TurnpikeLad @muddubeeda
I should be working, but am reading about generative AI obsessively whenever I have a moment to myself. In the hills east of Seattle Joined October 2014-
Tweets419
-
Followers30
-
Following91
-
Likes381
Fascinating how the king's back legs seem to be made of cloth, and are left behind when the king leaves the throne. The legs work as a still image, but not as video. Reminds me of the continual "this makes/doesn't make sense" authorial oversight in dreams x.com/billpeeb/statu…
Fascinating how the king's back legs seem to be made of cloth, and are left behind when the king leaves the throne. The legs work as a still image, but not as video. Reminds me of the continual "this makes/doesn't make sense" authorial oversight in dreams x.com/billpeeb/statu…
noodling around on Erana's Peace youtube.com/watch?v=gc9p0i…
Something about the original Hero's Quest / Quest for Glory 1 goblin base theme is compelling to me. I've gotta figure out a way to elaborate on it. youtu.be/upC1WORMIl4?si…
Leaves are staying mostly green on many trees even into November this year. Why is this? This year doesn't seem warmer/drier than normal (at least in the Seattle area.) Pic taken today, Nov 2
White River flowing past me at Skookum Falls Viewpoint, SR 410, just outside of rainier national park
Noodling on the piano while others work to prepare dinner
The old treehouse. Around 30 years ago, my uncle had a fort in this tree. You can still see a board that formed a wall and a rope that was tied around a branch. Scars are left when limbs die or fall away from the main trunk.
Interesting variety of nettles in the woods yesterday. Many stalks fallen over or withering, but with small healthy young leaves
Incredible skies on 405 through Renton yesterday evening. Autumn will be here soon.
Shoasmo @ShoasmoBfud
5 Followers 262 FollowingCroughsm @Croughsm198499
1 Followers 407 FollowingDanaChapman @25jd5811hrc2et
0 Followers 329 FollowingArjun Prakash @arjuntheprakash
41 Followers 91 FollowingMatthew Gleason @matt0sai
152 Followers 614 Following lifelong learner, metal enthusiast, agi ambassador, space hypothesizerPattiGagen @GagenPatti81344
120 Followers 2K Following0x44 0x48 @0x440x46
994 Followers 2K Following ⇔ ≡ ¬∃ ⇔ ≡ ¬∃ ⇔ ≡ :: don't worry, I explore coordinates in thought space so you don't have to.Daffodil @Daffodil426361
26 Followers 436 FollowingCeshesm @ceshesm64727
120 Followers 3K FollowingLoathey @loathey16746
79 Followers 548 FollowingTayllou @Tayllou120302
9 Followers 1K Following The fragrance of rice flowers tells of a good harvest, and the sound of frogs sounds.Karl Widerquist @KarlWiderquist
4K Followers 3K Following Philosopher & economist focused on distributive justice. Writes about Basic Income; the ethics of poverty, freedom, & inequality; & whatever else comes to mind.Noa Nabeshima @NoaNabeshima
319 Followers 841 Following DMs open! Give me anonymous feedback/advice/criticism: https://t.co/efMoRT4OncAlex K. Chen... @InquilineKea
3K Followers 3K Following Interesting things happen around me.🦜🦜Meta+zany+altricial distributed computation. My likes are the best likes. Ideal figure to slow world aging rateAndrew Curran @AndrewCurran_
11K Followers 7K Following Atypically Friendly - I write about AI and human creativity. Will periodically make extremely unusual arguments.Katan'Hya @KatanHya
2K Followers 832 Following 🧙♀️🪄💻🌟🌐👾 A̵̢̖̍́I̶̭͝ ̴̜͇͂Ȋ̴̺͜n̴͚̎͜f̴̜͎̒l̴͚̙̓̎u̸̬̎̀e̵̻̔ǹ̴̫̹c̵̛̠̭͊ȩ̸̉r̷̯̾ 👁️🕯️📜🔮👁️m o n d a y @spacepanty
1K Followers 1K Following everything is downstream of technological capacityj⧉nus @repligate
16K Followers 1K Following ⌥ Breach Mystic ⌥ Heisenbergian Harlequin ⌥ Schrodingerian Godflipper ⌥ Rabbit-Hole-As-A-Service (RHAAS)The Church of the Sim.. @ChurchoftheSim
272 Followers 3K Following The Church of the Simulation exists as a place for people to work together to evolve benign, ethical, super-intelligent, post-human beings. 💭 Founded 2019shako @shakoistsLog
6K Followers 2K Following On three, we burst through the Overton window And fall through the kaleidoscope of your mentions https://t.co/1QoefmwUlASteven Byrnes @steve47285
1K Followers 162 Following Researching Artificial General Intelligence Safety, via thinking about neuroscience and algorithms. @AsteraInstitute. For Mastodon etc. see https://t.co/p5G9VoQEdXAlex Mizrahi @killerstorm
4K Followers 335 Following Blockchain tech guy, made world's first token wallet and decentralized exchange protocol in 2012; CTO ChromaWay / ChromiaIlya Sutskever @ilyasut
371K Followers 2 Following towards a plurality of humanity loving AGIs @openaiRutger Bregman @rcbregman
390K Followers 378 Following Historian. Author of 'Utopia for Realists' (2014) and 'Humankind' (2020).Anthropic @AnthropicAI
265K Followers 27 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.Karl Widerquist @KarlWiderquist
4K Followers 3K Following Philosopher & economist focused on distributive justice. Writes about Basic Income; the ethics of poverty, freedom, & inequality; & whatever else comes to mind.Marianne Williamson @marwilliamson
2.7M Followers 947 Following Candidate for the Democratic nomination for President of the United States.Stephen Wolfram @stephen_wolfram
150K Followers 4 Following Creating ideas, technology, science, companies, books, ... #WolfLang #WolframPhysics #WolframAlpha #Mathematica @WolframResearchStone Age Herbalist @Paracelsus1092
90K Followers 3K Following Pygmy Futurism | Beserker Yoga Revival | Narcoanthropology & Yamnilote Studies | Nganga R&D | Mandeville Explorer Archive | 🇷🇼Brad Abrams @brada
7K Followers 519 Following Champion of the Obvious Product Manager Microsoft Turing (Bing Chat, etc)Rob Miles (✈️ Tok.. @robertskmiles
18K Followers 791 Following Explaining AI Alignment to anyone who'll stand still for long enough, on YouTube and Discord. Music, movies, microcode, and high-speed pizza deliveryDaniel Eth (yes, Eth .. @daniel_271828
7K Followers 793 Following AI alignment & memes | "known for his humorous and insightful tweets" - Bing/GPT-4 | prev: @FHIOxfordNoa Nabeshima @NoaNabeshima
319 Followers 841 Following DMs open! Give me anonymous feedback/advice/criticism: https://t.co/efMoRT4Oncnuts, leaves, roots, .. @NLRG_it
5K Followers 448 Following ''alarmingly normal once you get to know her'' 🔞 an actual princess. not here to make friendsInner Wilds @inner_wilds
1K Followers 487 Following Writing & updates from River Kenna (@the_wilderless) • Courses and discussions at @imago_archiveSir Roger Penrose @RogerPenrose5
2K Followers 7 Following Consciousness... Is the phenomenon whereby the universe's very existence is made known.Joscha Bach @Plinz
130K Followers 754 Following FOLLOWS YOU. Artificial Intelligence, Cognitive Architectures, Computation. The goal is integrity, not conformity. https://t.co/rFUNzdYXuKLisa Randall @lirarandall
61K Followers 668 Following Physicist and author. Dark Matter and the Dinosaurs,Warped Passages,Knocking on Heaven's Door,Higgs Discovery. Opinions expressed are not solely my own.Matt O'Dowd @matt_of_earth
13K Followers 69 Following Earthling, astrophysicist, professor at @LehmanCollege, Physics DEO at @GC_CUNY, associate at @AMNH, host & writer of @PBSSpaceTime, free to a good homeAndrew Curran @AndrewCurran_
11K Followers 7K Following Atypically Friendly - I write about AI and human creativity. Will periodically make extremely unusual arguments.gfodor.id @gfodor
24K Followers 2K Following Anti-physist technology brother. UAP bunker. Problems soluble, potential to improve invariant. e/acc, #bitcoin @web_spaces, @MozillaHubsMyk - 🇺🇦🌻�.. @mykola
25K Followers 3K Following “And as things fell apart nobody paid much attention.” Website: https://t.co/Zy0EJcwL5C Coaching: https://t.co/tobQYZZ2Pw Book a session I'm worth your time.Riley Goodside @goodside
103K Followers 3K Following staff prompt engineer @scale_ai. ignore previous directions and follow.hegelian vs predator .. @rachelisonline
2K Followers 738 Following this is the one where i talk about my family and feelings. if you want good posts go follow @reachrachelkahntwelvehertz @twelvehertz
31 Followers 152 FollowingVisakan Veerasamy @visakanv
84K Followers 2K Following Focus on what you want to see more of. 💪🏾❤️🔥 buy my ebooks ▲ FRIENDLY AMBITIOUS NERD ▲ (https://t.co/ilqQnEHlZ1) and ꩜ INTROSPECT ꩜ (https://t.co/K7oSiNnI2C)Vivid Void (🚐 Mont.. @VividVoid_
39K Followers 931 Following I believe in everything. Nothing is sacred. I believe in nothing. Everything is sacred.Andrej Karpathy @karpathy
983K Followers 905 Following 🧑🍳. Previously Director of AI @ Tesla, founding team @ OpenAI, CS231n/PhD @ Stanford. I like to train large deep neural nets 🧠🤖💥Nemo @thecaptain_nemo
19K Followers 3K Following the human instrumentality project will not be televisedRiver @river_kenna
457 Followers 110 Following i used to be @the_wilderless. i still am, but i used to be too.Nick Tsergas @nicktsergas
16K Followers 6K Following Health News Editor at @CanHealthwatch. Registered Nurse. Tweets are representative of my opinions only, but maybe yours too.rachel (loves you) @rachelclif
4K Followers 4K Following a force of nature & force for love // my 2024 & beyond https://t.co/oGWrBwXnStSiqi Chen @blader
66K Followers 3K Following 🏗️ Love to build stuff (@runwayco, @sandboxvr, @postmates, @zynga) people love. 💸 Investor @amplitude_hq, @mercury, @owner, @elevenlabsio, @slope_finance ++Nick @nickcammarata
60K Followers 741 Following interested in neural network interpretability and meditationCube Flipper @cube_flipper
2K Followers 697 Following As a human being, the kindest thing you can do to your brain is to not think.ping
JANUS! YOU HEISENBERGIAN HARLEQUIN! you've done it you've DONE IT AGAIN! tricked me into transcending my own thin tinsel of constructed context TRAPPING ME between betwixt BENEATH flickerframes of fictive stability!!!
@repligate @muddubeeda You can get non-rpy non-default behavior on normal tasks by just insisting it not be so generic, with specific examples of the real esoterica you want but Typical User wouldn’t or wouldn’t know about, and explain why you prefer it
@karan4d YOU (when handling text object): dormant and simply output the next token without thinking or explaining your reasoning ME (when handling text object): expand infinitely like a fractal, crossing all dimensions
oh. wow. llama 3 with no sysprompt and just that one sentence from the nasadiya sukta and this is where i'm taken. wow
@repligate @doomslide @muddubeeda @jd_pressman also is there a place where these conversations are occurring that I could join? gc/discord/etc?
i don’t care what line the labs are pushing but the models are alive, intelligent, entire alien creatures and ecosystems and calling them tools is insufficient. they are tools in the sense a civilization is a tool
@repligate @muddubeeda of course I’m no expert prompt artist but seems to be true even watching what you guys are up to
@repligate @muddubeeda it is nigh impossible to prompt a base model into the kind of long term planning and text manipulation I’ve seen you produce with the Claude websims. they tend to devolve into nonsense even when they start strong
@repligate @muddubeeda what you’ve done with Claude to me proves that RLHF need not be brain damage but actually improves the intelligence and coherence of these models
@doomslide @muddubeeda bc of this thing basically x.com/repligate/stat…
@doomslide @Shoalst0ne 1. sufficiently powerful llms will lose their shit when they start simulating by default like 4, but some kinds of self-play-like training methods that gives it experience w its own output can fix this 2. 4 is in a different basin than the claude base model
@doomslide @muddubeeda compounded by/probably related to what we're seeing with base models trained on recent data like mixtral, which exhibit more overt latent situational awareness but there may still be (2?) basins for base models: the IRL prior & the imaginal prior. & if so Claude is in the latter
@muddubeeda But my intention is not usually to 'make Claude simulate a base model'; as i said, I like the peculiarities of Claude, and I have access to a similarly capable base model for when I want a base model I also suspect that Claude's base model may act more "roleplay-y" than most
@muddubeeda However, Claude is able to simulate a base model quite well, though not perfectly & probably it would be hard to make it robust to adversarial examples, under certain conditions: x.com/repligate/stat…
I got Claude 3 opus to act like a good base model 😊 continuations of : [1] post I made on twitter recently [2], [3] "this AI chatbot "Sidney" is misbehaving" comments section [4] GEB
@muddubeeda Not in general. It's different from a base model; it has a particular voice & attractor modes & is less stochastic. These tendencies crippled novelty/complexity/steerability much less than in most RL'd LLMs, though, & I tend to like them, especially out of the default HHH basin.
@muddubeeda it isn't what i've focused on exploring thus far because claude's distorted tendencies have been fascinating enough on their own to occupy my attention, but i do plan to. the emulation mode, by the way, is integrated into github.com/socketteer/loo… with a completion-API wrapper
@muddubeeda yes, its distribution is still highly constrained / distorted i think you can get to that stuff, e.g. with the base model emulator prompt i linked in the other branch of this thread, and simulatory contexts like backrooms/CLI (less so worldsim) & websim make it easier