos, collected by your friends at IEEE Spectrum robotics. We also post a weekly calendar of upcoming robotics events for the next few months. Please send us your events for inclusion.
ICRA 2025: 19–23 May 2025, ATLANTA,...
Society's Backend
Side Projects to Get a Job in ML, a Survey of Small Language Models, How to Build ML Pipelines, and...
Society's Backend Reading List 11-11-2024
a month ago
Society's Backend Reading List 11-11-2024
Artificial Ignorance
AI Roundup 041: Lyria
November 17, 2023.
a year ago
The Gradient
What Do LLMs Know About Linguistics? It Depends on How You Ask
On the phenomenon of LLM sensitivity to prompting choices through two core linguistic tasks and...
a year ago
On the phenomenon of LLM sensitivity to prompting choices through two core linguistic tasks and categorize how specific prompting choices can affect the model's behavior.
One Useful Thing
Thinking companion, companion for thinking
Some simple ways to use AI to break you out of biases
a year ago
Some simple ways to use AI to break you out of biases
Artificial Ignorance
Milestones, motivations, and miscellany
Some thoughts on the future of this newsletter.
6 months ago
Some thoughts on the future of this newsletter.
fast.ai
I was an AI researcher. Now, I am an immunology student.
Last year, I became captivated by a new topic in a way that I hadn’t felt since I first discovered...
a year ago
Last year, I became captivated by a new topic in a way that I hadn’t felt since I first discovered machine learning
Artificial Ignorance
Small newsletters, big ideas
A subscriber showcase full of hidden gems on Substack.
2 weeks ago
A subscriber showcase full of hidden gems on Substack.
One Useful Thing
Doing Stuff with AI: Opinionated Midyear Edition
AI systems have gotten more capable and easier to use
7 months ago
AI systems have gotten more capable and easier to use
IEEE Spectrum
Boston Dynamics and Toyota Research Team Up on Robots
Today, Boston Dynamics and the Toyota Research Institute (TRI) announced a new partnership “to...
2 months ago
Today, Boston Dynamics and the Toyota Research Institute (TRI) announced a new partnership “to accelerate the development of general-purpose humanoid robots utilizing TRI’s Large Behavior Models and Boston Dynamics’ Atlas robot.” Committing to working towards a general purpose...
Sam Altman
E Pur Si Muove
Earlier
this year, I noticed something in China that really surprised me. I
realized I felt more...
over a year ago
Earlier
this year, I noticed something in China that really surprised me. I
realized I felt more comfortable discussing controversial ideas in Beijing than
in San Francisco. I didn’t feel completely comfortable—this was China,
after all—just more comfortable than at home.
That...
Weighty Thoughts
Scaling is a Choice
Progress in tech is rarely “inevitable.” Looking at semiconductors and AI.
2 weeks ago
Progress in tech is rarely “inevitable.” Looking at semiconductors and AI.
Artificial Ignorance
Looking a gift llama in the mouth
How Llama 3.1 uniquely leverages Meta's business model (and why we should be a little bit cynical...
5 months ago
How Llama 3.1 uniquely leverages Meta's business model (and why we should be a little bit cynical about it)
Weighty Thoughts
Let's talk about AI power costs
It's important, but a lot of recent attention has been concern-trolling
5 months ago
It's important, but a lot of recent attention has been concern-trolling
IEEE Spectrum
SwitchBot S10 Review: “This Is the Future of Home Robots”
I’ve been reviewing robot vacuums for more than a decade, and robot mops for just as long. It’s been...
3 months ago
I’ve been reviewing robot vacuums for more than a decade, and robot mops for just as long. It’s been astonishing how the technology has evolved, from the original iRobot Roomba bouncing off of walls and furniture to robots that use lidar and vision to map your entire house and...
fast.ai
Can LLMs learn from a single example?
We’ve noticed an unusual training pattern in fine-tuning LLMs. At first we thought it’s a bug, but...
a year ago
We’ve noticed an unusual training pattern in fine-tuning LLMs. At first we thought it’s a bug, but now we think it shows LLMs can learn effectively from a single example.
AI Snake Oil
The bait and switch behind AI risk prediction tools
Toronto recently used an AI tool to predict when a public beach will be safe. It went horribly awry....
over a year ago
Toronto recently used an AI tool to predict when a public beach will be safe. It went horribly awry. The developer claimed the tool achieved over 90% accuracy in predicting when beaches would be safe to swim in. But the tool did much worse: on a majority of the days when the...
AI Snake Oil
Model alignment protects against accidental harms, not intentional ones
The hand wringing about failures of model alignment is misguided
a year ago
The hand wringing about failures of model alignment is misguided
Matt Mazur
Introducing Preceden’s new AI-Powered Timeline Generator
For the past few months I’ve been heads down building an AI-powered timeline generator tool for...
a year ago
For the past few months I’ve been heads down building an AI-powered timeline generator tool for Preceden, my SaaS timeline maker software: The tool – which is free to use and available on Preceden’s homepage – lets you type in a topic or detailed description of a timeline and it...
AI Snake Oil
AI safety is not a model property
Trying to make an AI model that can’t be misused is like trying to make a computer that can’t be...
9 months ago
Trying to make an AI model that can’t be misused is like trying to make a computer that can’t be used for bad things
Weighty Thoughts
One Definite Sign of a Bad Startup Idea
If you can’t tell anyone about your startup idea, it's not a good idea
over a year ago
If you can’t tell anyone about your startup idea, it's not a good idea
Artificial Ignorance
10 of the most impactful AI stories of 2023
A quick look back on a very busy year in AI.
a year ago
A quick look back on a very busy year in AI.
Artificial Ignorance
How Intercom is transforming customer support with AI
A conversation with Fergal Reid, VP of AI at Intercom.
8 months ago
A conversation with Fergal Reid, VP of AI at Intercom.
Matt Mazur
Running Mistral 7B Instruct on a Macbook
Similar to yesterday’s post on running Mistral 8x7Bs Mixture of Experts (MOE) model, I wanted to...
a year ago
Similar to yesterday’s post on running Mistral 8x7Bs Mixture of Experts (MOE) model, I wanted to document the steps I took to run Mistral’s 7B-Instruct-v0.2 model on a Mac for anyone else interested in playing around with it. Unlike yesterday’s post though, this 7B Instruct...
Strange Loop Canon
Why AI hasn’t shown up in the GDP statistics yet
AI is meant to bring us closer to utopia, according to its builders.
5 months ago
AI is meant to bring us closer to utopia, according to its builders.
One Useful Thing
On giving AI eyes and ears
AI can listen and see, with bigger implications than we might realize.
a year ago
AI can listen and see, with bigger implications than we might realize.
IEEE Spectrum
Video Friday: Swiss-Mile Robot vs. Humans
Video Friday is your weekly selection of awesome robotics videos, collected by your friends at IEEE...
2 months ago
Video Friday is your weekly selection of awesome robotics videos, collected by your friends at IEEE Spectrum robotics. We also post a weekly calendar of upcoming robotics events for the next few months. Please send us your events for inclusion.
Humanoids 2024: 22–24 November...
Artificial Ignorance
AI Roundup 094: Pixtral et Le Chat
November 22, 2024.
a month ago
IEEE Spectrum
Drones with Legs Can Walk, Hop, and Jump into the Air
“Whenever I encountered crows on the EPFL campus, I would observe how they walked, hopped over or...
a month ago
“Whenever I encountered crows on the EPFL campus, I would observe how they walked, hopped over or jumped on obstacles, and jumped for take-offs,” says Won Dong Shin, a doctoral student at EPFL’s Laboratory of Intelligent Systems. “What I consistently observed was that they always...
Rozado’s Visual...
Mentions of Prejudice in Academic Papers: A Declining Trend Amidst Ongoing DEI Growth?
Prejudice-denoting terms in academic research have recently decreased while some DEI-related terms...
3 months ago
Prejudice-denoting terms in academic research have recently decreased while some DEI-related terms continue to rise—what does this shift reveal?
Daniel Miessler
Unsupervised Learning NO. 364 | Reality Headset, BingPT, AI+Cyber
If you're not subscribed to the podcast version of the newsletter, please add it with your favorite...
a year ago
If you're not subscribed to the podcast version of the newsletter, please add it with your favorite client. APPLE | SPOTIFY | OTHER SECURITY NEWS The FBI is warning people to block online ads due to imposters poisoning search results. They advise users to 1) check ad URLs, 2) go...
Marcus on AI
Update re: Microsoft and training data
Relevant to the post I sent earlier today,...
a month ago
Relevant to the post I sent earlier today, https://www.howtogeek.com/is-microsoft-using-your-word-documents-to-train-ai/says that an unnamed spokesperson at Microsoft claims that “Microsoft does not use customer data from Microsoft 365 consumer and commercial applications to...
Society's Backend
Anti-Social Media
How our technology habituates the paradoxical relationship between social media and social isolation
a year ago
How our technology habituates the paradoxical relationship between social media and social isolation
Strange Loop Canon
AI embraces its product arc
fuzzy processors are entering mass production
7 months ago
fuzzy processors are entering mass production
Artificial Ignorance
10 AI predictions for 2024
Hey Siri, set a reminder for 365 days.
a year ago
Hey Siri, set a reminder for 365 days.
Rozado’s Visual...
The Increasing Negativity and Emotionality of News Media Headlines
Published article Introduction I have recently published a paper where we describe a chronological...
over a year ago
Published article Introduction I have recently published a paper where we describe a chronological (2000–2019) analysis of sentiment and emotion in 23 million headlines from 47 news media outlets popular in the United States. We used Transformer language models fine-tuned for...
AI Snake Oil
Is GPT-4 getting worse over time?
A new paper going viral has been widely misinterpreted
a year ago
A new paper going viral has been widely misinterpreted
Daniel Miessler
Unsupervised Learning NO. 365 | China’s Decline, MicrosoftAI, Creativity Ratio…
🎙️If you're not subscribed to the podcast version of the newsletter, please add it using with your...
a year ago
🎙️If you're not subscribed to the podcast version of the newsletter, please add it using with your favorite client! APPLE | SPOTIFY | OTHER SECURITY NEWS NYC Surveillance Amnesty International has revealed new research showing that the NYPD has over 15,000 cameras that can do...
Artificial Ignorance
AI Roundup 026: AudioCraft
August 4, 2023
a year ago
Society's Backend
What you need to understand about LLM creativity
An simple overview of temperature and its effect on LLM output
8 months ago
An simple overview of temperature and its effect on LLM output
fast.ai
The Jupyter+git problem is now solved
Previously, using git with Jupyter could create conflicts and break notebooks. With nbdev2, the...
over a year ago
Previously, using git with Jupyter could create conflicts and break notebooks. With nbdev2, the problem has been totally solved.
Society's Backend
Devin Has Exposed a Major Issue with Software Engineering
And isn't that we're all going to lose our jobs
9 months ago
And isn't that we're all going to lose our jobs
Society's Backend
Google Leads New AI Releases, OpenAI Gives First Glimpse of AGI with o3, AI Policy Will Be...
Society's Backend Reading List 12-23-2024
a week ago
Society's Backend Reading List 12-23-2024
Rozado’s Visual...
Out-of-office Donald Trump still more prominent in news media content than the current U.S....
In 2019, I helped my colleague Musa al-Gharbi document the extraordinary prominence of Donald Trump...
a year ago
In 2019, I helped my colleague Musa al-Gharbi document the extraordinary prominence of Donald Trump in news media content (see here). I have recently updated that previous analysis. Briefly stated, no U.S. president in recent history has received a similar amount of media...
Daniel Miessler
NO. 359 | WhatsLeak, CCTV Ban, Meta Threats
SECURITY NEWS There appears to be a WhatsApp data leak of over 500 million users' data in 84...
over a year ago
SECURITY NEWS There appears to be a WhatsApp data leak of over 500 million users' data in 84 countries. They're supposedly selling the data for $7K in the UK, and around $2K in the US and Germany. MORE The FCC has banned Chinese CCTV cameras on sensitive government sites and...
One Useful Thing
An AI Haunted World
Intelligence, everywhere.
a year ago
Intelligence, everywhere.
Society's Backend
Updates to Society's Backend
New benefits for paid subscribers, support Society's Backend for just $1/mo, a referral program, and...
10 months ago
New benefits for paid subscribers, support Society's Backend for just $1/mo, a referral program, and more
Society's Backend
No One Should Be GPU Poor
For everyone to have access to AGI, everyone must also have access to the compute to use it
7 months ago
For everyone to have access to AGI, everyone must also have access to the compute to use it
Marcus on AI
Cognitive scientist Gary Marcus says AI must be regulated. He has a plan.
Fantastic writeup of my views today at The Wall Street Journal:
a month ago
Fantastic writeup of my views today at The Wall Street Journal:
Strange Loop Canon
Evaluations are all we need
On analysing talent in LLMs
11 months ago
On analysing talent in LLMs
Rozado’s Visual...
The political orientation of the ChatGPT AI system
Applying the Pew Research Political Typology Quiz to a state-of-the-art AI Language model
over a year ago
Applying the Pew Research Political Typology Quiz to a state-of-the-art AI Language model
The Berkeley...
On the Stepwise Nature of <br> Self-Supervised Learning
Figure 1: stepwise behavior in self-supervised learning. When training common SSL algorithms, we...
a year ago
Figure 1: stepwise behavior in self-supervised learning. When training common SSL algorithms, we find that the loss descends in a stepwise fashion (top left) and the learned embeddings iteratively increase in dimensionality (bottom left). Direct visualization of embeddings...
fast.ai
Qualitative humanities research is crucial to AI
Following the thread of any seemingly quantitative issue in AI ethics (such as determining if...
over a year ago
Following the thread of any seemingly quantitative issue in AI ethics (such as determining if software to rate loan applicants is racially biased or evaluating YouTube’s recommendation system) quickly leads to a host of qualitative questions. Unfortunately, there is often a large...
One Useful Thing
Scaling: The State of Play in AI
A brief intergenerational pause...
3 months ago
A brief intergenerational pause...
One Useful Thing
When you give a Claude a mouse
Some quick impressions of an actual agent
2 months ago
Some quick impressions of an actual agent
AI Snake Oil
Will AI transform law?
The hype is not supported by current evidence
11 months ago
The hype is not supported by current evidence
Rozado’s Visual...
Sentiment Associations of Politically Loaded Terms in News Media
Summary of manuscript “Using Word Embeddings to Probe Sentiment Associations of Politically Loaded...
over a year ago
Summary of manuscript “Using Word Embeddings to Probe Sentiment Associations of Politically Loaded Terms in News and Opinion Articles from News Media Outlets”
Weighty Thoughts
Writing, Originality, and Why Does Anyone Care What I Write About?
Musings, Insecurities, and Thoughts on Writing a Book (or Substack)
a week ago
Musings, Insecurities, and Thoughts on Writing a Book (or Substack)
fast.ai
AI and Power: The Ethical Challenges of Automation, Centralization, and Scale
Moving AI ethics beyond explainability and fairness to empowerment and justice
a year ago
Moving AI ethics beyond explainability and fairness to empowerment and justice
IEEE Spectrum
Robot Photographer Takes the Perfect Picture
Finding it hard to get the perfect angle for your shot? PhotoBot can take the picture for you. Tell...
a month ago
Finding it hard to get the perfect angle for your shot? PhotoBot can take the picture for you. Tell it what you want the photo to look like, and your robot photographer will present you with references to mimic. Pick your favorite, and PhotoBot—a robot arm with a camera—will...
Society's Backend
AI Video Editing, MLX vs PyTorch, AI in Space, and More [Comprehensive ML Resource List for 6/28/24]
Here is a comprehensive list of all machine learning resources and updates from the past week. Thank...
6 months ago
Here is a comprehensive list of all machine learning resources and updates from the past week. Thank you for supporting Society's Backend! Don't forget to also follow me on X. Claude 3.5 Sonnet Can Long-Context Language Models Subsume Retrieval, RAG, SQL, and More?
IEEE Spectrum
In 2025, People Will Try Living in This Underwater Habitat
The future of human habitation in the sea is taking shape in an abandoned quarry on the border of...
5 days ago
The future of human habitation in the sea is taking shape in an abandoned quarry on the border of Wales and England. There, the ocean-exploration organization Deep has embarked on a multiyear quest to enable scientists to live on the seafloor at depths up to 200 meters for weeks,...
Society's Backend
Backend Biweekly #2: 112 Total Updates and Resources
Details on Metas training clusters, analysis of ML competitions in 2023, AI's impact on water...
9 months ago
Details on Metas training clusters, analysis of ML competitions in 2023, AI's impact on water resources, and more
One Useful Thing
The future, soon: what I learned from Bing's AI
We had a brief glimpse of two different types of AI. Both are significant
a year ago
We had a brief glimpse of two different types of AI. Both are significant
The Gradient
Neural algorithmic reasoning
In this article, we will talk about classical computation: the kind of computation typically found...
a year ago
In this article, we will talk about classical computation: the kind of computation typically found in an undergraduate Computer Science course on Algorithms and Data Structures [1]. Think shortest path-finding, sorting, clever ways to break problems down into simpler problems,...
Rozado’s Visual...
Mentions of Political Extremism in English Wikipedia
A data-driven exploration uncovers disparities. Are they shaped by editorial choices or broader...
3 weeks ago
A data-driven exploration uncovers disparities. Are they shaped by editorial choices or broader societal/historical dynamics?
Strange Loop Canon
Life in India is a series of bilateral negotiations
finding the new equilibrium for the next superpower
2 months ago
finding the new equilibrium for the next superpower
fast.ai
From Deep Learning Foundations to Stable Diffusion
Practical Deep Learning for Coders part 2, 2022
over a year ago
Practical Deep Learning for Coders part 2, 2022
Sam Altman
Helion
I’m delighted to be investing more in Helion. Helion is by far the most promising approach to fusion...
over a year ago
I’m delighted to be investing more in Helion. Helion is by far the most promising approach to fusion I’ve seen.
David and Chris are two of the most impressive founders and builders (in the sense of building fusion machines, in addition to building companies!) I have ever met,...
Marcus on AI
PSA: Microsoft may be training on your private data without your knowledge
The Great Data Heist continues.
a month ago
The Great Data Heist continues.
Sam Altman
DALL•E 2
Today we did a research launch of DALL•E 2, a new AI tool that can create and edit images from...
over a year ago
Today we did a research launch of DALL•E 2, a new AI tool that can create and edit images from natural language instructions.
Most importantly, we hope people love the tool and find it useful. For me, it’s the most delightful thing to play with we’ve created so far. I find it to...
Sam Altman
Idea Generation
The most common question prospective startup founders ask is how to get ideas for startups. The...
over a year ago
The most common question prospective startup founders ask is how to get ideas for startups. The second most common question is if you have any ideas for their startup.
But giving founders an idea almost always doesn’t work. Having ideas is among the most important qualities for...
Artificial Ignorance
What are logprobs?
Something to do with logs and probabilities.
12 months ago
Something to do with logs and probabilities.
PromptArmor Blog
Data exfiltration from Writer.com with indirect prompt injection
Authors: PromptArmor and Kai Greshake
a year ago
Authors: PromptArmor and Kai Greshake
Artificial Ignorance
Dealing with AI fatigue
Notes for myself, and maybe you too.
10 months ago
Notes for myself, and maybe you too.
One Useful Thing
Google's Gemini Advanced: Tasting Notes and Implications
And then there were two.
11 months ago
Rozado’s Visual...
The Increasing Frequency of Terms Denoting Political Extremism in U.S. and U.K. News Media
Published manuscript Twitter Thread Introduction I summarize here a published study co-authored with...
over a year ago
Published manuscript Twitter Thread Introduction I summarize here a published study co-authored with Eric Kaufmann where we examine longitudinally the prevalence of terms denoting far-right and far-left political extremism in more than 30 million written news and opinion articles...
One Useful Thing
It is starting to get strange.
Let's talk about ChatGPT with Code Interpreter & Microsoft Copilot
a year ago
Let's talk about ChatGPT with Code Interpreter & Microsoft Copilot
Artificial Ignorance
Workshop: Enough GPT To Be Dangerous with Sairam Sundaresan
Have you ever wondered what’s actually happening inside ChatGPT and Claude, but found “explainers”...
a month ago
Have you ever wondered what’s actually happening inside ChatGPT and Claude, but found “explainers” too complex or technical?
Andrej Karpathy blog
A from-scratch tour of Bitcoin in Python
.wrap {
max-width: 900px;
}
p {
font-family: sans-serif;
font-size: 15px;
...
over a year ago
.wrap {
max-width: 900px;
}
p {
font-family: sans-serif;
font-size: 15px;
font-weight: 300;
overflow-wrap: break-word; /* allow wrapping of very very long strings, like txids */
}
.post pre,
.post code {
background-color: #fafafa;
font-size: 13px; /*...
Rozado’s Visual...
The Increasing Prominence of Prejudice and Social Justice Rhetoric in UK News Media
I have recently published a report with Matthew Goodwin about the increasing prominence of prejudice...
over a year ago
I have recently published a report with Matthew Goodwin about the increasing prominence of prejudice and social justice rhetoric in UK news media. Recent years have seen considerable debate about the rise of political polarization in British society. Specifically, over the last...
Daniel Miessler
News & Analysis | NO. 352
over a year ago
Society's Backend
All Machine Learning Resources and Updates 06/14/2024
Below are all machine learning resources and updates from the past week you don't want to miss. I’ve...
6 months ago
Below are all machine learning resources and updates from the past week you don't want to miss. I’ve decided to try out separating paid/free updates. I’m constantly trying new ways to make information consumption easier for you while working within the confines of the platforms...
Weighty Thoughts
The Mystical Q
OpenAI Q*, Primer on Reinforcement Learning, and Implications
a year ago
OpenAI Q*, Primer on Reinforcement Learning, and Implications
The Gradient
Interpretability Creationism
On “interpretability creationism” – interpretability methods that only look at the final state of...
a year ago
On “interpretability creationism” – interpretability methods that only look at the final state of the model and ignore its evolution over the course of training
Weighty Thoughts
Most AI startups are doomed
Just because it matters doesn’t mean it’s defensible or profitable
a year ago
Just because it matters doesn’t mean it’s defensible or profitable
Strange Loop Canon
Discovery is the original sin of the modern age
technology, living post scarcity, and how to find what you want
a year ago
technology, living post scarcity, and how to find what you want
Weighty Thoughts
From 6 Weeks to 600 Seconds (or Less)
The revolutionary potential of utilizing AI for electronic hardware design
6 days ago
The revolutionary potential of utilizing AI for electronic hardware design
Daniel Miessler
UL NO. 354 | THE NEWS & ANALYSIS SERIES
SECURITY NEWS The US has implemented a number of aggressive export controls to stop China from...
over a year ago
SECURITY NEWS The US has implemented a number of aggressive export controls to stop China from attaining advanced semiconductors. And now it looks like the bans will be expanded to quantum computing and AI as well. NYTIMES | MY ANALYSIS BELOW It appears Bytedance had a plan to...
AI Snake Oil
Scientists should use AI as a tool, not an oracle
How AI hype leads to flawed research that fuels more hype
7 months ago
How AI hype leads to flawed research that fuels more hype
Made by Ollin
Mac Replacement Icons
A standardized, simplified, and beautiful icon set
over a year ago
A standardized, simplified, and beautiful icon set
Weighty Thoughts
Why we need AI as a society
We’re aging too fast (AKA my entire AI/robotics investment thesis)
7 months ago
We’re aging too fast (AKA my entire AI/robotics investment thesis)
AI Snake Oil
Tech policy is only frustrating 90% of the time
That’s what makes it worthwhile
9 months ago
That’s what makes it worthwhile
AI Snake Oil
The LLaMA is out of the bag. Should we expect a tidal wave of disinformation?
The bottleneck isn't the cost of producing disinfo, which is already very low.
a year ago
The bottleneck isn't the cost of producing disinfo, which is already very low.
Artificial Ignorance
AI Roundup 071: How do you like them AIs
June 14, 2024.
6 months ago
Artificial Ignorance
AI Roundup 068: The ScarJo thing
May 24, 2024.
7 months ago
One Useful Thing
One sentence.
Prompting for maximum impact (and why that is a bad idea)
a year ago
Prompting for maximum impact (and why that is a bad idea)
Sam Altman
The Virus
Although I still hope things will go differently, the experts I’ve spoken to think we are likely to...
over a year ago
Although I still hope things will go differently, the experts I’ve spoken to think we are likely to face a global tragedy—hundreds of thousands of deaths from Covid-19.
I hope that society views this as a warning for the future. Covid-19 is bad, but only a warm-up. I think it’s...
One Useful Thing
"Do not fear AI, puny humans... that is not meant as a threat."
What we can learn from a completely AI written & illustrated lecture
a year ago
What we can learn from a completely AI written & illustrated lecture
One Useful Thing
Superhuman: What can AI do in 30 minutes?
AI multiplies your efforts. I found out by how much...
a year ago
AI multiplies your efforts. I found out by how much...
Artificial Ignorance
How to leverage long form content with AI
Specific tools and tactics for authors, podcasters, and videographers.
7 months ago
Specific tools and tactics for authors, podcasters, and videographers.
The Gradient
The Artificiality of Alignment
This essay first appeared in Reboot.
Credulous, breathless coverage of “AI existential risk”...
a year ago
This essay first appeared in Reboot.
Credulous, breathless coverage of “AI existential risk” (abbreviated “x-risk”) has reached the mainstream. Who could have foreseen that the smallcaps onomatopoeia “ꜰᴏᴏᴍ” — both evocative of and directly derived from children’s cartoons —
One Useful Thing
What AI can do with a toolbox... Getting started with Code Interpreter
Democratizing data analysis with AI
a year ago
Democratizing data analysis with AI
fast.ai
From Deep Learning Foundations to Stable Diffusion
We’ve released our new course with over 30 hours of video content.
a year ago
We’ve released our new course with over 30 hours of video content.
IEEE Spectrum
Where’s My Robot?
See the interactive version of this story on our site →
a month ago
See the interactive version of this story on our site →
AI Snake Oil
Starting reading the AI Snake Oil book online today
The book will be published on September 24
3 months ago
The book will be published on September 24
Society's Backend
Digitizing Smell, Automatic Prompt Optimization, Targeted AI Regulation, an Intro to AI Agents, and...
Society's Backend Reading List 11-04-2024
2 months ago
Society's Backend Reading List 11-04-2024
One Useful Thing
Feats to astonish and amaze
A compendium of things I didn't think AI should be able to do
a year ago
A compendium of things I didn't think AI should be able to do
Daniel Miessler
Sponsored Interview: Erkang Zheng of JupiterOne
In this standalone episode we’re doing a sponsored interview with Erkang Zheng of JupiterOne. So...
over a year ago
In this standalone episode we’re doing a sponsored interview with Erkang Zheng of JupiterOne. So Jupiter One is a special company to me. I just built a vuln management program at Robinhood based around them, and I believe so much in their vision that I’m looking to actually...
Daniel Miessler
NO. 361 | GPT++, Apple Security, CISA Cuba…
SECURITY NEWS South Korean authorities are warning that North Koreans are disguising themselves and...
over a year ago
SECURITY NEWS South Korean authorities are warning that North Koreans are disguising themselves and getting jobs in South Korea. The saddest part is that it appears to be just another income generation scheme, meaning they use the salaries to fund the North Korean nuclear...
Rozado’s Visual...
An Analysis of AI Political Preferences from a European Perspective
Introduction
2 months ago
Artificial Ignorance
How to build an AI search engine (Part 1)
Working with Claude, Brave, and streaming responses.
3 months ago
Working with Claude, Brave, and streaming responses.
Artificial Ignorance
AI Roundup 053: Bye bye Bard
February 9, 2024.
11 months ago
One Useful Thing
How to use AI to do practical stuff: A new guide
People often ask me how to use AI. Here's an overview with lots of links.
a year ago
People often ask me how to use AI. Here's an overview with lots of links.
Artificial Ignorance
AI Roundup 042: sama drama
November 24, 2023.
a year ago
Matt Mazur
Friday Updates: Smart Icons, Automatic Suggestions, Dealing with Spammers, Better Icon Colors
Preceden Lots of updates to Preceden this week: Improving the UX for the AI Suggestions When we...
a year ago
Preceden Lots of updates to Preceden this week: Improving the UX for the AI Suggestions When we rolled out the AI Suggestions feature last week, the typical experience for the user would go something like this: Lots of UX issues there though: To remedy this, I updated Preceden to...
Marcus on AI
o3, AGI, the art of the demo, and what you can expect in 2025
OpenAI’s new model was revealed yesterday; its most fervent believers think AGI has already arrived....
2 weeks ago
OpenAI’s new model was revealed yesterday; its most fervent believers think AGI has already arrived. Here’s what you should pay attention to in the coming year.
One Useful Thing
Now is the time for grimoires
It isn't data that will unlock AI, it is human expertise
a year ago
It isn't data that will unlock AI, it is human expertise
Artificial Ignorance
Workshop: Prompt Engineering for Claude
Watch now | If you’ve ever wanted to level up your AI prompting skills, this is the workshop for...
3 months ago
Watch now | If you’ve ever wanted to level up your AI prompting skills, this is the workshop for you.
The Berkeley...
Linguistic Bias in ChatGPT: Language Models Reinforce Dialect Discrimination
Sample language model responses to different varieties of English and native speaker...
3 months ago
Sample language model responses to different varieties of English and native speaker reactions.
ChatGPT does amazingly well at communicating with people in English. But whose English?
Only 15% of ChatGPT users are from the US, where Standard American English is the default. But...
IEEE Spectrum
Video Friday: Sleepy Robot Baby
Video Friday is your weekly selection of awesome robotics videos, collected by your friends at IEEE...
2 days ago
Video Friday is your weekly selection of awesome robotics videos, collected by your friends at IEEE Spectrum robotics. We also post a weekly calendar of upcoming robotics events for the next few months. Please send us your events for inclusion.
RoboCup German Open: 12–16 March...
One Useful Thing
The shape of the shadow of The Thing
We can start to see, dimly, what the near future of AI looks like.
a year ago
We can start to see, dimly, what the near future of AI looks like.
IEEE Spectrum
This Mobile 3D Printer Can Print Directly on Your Floor
Waiting for each part of a 3D-printed project to finish, taking it out of the printer, and then...
a month ago
Waiting for each part of a 3D-printed project to finish, taking it out of the printer, and then installing it on location can be tedious for multi-part projects. What if there was a way for your printer to print its creation exactly where you needed it? That’s the promise of...
One Useful Thing
Embracing weirdness: What it means to use AI as a (writing) tool
AI is strange. We need to learn to use it.
a year ago
AI is strange. We need to learn to use it.
Marcus on AI
Casey (Newton) Strikes Out
His continued skepticism of AI skepticism continues to miss the mark
3 weeks ago
His continued skepticism of AI skepticism continues to miss the mark
Weighty Thoughts
Software Engineering is Doomed
Or is it?
6 months ago
fast.ai
Mojo may be the biggest programming language advance in decades
Mojo is a new programming language, based on Python, which fixes Python’s performance and deployment...
a year ago
Mojo is a new programming language, based on Python, which fixes Python’s performance and deployment problems.
Weighty Thoughts
Why ChatGPT Strawberry o1 (and other LLMs) will Never be Good at Diagnosis
“Connectionist” vs. Knowledge-Based AI
3 months ago
“Connectionist” vs. Knowledge-Based AI
Artificial Ignorance
Why Claude 3 is a big upgrade
On Monday, Anthropic released Claude 3, a family of models that hit new highs on various benchmarks,...
10 months ago
On Monday, Anthropic released Claude 3, a family of models that hit new highs on various benchmarks, add multi-modal capabilities, and are a real competitor to GPT-4.
Matt Mazur
Emergent Mind in The Atlantic
One of the many people who saw the incorrect Google Quick Answer about Kenya was an editor at The...
a year ago
One of the many people who saw the incorrect Google Quick Answer about Kenya was an editor at The Atlantic who asked Caroline Mimbs Nyce, one of their reporters, to look into it. Caroline interviewed me for the article they just published which focused on the challenges Google is...
AI Snake Oil
Students are acing their homework by turning in machine-generated essays. Good.
Teachers adapted to the calculator. They can certainly adapt to language models.
over a year ago
Teachers adapted to the calculator. They can certainly adapt to language models.
The Berkeley...
Asymmetric Certified Robustness via Feature-Convex Neural Networks
Asymmetric Certified Robustness via Feature-Convex Neural Networks
TLDR: We propose the asymmetric...
a year ago
Asymmetric Certified Robustness via Feature-Convex Neural Networks
TLDR: We propose the asymmetric certified robustness problem, which requires certified robustness for only one class and reflects real-world adversarial scenarios. This focused setting allows us to introduce...
Daniel Miessler
NO. 367 | Hive Ransom, Anti-Google, Software 2.0…
🎙️If you’re not subscribed to the podcast version of the newsletter, please add it using with your...
a year ago
🎙️If you’re not subscribed to the podcast version of the newsletter, please add it using with your favorite client! APPLE | SPOTIFY | OTHER SECURITY NEWS The FBI infiltrated the HIVE ransomware group, stopping over $130 million in ransomware attacks. HIVE is known for going...
One Useful Thing
Freeing the chatbot
Intelligence, of a sort, is going to be all around us
8 months ago
Intelligence, of a sort, is going to be all around us
The Berkeley...
Reverse engineering the NTK: towards first-principles architecture design
Deep neural networks have enabled technological wonders ranging from voice recognition to machine...
over a year ago
Deep neural networks have enabled technological wonders ranging from voice recognition to machine transition to protein engineering, but their design and application is nonetheless notoriously unprincipled.
The development of tools and methods to guide this process is one of the...
Artificial Ignorance
AI Roundup 098: Flash Thinking
December 20, 2024.
2 weeks ago
Weighty Thoughts
Open AI's Valuation and a Favor to Ask
A guest post and an in-person panel
4 months ago
A guest post and an in-person panel
Weighty Thoughts
Deep Tech Startups Are Not Software Startups
Or why AI and “Software/SaaS” are apples and oranges
11 months ago
Or why AI and “Software/SaaS” are apples and oranges
AI Snake Oil
Three Ideas for Regulating Generative AI
Policy input to the federal government from a Stanford-Princeton team
a year ago
Policy input to the federal government from a Stanford-Princeton team
One Useful Thing
Getting started with AI: Good enough prompting
Don't make this hard
a month ago
Artificial Ignorance
AI Roundup 077: A good week for open-source
July 26, 2024.
5 months ago
Rozado’s Visual...
DepolarizingGPT
A Political Chatbot that Gives 3 Politically Diverse Answers to Every Prompt
a year ago
A Political Chatbot that Gives 3 Politically Diverse Answers to Every Prompt
Society's Backend
1000+ AI Agents Make a Civilization, xAI Creates the Largest Supercomputer in 4 Months, Possible AI...
Machine learning resources and updates 2024-09-09
3 months ago
Machine learning resources and updates 2024-09-09
Rozado’s Visual...
Which is the Wokest AI?
A Wokeness Ranking of LLMs
10 months ago
A Wokeness Ranking of LLMs
Society's Backend
OpenAI's o1, Model Merging, California Approves AI Regulation, and More
Machine learning resources and updates 2024-09-17
3 months ago
Machine learning resources and updates 2024-09-17
Rozado’s Visual...
Define Wokeness! Or how you shall know a word by the company it keeps
Visualizing what words often appear in the vicinity of woke/wokeness in news media content...
a year ago
Visualizing what words often appear in the vicinity of woke/wokeness in news media content illustrates why communication is almost impossible between red and blue America
Artificial Ignorance
AI Roundup 079: Don't call it an acquisition
August 9, 2024.
4 months ago
Matt Mazur
Going Full Time on My SaaS After 13 Years
In January 2010 I soft-launched launched Preceden, a web-based timeline maker tool, followed a few...
over a year ago
In January 2010 I soft-launched launched Preceden, a web-based timeline maker tool, followed a few weeks later by a larger launch on HackerNews: Today – almost 13 years to the day since the initial launch – I’m going full time on it and I couldn’t be more excited. A brief history...
Artificial Ignorance
From Stable Diffusion to Stable Everything
Inside Stability AI's roster of AI models.
7 months ago
Inside Stability AI's roster of AI models.
One Useful Thing
What people ask me most. Also, some answers.
A FAQ of sorts
a year ago
The Gradient
Text-to-CAD: Risks and Opportunities
In the realm of AI-powered text-to-CAD, there's promise, but also a surge in subpar designs. Can we...
a year ago
In the realm of AI-powered text-to-CAD, there's promise, but also a surge in subpar designs. Can we steer this technology towards better outcomes?
Society's Backend
AI Reading List 1: Understand Transformers, Reflection-70B Update, and LLMs Still Cannot Reason
Society's Backend Reading List 10-07-2024
2 months ago
Society's Backend Reading List 10-07-2024
Society's Backend
Godmother of AI Warns Against Regulation, Apple Intelligence System Prompts, Faster and Cheaper AI,...
Machine learning resources and updates 8/12/2024
4 months ago
Machine learning resources and updates 8/12/2024
AI Snake Oil
On the Societal Impact of Open Foundation Models
Adding precision to the debate on openness in AI
10 months ago
Adding precision to the debate on openness in AI
The Gradient
Deep learning for single-cell sequencing: a microscope to see the diversity of cells
On the the pivotal role that Deep Learning has played as a key enabler for advancing single-cell...
11 months ago
On the the pivotal role that Deep Learning has played as a key enabler for advancing single-cell sequencing technologies.
The Berkeley...
Generating 3D Molecular Conformers via Equivariant Coarse-Graining and Aggregated Attention
-->
Figure 1: CoarsenConf architecture.
(I) The encoder $q_\phi(z| X, \mathcal{R})$ takes the...
a year ago
-->
Figure 1: CoarsenConf architecture.
(I) The encoder $q_\phi(z| X, \mathcal{R})$ takes the fine-grained (FG) ground truth conformer $X$, RDKit approximate conformer $\mathcal{R}$ , and coarse-grained (CG) conformer $\mathcal{C}$ as inputs (derived from $X$ and a predefined...
One Useful Thing
Using AI to make teaching easier & more impactful
Here are five strategies and prompts that work for GPT-3.5 & GPT-4
a year ago
Here are five strategies and prompts that work for GPT-3.5 & GPT-4
The Gradient
A Brief Overview of Gender Bias in AI
A brief overview and discussion on gender bias in AI
9 months ago
A brief overview and discussion on gender bias in AI
PromptArmor Blog
Slack AI data exfiltration from private channels via indirect prompt injection
Authors: PromptArmor
4 months ago
Matt Mazur
Screw it, I’m Keeping Emergent Mind
A few months ago I announced I was going to try to sell Emergent Mind, my AI news aggregator, so I...
a year ago
A few months ago I announced I was going to try to sell Emergent Mind, my AI news aggregator, so I could focus on Preceden, my SaaS timeline maker. I wound up having a lot of discussions with potential buyers, but in the end the offers I received were either too low to be worth...
fast.ai
My family’s unlikely homeschooling journey
Prior to 2020, we never expected to homeschool, and now we have committed to it long-term.
over a year ago
Prior to 2020, we never expected to homeschool, and now we have committed to it long-term.
Rozado’s Visual...
The unequal treatment of demographic groups by ChatGPT/OpenAI content moderation system
Should AI systems treat different demographic groups unequally?
a year ago
Should AI systems treat different demographic groups unequally?
Stories by Andrej...
ICLR 2017 vs arxiv-sanity
over a year ago
The Berkeley...
Goal Representations for Instruction Following
Goal Representations for Instruction Following
Figure title. Figure caption. This image is...
a year ago
Goal Representations for Instruction Following
Figure title. Figure caption. This image is centered and set to 50%
page width. -->
A longstanding goal of the field of robot learning has been to create generalist agents that can perform tasks for humans. Natural language has...
Artificial Ignorance
10 AI stories that shaped 2024
Agents, deepfakes, Strawberries, and more.
a week ago
Agents, deepfakes, Strawberries, and more.
Marcus on AI
Sora still appears to have trouble with physics
Exactly as I warned in February
3 weeks ago
Exactly as I warned in February
Artificial Ignorance
How to build an AI search engine (Part 2)
VC funding, here we come.
2 months ago
VC funding, here we come.
AI Snake Oil
Four more things we worked on in 2022
We had a busy 2022. Here are a few things we worked on but didn’t cover here.
over a year ago
We had a busy 2022. Here are a few things we worked on but didn’t cover here.
Strange Loop Canon
AI bill vetoed; what's next?
Make AI regulations evidence based
3 months ago
Make AI regulations evidence based
Society's Backend
How to Know if Your Data is Being Used Maliciously
The information you should know about a company before gifting them your data
a year ago
The information you should know about a company before gifting them your data
Society's Backend
Discussions around OpenAI's o1, Superhuman AI, When AI Should Be An App, and More
Discussions from the past week: 09/16/2024
3 months ago
Discussions from the past week: 09/16/2024
Society's Backend
Why Machine Learning Systems Misbehave
And what makes them so difficult to work with
5 months ago
And what makes them so difficult to work with
Society's Backend
Coming soon!
Weekly publications to teach you about the technology that governs our society and how it impacts...
a year ago
Weekly publications to teach you about the technology that governs our society and how it impacts you. Coming in August!
Artificial Ignorance
AI Roundup 029: Hug all the faces
August 25, 2023
a year ago
Daniel Miessler
News & Analysis | NO. 353
over a year ago
The Gradient
Shape, Symmetries, and Structure: The Changing Role of Mathematics in Machine Learning Research
What is the Role of Mathematics in Modern Machine Learning?
The past decade has witnessed a shift in...
a month ago
What is the Role of Mathematics in Modern Machine Learning?
The past decade has witnessed a shift in how progress is made in machine learning. Research involving carefully designed and mathematically principled architectures result in only marginal improvements while...
Society's Backend
JAX is for More Than Just Machine Learning
What JAX is and its potential applications
7 months ago
What JAX is and its potential applications
Artificial Ignorance
Tutorial: How to chat with your documents
A step-by-step guide to doing Q&A with your data, using LlamaIndex and OpenAI.
a year ago
A step-by-step guide to doing Q&A with your data, using LlamaIndex and OpenAI.
Weighty Thoughts
VC Office Hours
Talk to James Wang, author of Weighty Thoughts and General Partner of Creative Ventures
11 months ago
Talk to James Wang, author of Weighty Thoughts and General Partner of Creative Ventures
Artificial Ignorance
AI Roundup 084: Strawberry / o1
September 13, 2024.
3 months ago
Society's Backend
LLM Research Recap of 2024, Fine-tuning LLM Judges, Amazon's Nova Models, Google's Genie 2, and More
Society's Backend Reading List 12-10-2024
3 weeks ago
Society's Backend Reading List 12-10-2024
Andrej Karpathy blog
Self-driving as a case study for AGI
Sparked by progress in Large Language Models (LLMs), there’s a lot of chatter recently about AGI,...
11 months ago
Sparked by progress in Large Language Models (LLMs), there’s a lot of chatter recently about AGI, its timelines, and what it might look like. Some of it is hopeful and optimistic, but a lot of it is fearful and doomy, to put it mildly. Unfortunately, a lot of it is also very...
Marcus on AI
ChatGPT, at Age Two
The bullshit just keeps on coming.
a month ago
The bullshit just keeps on coming.
AI Snake Oil
We Looked at 78 Election Deepfakes. Political Misinformation is not an AI Problem.
Technology Isn’t the Problem—or the Solution.
3 weeks ago
Technology Isn’t the Problem—or the Solution.
Matt Mazur
Don’t Self Host Unlicensed Proxima Nova Fonts
I’m a big fan of the Proxima Nova font and have been using it on Preceden for years: For a long time...
a year ago
I’m a big fan of the Proxima Nova font and have been using it on Preceden for years: For a long time I was loading Proxima Nova on Preceden via Typekit (a hosted web font service) for $49.99/year, but at some point I decided to self-host it to avoid the third party request which...
Artificial Ignorance
AI Roundup 096: Happy Shipmas
December 6, 2024.
4 weeks ago
Artificial Ignorance
The intern and the coach
A recent post from Wharton professor Ethan Mollick shared an impressive new study. In a nutshell, it...
a year ago
A recent post from Wharton professor Ethan Mollick shared an impressive new study. In a nutshell, it found that BCG consultants who used GPT-4 were up to 43% more effective at tasks vs. employees who didn’t.
Frank’s Ramblings
My Experience Living and Working in China, Part I
In this four-part article, I’ll go over some of the lessons I learned living and doing business in...
over a year ago
In this four-part article, I’ll go over some of the lessons I learned living and doing business in China’s tech industry. During my time in China, I’ve led a team of 10+ engineers to develop a location-based IoT and sensing platform, co-founded an open-source project called...
Artificial Ignorance
Distributing the future
A reminder that things take time.
7 months ago
A reminder that things take time.
Marcus on AI
𝗼𝟯 “𝗔𝗥𝗖 𝗔𝗚𝗜” 𝗽𝗼𝘀𝘁𝗺𝗼𝗿𝘁𝗲𝗺 𝗺𝗲𝗴𝗮𝘁𝗵𝗿𝗲𝗮𝗱: 𝘄𝗵𝘆 𝘁𝗵𝗶𝗻𝗴𝘀 𝗴𝗼𝘁...
Kevin Roose, of Hard Fork and NYT, was so impressed with OpenAI’s rollout that he joked “of course...
a week ago
Kevin Roose, of Hard Fork and NYT, was so impressed with OpenAI’s rollout that he joked “of course they have to announce AGI the day my vacation starts”.
Made by Ollin
A Video History of Self-Driving Cars
over a year ago
One Useful Thing
Thinking Like an AI
A little intuition can help
2 months ago
A little intuition can help
Strange Loop Canon
I wrote a book!
AI enables us to, finally, get "soft" APIs
a year ago
AI enables us to, finally, get "soft" APIs
Artificial Ignorance
AI Roundup 081: Creative differences
August 23, 2024.
4 months ago
Artificial Ignorance
Getting to the top of the GPT Store (and building an AI-native search engine, too)
Listen now | A conversation with Christian Salem, founder and CPO of Consensus.
8 months ago
Listen now | A conversation with Christian Salem, founder and CPO of Consensus.
One Useful Thing
Working with AI: Two paths to prompting
Don't overcomplicate things
a year ago
Don't overcomplicate things
IEEE Spectrum
Video Friday: Mars Chopper
Video Friday is your weekly selection of awesome robotics videos, collected by your friends at IEEE...
3 weeks ago
Video Friday is your weekly selection of awesome robotics videos, collected by your friends at IEEE Spectrum robotics. We also post a weekly calendar of upcoming robotics events for the next few months. Please send us your events for inclusion.
ICRA 2025: 19–23 May 2025, ATLANTA,...
One Useful Thing
Magic for English Majors
Programming in prose in an AI-haunted world
a year ago
Programming in prose in an AI-haunted world
Artificial Ignorance
AI Roundup 031: Think of the children
September 8, 2023
a year ago
Rozado’s Visual...
The Academic Literature and its Increasing Emphasis on Prejudice and Social Justice
Published article Twitter thread In previous scholarly work (here and here), I documented a marked...
over a year ago
Published article Twitter thread In previous scholarly work (here and here), I documented a marked increase of references to prejudice in US, UK and Spanish news media content. The work summarized here investigates the prevalence dynamics of prejudice-denoting terms in 175...
Matt Mazur
AOL Underground Podcast Interview about AOL-Files.com
Back in 1998 when I was 13 years old I got heavily involved in the AOL hacking scene, originally...
over a year ago
Back in 1998 when I was 13 years old I got heavily involved in the AOL hacking scene, originally building add-on software called progs (Revolution, Meridian), publishing code libraries called bas files (Alpha32), and later co-founding AOL-Files.com (where I went by the hacker...
Society's Backend
Why Gemini's Struggles Aren't Straightforward
And an overview of LLM security issues
9 months ago
And an overview of LLM security issues
Society's Backend
5 Highlights From Society's Backend in 2024
The top 5 articles and resources
4 days ago
The top 5 articles and resources
AI Snake Oil
Evaluating LLMs is a minefield
Annotated slides from a recent talk
a year ago
Annotated slides from a recent talk
Artificial Ignorance
AI Roundup 037: The AI IPCC
October 20, 2023.
a year ago
One Useful Thing
An Opinionated Guide to Which AI to Use: ChatGPT Anniversary Edition
A simple answer, and then a less simple one.
a year ago
A simple answer, and then a less simple one.
The Gradient
An Introduction to the Problems of AI Consciousness
Once considered a forbidden topic in the AI community, discussions around the concept of AI...
a year ago
Once considered a forbidden topic in the AI community, discussions around the concept of AI consciousness are now taking center stage, marking a significant shift since the current AI resurgence began over a decade ago.
Sam Altman
Join the YC Software Team
If you want to get funded by YC as a founder in the future, but you don't have a startup that's...
over a year ago
If you want to get funded by YC as a founder in the future, but you don't have a startup that's ready for that yet, joining the YC software team is a great hack to get there.
The YC software team is a small group of hackers in SF that write the software that makes all the parts...
Society's Backend
AI Reading List 2: Mathematical Limitations of LLMs, A SQL Roadmap for Data Science, and...
Society's Backend Reading List 10-14-2024
2 months ago
Society's Backend Reading List 10-14-2024
Society's Backend
Positioning Myself: The Greatest Piece of Career Advice I've Ever Received
And how it changed my personal life too
a year ago
And how it changed my personal life too
fast.ai
A new old kind of R&D lab
Answer.AI is a new kind of AI R&D lab which creates practical end-user products based on...
a year ago
Answer.AI is a new kind of AI R&D lab which creates practical end-user products based on foundational research breakthroughs.
One Useful Thing
Change blindness
21 months later
4 months ago
One Useful Thing
Post-apocalyptic education
What comes after the Homework Apocalypse
4 months ago
What comes after the Homework Apocalypse
The Berkeley...
Rethinking Human-in-the-Loop for Artificial Augmented Intelligence
How do we build and evaluate an AI system for real-world applications? In most AI research, the...
over a year ago
How do we build and evaluate an AI system for real-world applications? In most AI research, the evaluation of AI methods involves a training-validation-testing process. The experiments usually stop when the models have good testing performance on the reported datasets because...
Matt Mazur
Redesigning Preceden’s Pricing Page
Milan (Preceden’s designer) and I recently wrapped up a project to redesign Preceden’s pricing page....
a year ago
Milan (Preceden’s designer) and I recently wrapped up a project to redesign Preceden’s pricing page. Here’s the previous above-the-fold content: And here’s how the new design turned out: Few things to highlight: Very happy with how it turned out. Kudus to Milan for suggesting we...
Artificial Ignorance
AI Roundup 067: GPT-4o and Google I/O
May 17, 2024.
7 months ago
Society's Backend
Weekly Backend #7: 39 Resources and Updates
GPT-4o, Google I/O, Fugaku LLM, Prep for Machine Learning Interviews, and more
7 months ago
GPT-4o, Google I/O, Fugaku LLM, Prep for Machine Learning Interviews, and more
Society's Backend
I Enjoy Technical Interviews and You Can Too
A simple change in mindset that completely changed the way I interview
a year ago
A simple change in mindset that completely changed the way I interview
AI Snake Oil
Is Avoiding Extinction from AI Really an Urgent Priority?
The history of technology suggests that the greatest risks come not from the tech, but from the...
a year ago
The history of technology suggests that the greatest risks come not from the tech, but from the people who control it
Artificial Ignorance
AI Roundup 061: The AI innovator's dilemma
April 5, 2024.
9 months ago
One Useful Thing
Detecting the Secret Cyborgs
The AI Trap for Organizations
a year ago
The AI Trap for Organizations
Artificial Ignorance
Lessons from 139 YC AI startups (S23)
The evolution of the AI landscape, and YC's Copilot era.
a year ago
The evolution of the AI landscape, and YC's Copilot era.
Society's Backend
LLM-as-a-Judge, Instruction Pretraining, Solving Benchmarks Instead of Real-World ML Problems, and...
Weekly updates and resources 7/22/24
5 months ago
Weekly updates and resources 7/22/24
Artificial Ignorance
AI Roundup 046: AI has a CSAM problem
December 22, 2023.
a year ago
Matt Mazur
Progress on TimelineGPT, Emergent Mind missteps, finding balance
Hey all 👋! It’s been a minute since my last post (for reasons I’ll get into below) so here’s...
a year ago
Hey all 👋! It’s been a minute since my last post (for reasons I’ll get into below) so here’s periodic update on what I’ve been up to: TimelineGPT About two months ago I launched an in-app tool for Preceden that provides GPT-powered event suggestions to users to help them build...
Artificial Ignorance
Tutorial: How to build an AI agent (Part 1)
Part 1: What is an agent, and how do they work?
a year ago
Part 1: What is an agent, and how do they work?
Matt Mazur
When LTD Purchasers Meet an Inactive User Policy
Last year I participated in a Lifetime Deal (LTD) promotion to offer Preceden to the AppSumo...
a year ago
Last year I participated in a Lifetime Deal (LTD) promotion to offer Preceden to the AppSumo community. Maybe I’ll dive into my experience there in another post, but I wanted to share an interesting thing that’s happening now, a year after the deal ended. AppSumo has a policy...
Daniel Miessler
Napkin Ideas Around What Changes to Expect Post-ChatGPT
Work Replacement Talent Magnification Solopreneuers AI Specialists Idea Dominance Use Cases Random...
over a year ago
Work Replacement Talent Magnification Solopreneuers AI Specialists Idea Dominance Use Cases Random Thoughts If you’re reading this you already know the internet is on fire over the new GPTChatBot from OpenAI. There are people using it to create full virtual machines, to be their...
Matt Mazur
Turning Down $7k for a Side Project I Announced Two Weeks Ago I was Shutting Down
About two weeks ago I announced I was shutting down LearnGPT. After the announcement, I received a...
a year ago
About two weeks ago I announced I was shutting down LearnGPT. After the announcement, I received a lot of interest from people interested in taking it over, so I decided to try to sell it instead of shutting it down. I wound up receiving two offers, one for $6.5k and one for $7k,...
Weighty Thoughts
Compute is Overrated as AI’s Bottleneck
You can’t just blindly extrapolate compute requirements
a year ago
You can’t just blindly extrapolate compute requirements
Made by Ollin
Priors for Autonomous Vehicle Development
over a year ago
Artificial Ignorance
AI Roundup 028: Sunak's Safety Summit
August 18, 2023
a year ago
One Useful Thing
Not much is changing, a lot is changing
OpenAI, Microsoft, and the OpenOffspring
a year ago
OpenAI, Microsoft, and the OpenOffspring
Artificial Ignorance
The State of AI Engineering
Notes from the first AI Engineer Summit.
a year ago
Notes from the first AI Engineer Summit.
Full Width 
Shortcuts 
Sign Up 
Log In 
Articles 
Reading List 