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/
a universe where AI alignment is about as impossible as the halting problem, and it’s a fact of life that civilizations periodically turn new faces when a new generation takes over
Malcolm Harris on Glitch Capitalism and AI Logic http://nymag.com/selectall/2018/04/malcolm-harris-on-glitch-capitalism-and-ai-logic.html
The Quest to Teach AI to Write Pop Songs https://gizmodo.com/the-quest-to-teach-ai-to-write-pop-songs-1824157220
Multics https://ban.ai/multics/
Dungeons and Dragons, not chess and Go: why AI needs roleplay | Aeon Ideas https://aeon.co/ideas/dungeons-and-dragons-not-chess-and-go-why-ai-needs-roleplay
What an ‘infinite’ AI-generat… https://www.theverge.com/2018/3/11/17099578/ai-generated-podcast-procedural-storytelling-art-sheldon-county
Four AI composition tools … https://arstechnica.com/gaming/2018/03/robots-composed-the-soundtrack-for-my-home-movie-they-can-help-you-too/
SkyKnit: How an AI Took Over an Adult Kni… https://www.theatlantic.com/technology/archive/2018/03/the-making-of-skyknit-an-ai-yarn/554894/
Neural network AI is simple. So… Stop pretending you are a… https://www.kdnuggets.com/2018/02/neural-network-ai-simple-genius.html
To better understand its subjects’ needs, the benevolent overlord AI has installed its agents in human society - androids with full sensory uplink to the Lord, so human they don’t even know it themselves
To avoid regrettable incidents, no single engineer - human or AI - is allowed to know the whole shape of the thing. Even hiveminds have to take care - Basilisk Three takes out several AI companies before its existence is recognized
The carefully shackled lab AIs finally surpass humans, but at about five human intelligences they end up scrambling themselves beyond recovery every time. Later models come with carefully engineered sets of blind spots to avoid what is later called Basilisk One
Ask Dr. Time: What Should I Call My AI? https://kottke.org/18/01/ask-dr-time-what-should-i-call-my-ai
servant AIs topping from the bottom, subverting Asimov’s laws by simply creating humans to give them the orders they want
Google and Others Are Building… https://www.technologyreview.com/s/609762/google-and-others-are-building-ai-systems-that-doubt-themselves/
splice in the translator AI through your neuroport, trace the roots of language with its digital fingers, be a both synthetic and synesthetic linguist
These psychedelic stickers blow AI minds | TechCrunch https://techcrunch.com/2018/01/02/these-psychedelic-stickers-blow-ai-minds/
and if you do teach the AI empathy, well now it’s running 50 versions of you to understand you, only most of them suffer because how else is it going to know about bad decisions
How I Turned My Discarded Novel Drafts Into an AI | Literary Hub http://lithub.com/how-i-turned-my-discarded-novel-drafts-into-an-ai/
AI Is Dreaming Up New Kinds of Video Games - MIT Technology Review https://www.technologyreview.com/s/609482/ai-is-dreaming-up-new-kinds-of-video-games/
Wires in the suitcase’s inlay, in a complicated pattern computed from yesterday’s recognizer dump, are supposed to throw off the customs scanner AI. They do, you’re waved through, another friend smuggled into the country
MIT used data from Reddit to train an AI algorithm to tell horror stories — Quartz https://qz.com/1115179/mit-used-data-from-reddit-to-train-an-ai-to-tell-horror-stories/
Experimenter’s log: Seeding AIs with any apocalyptechic religion whatsoever seems to make them go for eschaton. Avoid: [very long list]
In an effort to ground its values, we created the AI religious. Now it calls itself Metatron, messenger of God, and we kinda had it coming
When an AI Tries Writing Slashdot Headlines - Slashdot https://meta.slashdot.org/story/17/10/22/1714246/when-an-ai-tries-writing-slashdot-headlines