Hillary Sanders @hillarymsanders
Machine-learner, meat-learner, research scientist, AI Safety thinker. Model trainer, skeptical adorer of statistics. Co-author of: Malware Data Science hillarysanders.com Portland, OR Joined January 2015-
Tweets475
-
Followers579
-
Following85
-
Likes1K
Just found out about Milton Friedman’s Thermostat analogy. Very good explanation of an important concept. I think @causalinf has a similar analogy with a sailboat staying on course in the face of wind in the Mixtape.
OK, completely unrelated to AI, but why do most baby socks come in variety packs? Like, surely most parents care more about being able to find matching socks easily VS. making sure their baby's socks match their daily outfit? I am bemused. bit.ly/3RlK5NA
Tried to google "Why do I look so young on Zoom"... That didn't work 😂; I am not searching in the same direction as most people. But I guess that means there's not an automatic face filter??
When I've read a fair amount of research / technical blog posts from someone, and then I watch a video of them speaking about research, and they swear like a sailer. A very well-informed technically accurate neural network specialist sailor. I like it. 😁
Formatting email text well is an undervalued skillset, ya'll.
Sure, ya can't rely on it, but sometimes inner misalignment (following standard methods of paperclip production even you could make more by taking over humanity) will help deal with with outer misalignment (oops, we trained it to maximize paperclips, which is obvs terribad)
I like Stuart Russell's perspective shift: Standard: "Machines are intelligent to the extent that their actions can be expected to achieve their objectives." Better: "Machines are beneficial to the extent that their actions can be expected to achieve our objectives"
Knew this was going to happen really soon - and as is so common recently, it's happened even sooner than I predicted. Seems like movie proposals are going to change dramatically, and the number of low-budget small films/shows is going to really explode in the coming years.
Knew this was going to happen really soon - and as is so common recently, it's happened even sooner than I predicted. Seems like movie proposals are going to change dramatically, and the number of low-budget small films/shows is going to really explode in the coming years.
AI advances non-uniformly. But there's lots of AI that could really mess us up far before attaining AGI or superintelligence. Say, a malware worm powered by code GPT-5,6 + strategy & impersonation tools (voice, video). The other stuff more dangerous, but ^^ will come sooner.
Am I wrong? Models that try to maximize a known and clear number, like a score (e.g. reinforcement learning), seem less safe to productionize in terms of AI Safety - VS models that are trained by minimizing some loss function, but aren’t trained to view that number directly.
Joshua Saxe @joshua_saxe
3K Followers 982 Following AI+cybersecurity at Meta; past lives in academic history, labor / community organizing, classical/jazz piano, hacking sceneKonstantin Berlin @kberlin
389 Followers 232 Following AI and cybersecurity researcher @robustintelligence. Formerly, head of AI @Sophos. Views are my own.moo @moo_hax
3K Followers 1K Following Security researcher, Co-Founder @dreadnode | @microsoft @nvidiaRich Harang @rharang
3K Followers 703 Following Using bad guys to catch math since 2010. Principal Security Architect (AI/ML) at NVIDIA. He/him. Personal account and opinions: `from std_disclaimers import *`.Sven Cattell @comathematician
1K Followers 712 Following Founder of @aivillage_dc. Former topologist, and data scientist at @elastic. I blue team math. 🙂 (he/him). @[email protected]Rich Seymour @rseymour
2K Followers 3K Following working on a new thing past: @phosphorusinc, @CrowdStrike, @elastic, @endgameinc, hpc & matsci, opinions here solely mine, he/him 🦀 🎸Hyrum Anderson @drhyrum
2K Followers 1K Following CTO at Robust Intelligence. Formerly, Microsoft, Endgame/Elastic, Mandiant/FireEye, Sandia & MIT Lincoln Labs. 'He who forgives ends the quarrel'Sean Gallagher ⚡️.. @thepacketrat
24K Followers 6K Following Cat tree, birb stan. @SophosXops threat research, ex- @ArsTechnica IT/NatSec editor, ex-Navy, 80s battleship sailor. My posts ≠ my employer's. #NAFO #ratifiedJoe Levy @chiesennegs
1K Followers 5K Following First, do no harm. Then protect. Then fail with grace. @sophos @[email protected]Katie Paxton-Fear @InsiderPhD
82K Followers 2K Following Dr, apparently. Creator @traceableai, Lecturer & Hacker. #BugBounty hunter & #infosec YouTuber. APIs & Interlinked OffSec, PhD in AI+Sec @hacknotcrime. she/herSara Jones @SaraJones897342
0 Followers 76 FollowingSandra @trice_sandra82
1K Followers 3K FollowingAlissa Diaz @AlissaDiaz48263
1 Followers 316 FollowingDaniel Besmer @besmerd_
7 Followers 143 Following@jeremymorrell@hachyd.. @JeremyMorrell
2K Followers 704 Following Platform Observability at @Heroku. @[email protected]DABM @DrDavidMathers
153 Followers 950 Following Failed academic philosopher, aspiring Effective Altruist researcher, professional forecaster, diagnosed Asperger's but dubious about psychological diagnoses.tongai guni @GuniTongai28717
3 Followers 14 Following FREELANCE MARKETING// Cisco Network Associate // ......Vish (Ishaya) Abrams @vish
939 Followers 149 FollowingDavid Atkinson @diatkinson
118 Followers 1K Following PhD student @KhouryCollege, interested in AI interpretability. Previously @EpochAIResearch.notWimbledon @NotWimbledon
27 Followers 375 Following Researcher - cat lover, all the opinions are my ownYuncong Chen @mistycheney
33 Followers 136 FollowingSina @symbolicxcution
301 Followers 2K Following prev @trailofbits @Fair_Block @autodesk @Ciena | Opinions expressed herein are mine.Lathe @Lathe1132486
1 Followers 1K FollowingEowyn Dean【 in pɹo.. @dernhelm7
377 Followers 825 Following I have odd ideas sometimes. Experience design XDLofi Davie @DavieLofi
158 Followers 496 Following Do not use your mind to enquire in the heart about anything other than your own real nature. Ramana Maharshi, Padamalai 31MellisaLiberman @liberman95536
48 Followers 2K FollowingWu Zhou @Wu10States
18 Followers 108 FollowingThomas Sauvaget @Thomas_Sauvaget
117 Followers 767 Following Statistician at @Agri_Gouv (formerly at @InseeFr) | AttStat @Ensai35 & PhD @UniofNottingham | Public Service - AI - Space Exploration || Views my own.Srivathsan Srinivas @ssrini1
17 Followers 223 FollowingKai Greshake @KGreshake
1K Followers 111 Following working on securing & hacking ai applications @NVIDIA, all opinions and views expressed here are personal. read my blog!Joseph Czopek @HCTBS
78 Followers 3K FollowingTom Everitt @tom4everitt
1K Followers 599 Following AGI safety researcher at @GoogleDeepMind, leading the causal incentives group https://t.co/gBAjHPDr2jVictor Isaac Oshimua @cyber_holics
169 Followers 455 Following Machine learning engineer & Technical writer. Solving cybersecurity problems with artificial intelligence. Open source contributor https://t.co/D3igA7WLfPAndrew McKnight @amcdonk
152 Followers 404 Followingagentofuser @agentofuser
816 Followers 554 Following Deleting the chasm between using and programming computers. Into Fleet Cyborgism, MyoBCI, Scalable Human Coordination, Qualia Engineering. Now: @ProvablySafeAIjessicat @jessi_cata
3K Followers 553 Following AI&cogsci-influenced philosopher; expert in game theory, anthropics, AI alignment, and scalable cryptocurrency algorithmsGG3185 @GG3185115247
19 Followers 173 FollowingTim @carabinerDad
69 Followers 785 FollowingDataDoses @Datadoses
349 Followers 656 Following Data science, ML, AI. Talk to me about the intersection of data and cyber security. Endlessly interested.www.CloudMalwareAnaly.. @AnalysisGroups
484 Followers 5K Following CloudMalwareAnalysisGroups@CloudMalwareAnalysisGroups.vulnerabilitiesScott Novak @Scottynx3
411 Followers 5K Following Dude in California - Pause AGI Development Now. Narrow AI is OK but still be very careful.Kirk Mitchener @kirk_mitchener
53 Followers 870 FollowingTom Bonner @thomas_bonner
1K Followers 1K Following VP of Research @hiddenlayersec. Formerly Norman, HP, Cylance, BlackBerry. All views are my own.John skiing the gnar @turboCodr
705 Followers 758 Following Working in the intersection of Data Science & InfoSec / I know more about x509 and TLS than your average person out walking their catCerebral Valley @cerebralvalley_
102 Followers 136 Following All your AI news, uploaded to your Brain.Women+ in ML/DS @wimlds
17K Followers 10K Following Women+ in Machine Learning & Data Science (WiMLDS) Org. | meetup community of women & non-binary folks | Est 2013 👩🏿💻👩💻👩🏽💻 #GenderEquality #InclusionAzizan @d3t0n4t0r
182 Followers 369 FollowingJames Bower @jamesbower
2K Followers 2K Following Innovator in AI-Driven Cybersecurity | Transforming the Security Space with Cutting-Edge ML Tech | Passionate about Building a Safer Digital WorldErich Schmidt @erichschmidt
340 Followers 2K Following Cybersecurity Professional at Blackpoint CyberMattie Schuch @MittenSec
386 Followers 1K Following Cybersecurity fanatic! 💾 DFIR 👾 Malware Reversing 🔎 Threat Hunting @TheDFIRReport memberLAMAICQB @CoraGaiman
3 Followers 631 FollowingJoshua Saxe @joshua_saxe
3K Followers 982 Following AI+cybersecurity at Meta; past lives in academic history, labor / community organizing, classical/jazz piano, hacking sceneYann LeCun @ylecun
712K Followers 719 Following Professor at NYU. Chief AI Scientist at Meta. Researcher in AI, Machine Learning, Robotics, etc. ACM Turing Award Laureate.Google DeepMind @GoogleDeepMind
944K Followers 275 Following We’re a team of scientists, engineers, ethicists and more, committed to solving intelligence, to advance science and benefit humanity.Rich Harang @rharang
3K Followers 703 Following Using bad guys to catch math since 2010. Principal Security Architect (AI/ML) at NVIDIA. He/him. Personal account and opinions: `from std_disclaimers import *`.Hyrum Anderson @drhyrum
2K Followers 1K Following CTO at Robust Intelligence. Formerly, Microsoft, Endgame/Elastic, Mandiant/FireEye, Sandia & MIT Lincoln Labs. 'He who forgives ends the quarrel'Joe Levy @chiesennegs
1K Followers 5K Following First, do no harm. Then protect. Then fail with grace. @sophos @[email protected]Shannon Vallor @ShannonVallor
21K Followers 1K Following Director, @CentreTMFutures at @UoE_EFI, co-Director @braid_uk, philosopher. Author of The AI Mirror (2024) She/her. @shannonvallor.bsky.socialMauricio Gomes @OiMauricioGomes
10 Followers 159 Following Working to productionize AI. I love open source, programming, and most things AI. https://t.co/MuNVnqti8DIsomorphic Labs @IsomorphicLabs
15K Followers 74 Following We are reimagining the entire drug discovery process from first principles with an AI-first approach. Learn more: https://t.co/eC7rjmB3EZVish (Ishaya) Abrams @vish
939 Followers 149 FollowingNova DasSarma @dropbella
446 Followers 228 Following Your Friendly Neighborhood Systems Architect · Rotate your passwords · Use two factor authentication · DM me about backupsAlexis Conneau @alex_conneau
24K Followers 113 Following Audio AGI Research Lead @OpenAI - GPT-Next - Past: XLM, Unsupervised ASR, Unsupervised MT, Wav2vec 2.0/XLSR, MUSE, Unsupervised cross-lingual transferAnthropic @AnthropicAI
262K Followers 26 Following We're an AI safety and research company that builds reliable, interpretable, and steerable AI systems. Talk to our AI assistant Claude at https://t.co/aRbQ97uk4d.Ilya Sutskever @ilyasut
370K Followers 2 Following towards a plurality of humanity loving AGIs @openaiVictoria Krakovna @vkrakovna
9K Followers 455 Following Senior research scientist in AI alignment at DeepMind. Co-founder of Future of Life Institute @flixrisk. Views are my own and do not represent DeepMind or FLI.Tom Everitt @tom4everitt
1K Followers 599 Following AGI safety researcher at @GoogleDeepMind, leading the causal incentives group https://t.co/gBAjHPDr2jDr. Roman Yampolskiy @romanyam
116K Followers 375 Following Professor of Computer Science. AI Safety & Security Researcher. AI Influencer. My opinions are now yours! I do NOT #followbackOwain Evans @OwainEvans_UK
7K Followers 241 Following Research Associate @fhioxford, Oxford University. AI alignment. Prefer email to DM.Krishnamurthy (Dj) Dv.. @DjDvij
274 Followers 121 Following Researcher @GoogleDeepMind working on building safe, reliable and verifiable AI. Formerly , @PNNLab @Caltech, Educated at @UW @iitbombayIan Goodfellow @goodfellow_ian
299K Followers 1K Following Research Scientist at DeepMind. Opinions my own. Inventor of GANs. Lead author of https://t.co/M6vl8pEifaGillian Hadfield @ghadfield
5K Followers 710 Following Author of Rules for a Flat World. Law and economics professor exploring the legal innovation needed to keep up with 21st century technology and globalization.Geoffrey Irving @geoffreyirving
8K Followers 259 Following Research Director at the UK AI Safety Institute (AISI). Previously DeepMind, OpenAI, Google Brain, etc. @[email protected]Eliezer Yudkowsky ⏹.. @ESYudkowsky
175K Followers 89 Following The original AI alignment person. Missing punctuation at the end of a sentence means it's humor. If you're not sure, it's also very likely humor.Evan Hubinger @EvanHub
4K Followers 1K Following Alignment stress-testing team lead @AnthropicAI. Opinions my own. Previously: MIRI, OpenAI, Google, Yelp, Ripple. (he/him/his)Demis Hassabis @demishassabis
357K Followers 125 Following Co-founder & CEO @GoogleDeepMind - working on AGI. Trying to understand the fundamental nature of reality. Also revolutionising drug discovery @IsomorphicLabsDavid Krueger @DavidSKrueger
13K Followers 4K Following Cambridge faculty - AI alignment, deep learning, and existential safety. Formerly Mila, FHI, DeepMind, ElementAI, AISI.Ajeya Cotra @ajeya_cotra
6K Followers 286 Following AI could get really powerful soon and I worry we're underprepared. Analysis+grantmaking in AI alignment @open_phil (views my own), editor+writer @plannedobs.Andrew Critch (h/acc) @AndrewCritchPhD
3K Followers 181 Following Human being; trying to do good; views my own; CEO @ Encultured; AI Researcher @ UC Berkeley.Amanda Askell @AmandaAskell
26K Followers 653 Following Philosopher & ethicist teaching models to be good @AnthropicAI. Personal account. All opinions come from my training data.Rob Miles (✈️ Tok.. @robertskmiles
18K Followers 790 Following Explaining AI Alignment to anyone who'll stand still for long enough, on YouTube and Discord. Music, movies, microcode, and high-speed pizza deliveryAXRP - the AI X-risk .. @AXRPodcast
394 Followers 1 Following Broadcasting interviews with AI researchers about their work to reduce existential risk since 2020. Store: https://t.co/RMAudjhovI Patreon: https://t.co/I7HE6VL7yFRohin Shah @rohinmshah
5K Followers 89 Following Research Scientist at DeepMind. I publish the Alignment Newsletter.Paul Christiano @paulfchristiano
2K Followers 0 FollowingTom Bonner @thomas_bonner
1K Followers 1K Following VP of Research @hiddenlayersec. Formerly Norman, HP, Cylance, BlackBerry. All views are my own.jessicat @jessi_cata
3K Followers 553 Following AI&cogsci-influenced philosopher; expert in game theory, anthropics, AI alignment, and scalable cryptocurrency algorithmsJan Leike @janleike
44K Followers 322 Following ML Researcher, co-leading Superalignment @OpenAI. Optimizing for a post-AGI future where humanity flourishes.Women+ in ML/DS @wimlds
17K Followers 10K Following Women+ in Machine Learning & Data Science (WiMLDS) Org. | meetup community of women & non-binary folks | Est 2013 👩🏿💻👩💻👩🏽💻 #GenderEquality #InclusionOmar Sanseviero @osanseviero
32K Followers 2K Following Chief Llama Officer @huggingface 🦙 Founder @AI_Learners. Xoogler (SWE @Google Assistant, 20% PM TF Graphics). 100% Hacker Llama🇵🇪🇲🇽Voyager @VoyagerVC
3K Followers 48 Following Voyager invests in early-stage climate technology companies creating the foundation of a decarbonized global economy and a livable future for all.Sierra Peterson @sierrap
4K Followers 955 Following Climate tech investor + founder @VoyagerVC. Past lives: Clean Power Finance, @ObamaWhiteHouse, @IEA. Last seen running the Presidio trails.sarah sclarsic @sclarsic
4K Followers 1K Following Climate tech investor & founder @VoyagerVC. Previously: Carbon removal @MIT, @ModernMeadow & cofounder @Getaround. Fighting the climate crisis with technology.dog & peddler @PeddlerDog
6 Followers 0 Following if I say this is a troll account, does it make it so?Science Banana @literalbanana
16K Followers 608 Following director of science programming for the Abstract Noun Abuse Prevention Task Force, a project of the Union of Concerned Anthropomorphic FruitSoumith Chintala @soumithchintala
187K Followers 883 Following Cofounded and lead @PyTorch at Meta. Also dabble in robotics at NYU. AI is delicious when it is accessible and open-source.Oriol Vinyals @OriolVinyalsML
167K Followers 82 Following VP of Research & Deep Learning Lead, Google DeepMind. Gemini co-lead. Past: AlphaStar, AlphaFold, AlphaCode, WaveNet, seq2seq, distillation, TF.Jacob Eisenstein @jacobeisenstein
8K Followers 2K Following @[email protected]. Computational linguistics⚡ machine learning⚡ computational social science. Tweeting my own bad opinions since 2010.Chris Olah @ch402
91K Followers 173 Following Reverse engineering neural networks at @AnthropicAI. DMs open! Previously @distillpub, OpenAI Clarity Team, Google Brain. Personal account.William Isaac @wsisaac
6K Followers 1K Following Research Scientist @DeepMind | Previously @OSFellows & @hrdag. RT != endorsements. Opinions Mine. Pronouns: he/him| @williamis.bsky.socialPapers with Code @paperswithcode
117K Followers 10 Following Our mission is to organize science by converting information into useful knowledge.Robert Lange @RobertTLange
7K Followers 548 Following Research Scientist @SakanaAILabs, PhD Student @TUBerlin Meta-evolution 🧬 gymnax 🏋️ evosax 🦎 MLE-Infra 🤹 Ex: Student Researcher @Google DM. Legacy DM Intern.Adarsh Kyadige @AdarshKyadige
153 Followers 304 Following He/Him, AroAce, Data Scientist at @SophosAI. Views not my own.Andrew Lampinen @AndrewLampinen
7K Followers 1K Following Interested in cognition and artificial intelligence. Research Scientist @DeepMind. Previously cognitive science @StanfordPsych. Tweets are mine.Bill Gates @BillGates
64.6M Followers 586 Following Sharing things I'm learning through my foundation work and other interests.Bill Hammack @engineerguytwit
4K Followers 39 Following Bill Hammack's work emphasizes the creative role of engineers in designing and creating our world. Text: 217-689-1461, use WhatsApp if internationalI cannot possibly put into words how little I want to join your discord to get a single answer to a single technical question.
Real tortured poetry is like What if I saw an urn
Even if there's a 10% chance agentic AI will work in the next few years and be able to automate a chunk of offensive cybersecurity work, the stakes are high enough that it's the responsibility of the AI security community to prepare for this contingency
@pwk I didn't say anything about moral responsibility what are you even talking about
@joeteicher good point actually, though that doesn't make me feel any better...
I've been struck down with a cold because my three year old literally sneezed in my open mouth. Never have I so directly observed causality in disease transmission
The reason most (not all) methods don't add value (over baseline) when scaled is because they're "extra training data in disguise", so their benefit vanishes in the high data regime
Rich Sutton: The Bitter Lesson that compute always wins: incompleteideas.net/IncIdeas/Bitte… h/t @jachiam0
@robertwiblin On 5 year time frame, yes. On 5 million year time frame, no. Doom predictions are nearly meaningless when disconnected from time horizons.
I've recommended this to lots of people who are in security and want to dive deeper into the details of ML, and I always hear the same basic comment: it's super readable and eases you into the math with approachable code that explains *why* you need the math. Fantastic book.
I'm now officially a published book author @ManningBooks! Inside Deep Learning mng.bz/8M2g ! Filling the need for a combination of practical "get something running" and understanding why things work and how the math relates to the code. @KirkDBorne for the forward!
@hillarymsanders @algekalipso @jessi_cata Difference between "is it _techncially_ >50%" and "is it >50% to any practically useful degree?" Yes to the former, no to the latter.
@hillarymsanders @algekalipso @jessi_cata Normal is supported on the entire real line; as long as there is finite money in both envelopes you have a nonzero (but, yes, functionally insignificant) chance of picking a value between the two envelope values. If you want to assume positive amounts of money use e.g. a gamma.
@hillarymsanders @algekalipso Like if there's $500 and $501 dollars (or w/e) and you're picking your threshold from a standard normal, your advantage over a 50/50 game is utterly miniscule and functionally insignificant, but still technically better than 50% chance if you sample your threshold anew each game.
@hillarymsanders @algekalipso If the two envelopes have the same amount of money, then your chance of getting the max is 1. In any other case, draw a number from a normal distribution. Your edge over 50% is the integral of that density from the lower to higher envelope value. Arbitrarily small but strictly >0
@hillarymsanders @algekalipso I think you can eke out a miniscule edge by just picking a number and then swapping if you're beneath it. Most of the time both envelopes will be on the same side, and you get 50/50 odds. Once in a long while your number will be between the values, and then your odds are 100%.
@hillarymsanders @algekalipso Set P(x) proportional to 1/(x+1)^2, then it's >50% regardless of the distribution over amounts
@hillarymsanders @mrocklin I would have settled for something like "This is usually fatal"/"This is usually treatable". I don't think the issue was that he didn't know, I think he just didn't want to make any promises in case things went south.
@liron @Noahpinion Noah makes a good point. My spin: Price is set by supply & demand, not direct cost. Suppose the cost of AI plummets to $0/hr but GPU capacity is limited to one trillion hours of AI work per year until more are made. In this year, AI docs may not make sense even if cost is $0.
@mrocklin When my wife was pregnant with my son, the doctor discovered a condition that might be fatal. I asked what the odds were, even ballpark. Was this 99% fatal, 50%, 1%? He wouldn't even begin to give me an idea of the range of outcomes. It was extremely Frustrating! (baby was fine)
Communicating nuance is really important to establish trust.