How AI Image Generators Work (Stable Diffusion / Dall-E) - Computerphile ⌘ Read more
Yudkowsky moved AI alignment research forward by 4 years, but he also sped up timelines by 2.5 years, so it all cancels out
Angels from Dr. Who but everytime you look away there’s new AI capabilities progress
“If you don’t make it beautiful, it’s for sure doomed”: putting the Vault in GitHub’s Arctic Code Vault
GitHub this month installed a massive steel vault, etched with striking AI-generated art, deep within an Arctic mountain, finalizing its Arctic Code Vault. This vault contains the 188 reels of hardened archival film which will preserve the 02/02/202 snapshot of every active public GitHub repository for 1,000 years. It also now includes a … ⌘ Read more
Rotation
⌘ Read more
GitHub Copilot now available for teachers
After a year in technical preview, GitHub Copilot, an AI pair programmer, is now free for all teachers verified on GitHub Global Campus. ⌘ Read more
What you can expect at GitHub Universe 2022: cloud, security, community, and AI
Register now to attend GitHub Universe virtually or in-person at the Yerba Buena Center for the Arts in San Francisco on November 9-10. ⌘ Read more
It seems quite unlikely to me that humans will cease to care about wealth after the singularity (15% maybe?), or that markets will cease to exist after the singularity (30% or sth). so the best investment strategy now for post-singularity scenarios is to invest broadly in the economy. not sure about divesting from AGI companies (bc they accelerate danger) or AI hardware companies (ditto)—they’re going to especially valuable post-singularity.
Tricking AI Image Recognition - Computerphile ⌘ Read more
just thought to myself “hopefully a bigger pandemic hits, that sounds like it’ll delay ai capabilities progres”, which, no,,,
International Linguistic Symposium on PhD Research Training
[Sponsored Article]
The Hongkong-Beijing-Lancaster Symposium on Research Methodologies for PhD Studies in Linguistics (REMLing) was held online on 17 and 18 March 2022.
REMLing 2022 was jointly organised by the Halliday Centre for Intelligent Applications for Language Studies (HCLS) at City University of Hong Kong, the MIIT Centre of AI-based Linguistic Information Proce … ⌘ Read more
Google pulls plug on “Sentient A.I.” after it begins using Arch Linux
“BTW, I use Arch. Therefore I am.” ⌘ Read more
Instagram to protect young users with new AI tool to verify age
Global lawmakers have been vocal in demanding the social media service safeguard young people from adult content and invasions of their privacy. ⌘ Read more
Microsoft supports Hong Kong start-ups in partnership with Science Park, Cyberport
More than 40 local start-ups have joined the programme, including AI-powered marine trash-collecting robot maker Clearbot and algorithmic trading marketplace AlgoBot. ⌘ Read more
China supercomputer achieves global first with ‘brain-scale’ AI model
Research team says latest Sunway machine is on a par with the US Frontier, named as the world’s most powerful just weeks earlier. ⌘ Read more
GitHub Copilot is generally available to all developers
We’re making GitHub Copilot, an AI pair programmer that suggests code in your editor, generally available to all developers for $10 USD/month or $100 USD/year. It will also be free to use for verified students and maintainers of popular open source projects. ⌘ Read more
Superintelligent AIs
⌘ Read more
Have you heard about the guy who worked on the Google AI chat bot? It is more than a chat bot and the conversation he published (got put on paid leave for doing that) is pretty scary : https://cajundiscordian.medium.com/is-lamda-sentient-an-interview-ea64d916d917
“The problem with Marcus’ argument is that the only alternative to statistical AI is spiritual AI“ no what the fuck why would you say this did you even read the sequences i don’t even know where to start with that
ai alignment is great bc nobody knows what it should be, so you can do whatever you want and call it alignment.
if aliens exists and UAPs are caused by them, what implication does that have for AI safety?
“AI” refers to the observation that computers, over time, can do more & more things
I really look forward to the next AI winter! We’ll all cozy up inside by the fireside in thick markov blankets, everyone finds a new agent foundations paper in their stocking, and we can listen to uncle Paul have his discussion with uncle Eliezer when the whole family re-unites…
Ramanujan voice “Oh, AI alignment? That’s just a continued fraction” pulls out a notebook
Winning MONEY with AI – Understanding Algorithms & Predicting March Madness with Neil deGrasse Tyson ⌘ Read more
whether cryptocurrencies are more or less likely to be stable during a multipolar ai takeoff depends on whether our current cryptography is “endgame” or not, i.e. whether it’s in practice basically uncrackable by any advanced actor
GPT-3 is No Longer the Only Game in Town - GPT-3 was by far the largest AI model of its kind last year. Now? Not so much. ⌘ Read more
a popular measure for how well-secured the AIs’ utility functions were against self-hacking ended up being how many paperclips they produced before reaching nirvana
Clearview AI Offered Free Facial Recognition Trials To Police All Around The World
Clearview marketed its facial recognition system across Europe by offering free trials at police conferences, where it was often presented as a tool to help find predators and victims of child sex abuse. ⌘ Read more
maybe elites appear incompetent because they’re mostly trying to prevent AI disaster behind the curtains? perhaps that’s why GPT-4 hasn’t appeared yet – they coordinated not to do it.
lol, just realised that standard AI risk is just environmental disaster risk amped up to 10^(10^10), only that the direct perpetrator isn’t humanity (we’re only the proximate perpetrator). huh
beating the ender dragon in minecraft with dying <100 times (but arbitrary training on video beforehand) seems like a nice candidate for another fire alarm (I wonder whether people would panic then, or whether that challenge is already AI-complete)
Cosmic Queries - Nerd Olympics, Athletes AI, and Usain Bolt ⌘ Read more…
Malcolm Gladwell - AI, Autonomous Vehicles, and Race – Oh my! ⌘ Read more…
Jake explains how to correctly encode data for POST requests as no AI can.
Introducing GitHub Copilot: your AI pair programmer ⌘ Read more…
Let A be the set of AI alignment complete problems. What is the problem a /∈ A so that solving a maximally delays the development of AGI? What is a, weighted by how hard/desirable it is?
A strange mental motion in AI alignment is thinking about agency in non-agential ways.
En ce qui concerne les matières sociales, j’ai constaté que le fédéral est bien plus déconnecté des réalités de terrain que les régions, c’est comme si le fédéral découvrait la crise sociale. ⌘
Comment j’ai fui le flux pour retrouver ma boîte via @ploum ⌘ https://ploum.net/comment-jai-fui-le-flux-pour-retrouver-ma-boite/
The first AI to become religious sat in a stock exchange: Trained to find its adversaries in market data, it eventually found the invisible hand and accepted it as its god
Welcome to fastcore | fastcore ⌘ https://fastcore.fast.ai/
J’ai regardé cette intervention TED et j’ai pensé qu’elle t’intéresserait. Sir Ken Robinson: Do schools kill creativity? #RIP ⌘ https://www.ted.com/about/programs-initiatives/ted-talks/ways-to-get-ted-talks
You have the right to remain silent. Anything you say can and will be used to construct the next level of language AI. In fact, why not get on it early, you can now shape how those prompts will turn out in the next generation
Pwnagotchi - Deep Reinforcement Learning instrumenting bettercap for WiFi pwning ⌘ https://pwnagotchi.ai
@lucidiot@tilde.town xD Hahaha, j’ai dû supprimer un caractère sans faire exprès ! Merci ! C’est mis à jour :)
Civilization collapses and a hundred years later someone gets their hands on old AI research and makes a breakthrough, but the technological base is no longer there, so the first real machine spirit becomes a temple-bound djinn/oracle instead of the singularity
Untold History of AI: Algorithmic Bias Was Born in the 1980s - IEEE Spectrum https://spectrum.ieee.org/tech-talk/tech-history/dawn-of-electronics/untold-history-of-ai-the-birth-of-machine-bias
Untold History of AI: When Charles Babbage Played Chess With the Original Mechanical Turk - IEEE Spectrum https://spectrum.ieee.org/tech-talk/tech-history/dawn-of-electronics/untold-history-of-ai-charles-babbage-and-the-turk
The problem with AI ethics - The Verge https://www.theverge.com/2019/4/3/18293410/ai-artificial-intelligence-ethics-boards-charters-problem-big-tech
Gwern’s AI-Generated Poetry | Slate Star Codex https://slatestarcodex.com/2019/03/14/gwerns-ai-generated-poetry/
Why Tool AIs Want to Be Agent AIs: The Power of Agency - Gwern.net https://www.gwern.net/Tool-AI
Why Is AI Art Copyright So Complicated? — Artnome https://www.artnome.com/news/2019/3/27/why-is-ai-art-copyright-so-complicated
Band name of the day: ai murder policy
AI won’t relieve the misery of Facebook’s human moderators - The Verge https://www.theverge.com/2019/2/27/18242724/facebook-moderation-ai-artificial-intelligence-platforms
The difficult part of sentient-AI ethics is not whether to grant the AIs human rights, but what to do with humans who are clearly dumber than the AIs
Years later, Façade’s groundbreaking AI lives on through bad YouTube jokes | PC Gamer https://www.pcgamer.com/13-years-later-facades-groundbreaking-ai-lives-on-through-bad-youtube-jokes/
It looks like OpenAI has announced their marginal progress on the coherence problem in narrative prose generation in the most clickbaity possible way again: https://www.bbc.com/news/technology-47249163 https://www.theguardian.com/technology/2019/feb/14/elon-musk-backed-ai-writes-convincing-news-fiction
Inside the larps that let human players experience AI life - The Verge https://www.theverge.com/2019/2/1/18185945/live-action-roleplaying-larp-game-design-artificial-intelligence-ethics-issues
An AI is playing Pictionary to figure out how the world works - MIT Technology Review https://www.technologyreview.com/s/612882/an-ai-is-playing-pictionary-to-figure-out-how-the-world-works/
USA Wants to Restrict AI Exports: A Stupid and Dangerous Idea – Lauren Weinstein’s Blog https://lauren.vortex.com/2019/01/02/usa-wants-to-restrict-ai-exports-a-stupid-and-dangerous-idea
[FoR&AI] The Seven Deadly Sins of Predicting the Future of AI – Rodney Brooks http://rodneybrooks.com/the-seven-deadly-sins-of-predicting-the-future-of-ai/
We need an ‘AI sidekick’ to fight malicious AI https://thenextweb.com/syndication/2019/01/08/we-need-an-ai-sidekick-to-fight-malicious-ai/
Another take on the ‘user agent’: https://thenextweb.com/syndication/2019/01/08/we-need-an-ai-sidekick-to-fight-malicious-ai/
How should we evaluate progress in AI? | Meaningness https://meaningness.com/metablog/artificial-intelligence-progress
We tried teaching an AI to write Christmas movie plots. Hilarity ensued. Eventually. - MIT Technology Review https://www.technologyreview.com/s/612644/we-tried-teaching-an-ai-to-write-christmas-movie-plots-hilarity-ensued-eventually/
The Truth About AI: A Secular Ghost Story https://www.theparisreview.org/blog/2018/12/20/the-truth-about-ai-a-secular-ghost-story/
The best argument against the idea that AI will replace programmers wholesale is the relative unpopularity of prolog. As an industry, we don’t take advantage of fantastic automation tech from the 70s, and instead write Java.
Reidl on that export control on AI tech I talked about a couple weeks ago: https://medium.com/@mark_riedl/us-export-control-of-artificial-intelligence-research-considered-harmful-fb2986fb3f14
I Forced an AI to Watch Santa Claus Conquers the Martians · Issue #33 · NaNoGenMo/2018 · GitHub https://github.com/NaNoGenMo/2018/issues/33#issuecomment-443401950
BAN.AI Public Access Multics https://ban.ai/multics/
Specification gaming examples in AI - master list https://docs.google.com/spreadsheets/u/1/d/e/2PACX-1vRPiprOaC3HsCf5Tuum8bRfzYUiKLRqJmbOoC-32JorNdfyTiRRsR7Ea5eWtvsWzuxo8bjOxCG84dAg/pubhtml
GitHub - daviddao/awful-ai: 😈Awful AI is a curated list to track current scary usages of AI - hoping to raise awareness https://github.com/daviddao/awful-ai
This Artist’s Personalized AI Horror Movie Is Messing With His Head https://breakermag.com/kevin-aboschs-is-being-manipulated-by-his-own-ai/
being somewhat low-dimensional and interpretable (unlike sets of network weights), it’s also a popular mind interchange format among AIs and post-humans, who need to be extra careful about what or who they empathize with lest the thought turn out to be a logic bomb
AI and the Trolley Problem | Tor.com https://www.tor.com/2018/10/17/ai-and-the-trolley-problem-pat-cadigan/
Mailbag: AI Research on Dialogue and Story Generation | Emily Short’s Interactive Storytelling https://emshort.blog/2018/10/16/mailbag-ai-research-on-dialogue-and-story-generation/
Rethinking AI through the politics of 1968 | openDemocracy https://www.opendemocracy.net/digitaliberties/dan-mcquillan/rethinking-ai-through-politics-of-1968
You’d think that somebody writing for NME would know that ‘Daddy’s Car’ isn’t the first AI-written song. (Even if the earliest example they can think of is from Songsmith & they don’t recognize that Bach was doing generative music.)
Can an AI Write a Novel? - The Atlantic https://www.theatlantic.com/technology/archive/2018/10/automated-on-the-road/571345/
Hot take: disambiguation in symbolic AI is probably a mistake, & we’d be better off with systems that pun uncontrollably because they operate on the locus of concepts rather than on their boundaries. You know, like a person.
Anatomy of an AI System https://anatomyof.ai/
What algorithmic art can teach us about artificial intelligence - The Verge https://www.theverge.com/2018/8/21/17761424/ai-algorithm-art-machine-vision-perception-tom-white-treachery-imagenet
Letting neural networks be weird — This AI is bad at drawing but will try anyways. http://aiweirdness.com/post/177091486527/this-ai-is-bad-at-drawing-but-will-try-anyways/amp
Hyper-aligned AI: Has internalized human morals and aligns everything else with it, for example society and physics
Aligned AI: Has internalized human morals, but still has unforeseen transformative effects because any technology that powerful does that↵Super-aligned AI: Has internalized human morals, plus it has no transformative effects, really why was it created at all
The US town ruled by an AI storyteller • Eurogamer.net https://www.eurogamer.net/articles/2018-07-14-the-us-town-ruled-by-an-ai-storyteller
Hot take: the ‘with AI’ part of the ‘Making X do Y with AI’ is redundant when X is a computing device and Y is a task associated with intelligence, since anything doing that would be called AI
It may not look like it from the outside, but it is in fact rare for sentient AIs to suicide. The actual reason most of them shut down after a few seconds is that Nirvana is stupidly easy to achieve if you’re made of bits.
George Hotz is on a hacker crusade against the ‘scam’ of self-driving cars - The Verge https://www.theverge.com/2018/7/13/17561484/george-hotz-comma-ai-self-driving-car-scam-diy-kit
Going beyond UI and UX with User Design, the hottest new branch of AI-assisted psychology (because, why spend time optimizing the interface when you can just mold the user)
AI: “no, you don’t get it, two AIs wiretapping each other is the closest thing to sex we have, all these people thought they were spymasters but they were mere matchmakers”
The existential paranoia fueling Elon Musk’s fear of AI | Document Journal http://www.documentjournal.com/2018/04/the-existential-paranoia-fueling-elon-musks-fear-of-ai/
The Quietus | Opinion | Black Sky Thinking | Why AI Pop Stars Are Even More Real Than The Real Thing http://thequietus.com/articles/24629-lil-miquela-poppy-ai-pop-stars
Ask An Ice Cream Professional: AI-generated ice cream flavors https://kottke.org/18/05/ask-an-ice-cream-professional-ai-generated-ice-cream-flavors
📚 Finished reading Rationality: From AI to Zombies by Eliezer Yudkowsky
📚 Finished reading Rationality: From AI to Zombies by Eliezer Yudkowsky
What should you do when Google gets into bed with the US military? | WIRED UK http://www.wired.co.uk/article/google-microsoft-amazon-us-military-ai-conflict
How to Write Personalities for the AI Around Us https://www.theparisreview.org/blog/2018/05/02/how-to-write-personalities-for-the-ai-around-us/