If you plug “AI” or “artificial intelligence” into Google Image Search, here’s what you get. Electric brains firing blue with neurons. Translucent robots and robot heads, code dribbling down their foreheads and noses.

This is a 1 hour general-audience introduction to Large Language Models: the core technical component behind systems like ChatGPT, Claude, and Bard. What they are, where they are headed, comparisons and analogies to present-day operating systems, and some of the security-related challenges of this new computing paradigm. As of November 2023 (this field moves fast!).

Context: This video is based on the slides of a talk I gave recently at the AI Security Summit. The talk was not recorded but a lot of people came to me after and told me they liked it. Seeing as I had already put in one long weekend of work to make the slides, I decided to just tune them a bit, record this round 2 of the talk and upload it here on YouTube. Pardon the random background, that's my hotel room during the thanksgiving break.

• Slides as PDF: https://drive.google.com/file/d/1pxx_ZI7O-Nwl7ZLNk5hI3WzAsTLwvNU7/view?usp=share_link (42MB)
• Slides. as Keynote: https://drive.google.com/file/d/1FPUpFMiCkMRKPFjhi9MAhby68MHVqe8u/view?usp=share_link (140MB)

Few things I wish I said (I'll add items here as they come up): - The dreams and hallucinations do not get fixed with finetuning. Finetuning just "directs" the dreams into "helpful assistant dreams". Always be careful with what LLMs tell you, especially if they are telling you something from memory alone. That said, similar to a human, if the LLM used browsing or retrieval and the answer made its way into the "working memory" of its context window, you can trust the LLM a bit more to process that information into the final answer. But TLDR right now, do not trust what LLMs say or do. For example, in the tools section, I'd always recommend double-checking the math/code the LLM did. - How does the LLM use a tool like the browser? It emits special words, e.g. |BROWSER|. When the code "above" that is inferencing the LLM detects these words it captures the output that follows, sends it off to a tool, comes back with the result and continues the generation. How does the LLM know to emit these special words? Finetuning datasets teach it how and when to browse, by example. And/or the instructions for tool use can also be automatically placed in the context window (in the “system message”). - You might also enjoy my 2015 blog post "Unreasonable Effectiveness of Recurrent Neural Networks". The way we obtain base models today is pretty much identical on a high level, except the RNN is swapped for a Transformer. http://karpathy.github.io/2015/05/21/rnn-effectiveness/ - What is in the run.c file? A bit more full-featured 1000-line version hre: https://github.com/karpathy/llama2.c/blob/master/run.c

Chapters: Part 1: LLMs 00:00:00 Intro: Large Language Model (LLM) talk 00:00:20 LLM Inference 00:04:17 LLM Training 00:08:58 LLM dreams 00:11:22 How do they work? 00:14:14 Finetuning into an Assistant 00:17:52 Summary so far 00:21:05 Appendix: Comparisons, Labeling docs, RLHF, Synthetic data, Leaderboard Part 2: Future of LLMs 00:25:43 LLM Scaling Laws 00:27:43 Tool Use (Browser, Calculator, Interpreter, DALL-E) 00:33:32 Multimodality (Vision, Audio) 00:35:00 Thinking, System 1/2 00:38:02 Self-improvement, LLM AlphaGo 00:40:45 LLM Customization, GPTs store 00:42:15 LLM OS Part 3: LLM Security 00:45:43 LLM Security Intro 00:46:14 Jailbreaks 00:51:30 Prompt Injection 00:56:23 Data poisoning 00:58:37 LLM Security conclusions End 00:59:23 Outro

A video about neural networks, function approximation, machine learning, and mathematical building blocks. Dennis Nedry did nothing wrong. This is a submission for #SoME3

My Links Patreon: https://www.patreon.com/emergentgarden Discord: https://discord.gg/ZsrAAByEnr

Links and Content: On Mathematical Maturity, Thomas Garrity: https://www.youtube.com/watch?v=zHU1xH6Ogs4 Earth Rotation Loop: https://www.youtube.com/watch?v=aiQdLP2mBJE Modeling Shell Surfaces: https://www.geogebra.org/m/xtv7zpn5 Fourier Features Paper: https://arxiv.org/abs/2006.10739 Code for mandelbrot/image approximations: https://github.com/MaxRobinsonTheGreat/mandelbrotnn Code for line/surface approximations: https://github.com/MaxRobinsonTheGreat/ManimApproximations

Music: https://youtube.com/@acolyte-compositions

Timestamps (0:00) Functions Describe the World (3:15) Neural Architecture (5:35) Higher Dimensions (11:55) Taylor Series (15:20) Fourier Series (21:25) The Real World (24:32) An Open Challenge

Hi, woke moralists! In today’s episode, we take a look at Jordan Peterson through the lens of Jungian archetypes – just as he would want – and explore some of his other ideas about climate change, gender, marriage, hierarchies, and lobsters. Don’t check the timecode!

Get your BILLIONAIRES ARE NOT YOUR FRIENDS merch here: https://www.teepublic.com/t-shirt/31298912-billionaires-are-not-your-friends?store_id=237592

Check out our new compilation series, CODY COMPS here: https://www.youtube.com/playlist?list=PLqFkH8uXvlJbGeJKUChW4VGxCMhKQ9cJ8

Please fill out our SURVEY: https://kastmedia.com/survey/

Check out our new series SOME THIS! - https://www.youtube.com/playlist?list=PLkJemc4T5NYbcqTbNmyH3uqutwcj8fHf3

Support us on our PATREON: http://patreon.com/somemorenews Check out our MERCH STORE: https://www.teepublic.com/stores/somemorenews?ref_id=9949 SUBSCRIBE to SOME MORE NEWS: https://tinyurl.com/ybfx89rh

Executive Producer - Katy Stoll (@KatyStoll). Written by David Christopher Bell (@MovieHooligan) and Cody Johnston (@drmistercody). Directed by Will Gordh (@will_gordh). Edited by Gregg Meller (@greggmeller). Graphics by F. Clint DeNisco. Head Writer - David Christopher Bell. Producer - Jonathan Harris. Researcher - Marco Siler-Gonzales (@mijo_marco). Associate Producer - Quincy Tucker (@quincyrashard).

Subscribe to the Even More News and SMN audio podcasts here: Apple Podcasts: https://podcasts.apple.com/us/podcast/some-more-news/id1364825229 Spotify: https://open.spotify.com/show/6ebqegozpFt9hY2WJ7TDiA?si=5keGjCe5SxejFN1XkQlZ3w&dl_branch=1 Stitcher: https://www.stitcher.com/show/even-more-news

Find your new favorite shoes for sunny days and upcoming travel at allbirds.com. That’s http://ALLBIRDS.com.

High-performance beauty and skin-care products made with clean, skin-loving ingredients. Right now, you can get 15% off your first order when you visit thrivecausemetics.com/MORENEWS.

Chapters: 00:00 - Intro 8:53 - Jordan Peterson Is Not A Climate Expert 19:24 - Jordan Peterson Demands That You Respect His Bigotry 26:28 - Who, What, Why, and How is Jordan Peterson? 34:56 - The Persona 1:28:44 - The Shadow 2:00:47 - The Anima And Animus 2:34:13 - The Self

Source list: https://docs.google.com/document/d/1ciPKIdTqGdHsxsj-7vjlcnyq4JDYlLBM-UyH3LBi3No/edit

jordanpeterson #jordanpetersonandjoerogan #jordanpetersonquotes #somemorenews

An artificial intelligence trained to mimic the logic circuits of an ordinary computer can run code within itself, potentially speeding up certain calculations An artificial intelligence that mimics the operation of a standard computer within its neural network could speed up certain calculations.

An artificial intelligence trained to mimic the logic circuits of an ordinary computer can run code within itself, potentially speeding up certain calculations An artificial intelligence that mimics the operation of a standard computer within its neural network could speed up certain calculations.

Share this so it isn't my last episode. Join our Discord! https://discord.gg/SrDAqQzQ

✳ PATREON: https://www.patreon.com/user?u=711561 ✳ Twitter: http://twitter.com/_lucasrizzotto ✳ Facebook: http://facebook.com/lucasrizzotto ✳ Instagram: http://instagram.com/_lucasrizzotto ✳ Website: http://lucasrizzotto.com]

Thanks for the wild ride, friends. This one was a joy (and hell) to make. Make sure you follow me closely on social media - got some big announcements to make this week!

I really appreciate you all, specially those of you who dig deep into the description to find hidden bits like this :)

This channel never really blew up the way I thought it would but it brought me close with some of the most passionate people I've ever seen and I'll always be eternally grateful.

Love ya,

Swooty McBooty

Special thanks to: Austin Beaulier (for helping me get the epic smooth shots in the memories sequence) Cix Liv (for all his expert insight as Dimitri) Max Noir (for dealing with all the bs of making this video for almost a year lmao) Stephen Hodgson (for helping out hugely with GPT-3 code)

Chapters!

00:00 - Teaser 00:38 - Introduction 02:35 - Jeff Bezos' Tale 04:22 - The Brain Transplant 05:07 - GIANT METAL DEATH MACHINES 06:00 - Okay Let Me Explain GPT-3 For Real Now lol 07:10 - Writing Fake Memories 09:30 - PLOT TWIST! An Old Friend 11:20 - Ad-Break 12:24 - The Conversation Begins 13:36 - Microwave Gets Personal 14:45 - Microwave Politics 15:43 - Oh God He Likes Hitler 16:18 - Dramatic Question Montage 19:34 - Unintended Consequences 20:19 - Microwave PTSD Montage 21:12 - The Murder Attempt 23:04 - Lucas Moves Far Away... 23:33 - ... To Uganda LOL 24:22 - Haunted By Nightmares 24:53 - Talking to an Expert 26:34 - Is Magnetron Fake? 27:06 - The Final Conversation 30:12 - Goodbye, Magnetron 31:34 - Is it over? 32:15 - Outro / Sad Announcements 34:20 - Goodbyes & Reminiscing

Production Crate: https://bit.ly/blackthornVFX

PLAY on Windows: https://blackthornprod-games.itch.io/piece-by-piece

PLAY on Web: https://www.newgrounds.com/portal/view/829585

Jonas Tryoller: https://www.youtube.com/c/JonasTyroller Sokpop: https://www.youtube.com/c/sokpopcollective/videos Code Monkey: https://www.youtube.com/c/CodeMonkeyUnity

Yän: https://www.youtube.com/c/Y%C3%A4nGamedev/videos

In this video, 6 game developers spend 4 hours each making a video game without any communication! Liam began work prototyping the game, then passed it on to me, then Aran from Sokpop, Jonas Tyroller, Code Monkey, and Yän!

We ended up creating a cool spaceship building/survival experience.

0:00 - Intro 0:15 - Liam 1:03 - Noa 2:27 - Sokpop 3:58 - Jonas Tryoller 7:12 - Code Monkey 9:08 - Yän

In her new book Race After Technology: Abolitionist Tools for the New Jim Code, Ruha Benjamin breaks down the “New Jim Code,” technology design that promises a utopian future but serves racial hierarchies and racial bias.

On March 13, Li Jin sold a GIF for $25,000. A link to the moving image, which was assigned a unique bit of code as proof of authenticity and stored on the blockchain, was purchased at auction via the cryptocurrency Ethereum and sold in 24 hours.

There are tens of thousands of genes in the human genome: minuscule twists of DNA and RNA that combine to express all of the traits and characteristics that make each of us unique. Each gene is given a name and alphanumeric code, known as a symbol, which scientists use to coordinate research.

So this neural network is called Jukebox, by OpenAI, which basically generates songs that don't exist. Also bonus AI upscaled gifs of Rick Astley, which definitely doesn't make him look like Satan spawn.

Never Gonna Give You Up (AI Versions): https://jukebox.openai.com/?song=787729588 https://jukebox.openai.com/?song=787732303 https://jukebox.openai.com/?song=787735267 https://jukebox.openai.com/?song=787733749 https://jukebox.openai.com/?song=787730953

[Blog] https://openai.com/blog/jukebox/ [Sample Explorer] https://jukebox.openai.com/ [Source Code] https://github.com/openai/jukebox/ (You can generate your own songs, but you need a pretty beefy computer, some programming knowledge, and a lot of time and patience) [AI Upscaler] https://topazlabs.com/gigapixel-ai/ (There's a 30-day trial if you don't feel like paying for it)

Wow, this is the most views I've gotten on a video, like, ever. It even surpasses videos on my other channel.

[PLEASE READ THE CHANNEL DESCRIPTION]

In her new book Race After Technology: Abolitionist Tools for the New Jim Code, Ruha Benjamin breaks down the “New Jim Code,” technology design that promises a utopian future but serves racial hierarchies and racial bias.

TASBot takes the Celeste All Red Berries train and discovers CovertMuffin blocks (with downtime edited out) by euni, fishmcmuffins, and kilaye, joined on the couch by TGH, buhbai, Kevin from PowerUp Audio, and dwangoAC at the SGDQ 2019 TAS block. This TAS is incredible, complete with a massive chain of red berries following Madeline up the mountain.

GDQ set the donation incentive to $175k to see this game and viewers responded resoundingly, filling the donation incentive with plenty of time to spare. This Celeste All Red Berries TAS represents over 2,000 hours of effort spearheaded by euni with help from fishmcmuffins and kilaye. This TAS was made with tools pioneered by DevilSquirrel and extended over time by others.

The commentary on this run is some of the best at any GDQ, with frequent asides from TGH on realtime viability and developer commentary right on the couch from Kevin who helped in the creation of the game. This run even includes a surprise beyond the jawdropping; unbeknownst to many, Kevin from PowerUp Audio secretly altered the sound played by his namesake blocks to spread some FWAHAHA love.

Note: During the event a minor startup issue occurred which required restarting the run; that section has been edited out of the video. Additionally, the GDQ sound board crashed during the run causing 10 seconds of lost audio after which the run was paused until the issue could be resolved. The downtime during that issue was also edited. The run is otherwise unbroken as evidenced by the countdown timer and everything presented is a legitimate playthrough; the tools used to create this TAS do not alter gameplay in any way and this represents the fastest possible way to navigate the game based on how the game is written. There is no cheating on display here, only hours of dedication by the TAS authors to determine the fastest possible route through the game while collecting every single Red Berry in the game.

Celeste community links: euni - https://www.twitch.tv/euni_i kilaye - https://www.youtube.com/user/KilayeFr fishmcmuffins - https://www.twitch.tv/fishmcmuffins Source code - https://github.com/EuniverseCat/CelesteTAS/tree/SGDQ

TASBot links: Discord - http://Discord.TAS.bot TASBot home - https://TAS.bot Twitch live streams - http://twitch.tv/dwangoAC Music by DJPIE1337 - https://dj-pie.bandcamp.com AngelWind: YouTube Editor for Hire - angelwind76@gmail.com

This video was posted with permission from Games Done Quick and with the permission of euni representing the TAS authors.

'Trust Me' - I'm an Algorithm

Trust is fundamental to all our relationships - not just with our family and friends. We trust banks with our money, we trust doctors with our really personal information. But what happens to trust in a world driven by algorithms? As more and more decisions are made for us by these complex pieces of code, the question that comes up is inevitable: can we trust algorithms?

Join Ali Rae in this first episode of ‘All Hail The Algorithm’ - a 5 part series exploring the impact of these invisible codes on our everyday lives.

In this episode, Ali speaks with Asher Wolf about Australia’s robo-debt scandal, Virginia Eubanks about how inequality is being automated and Sharad Goel about COMPAS - an algorithm being used in courts across the US to assist with sentencing.

Subscribe to our channel http://bit.ly/AJSubscribe Follow us on Twitter https://twitter.com/AJEnglish Find us on Facebook https://www.facebook.com/aljazeera Check our website: http://www.aljazeera.com/

AlJazeeraEnglish #Algorithms #AllHailtheAlgorithm

When biologists synthesize DNA, they take pains not to create or spread a dangerous stretch of genetic code that could be used to create a toxin or, worse, an infectious disease.

The death of a woman hit by a self-driving car highlights an unfolding technological crisis, as code piled on code creates ‘a universe no one fully understands’ by The 18th of March 2018, was the day tech insiders had been dreading.

Casting Code: Reflections on 3D Printing Blog The letterhead of Henri Lebossé announces that his firm uses a ‘mathematically perfected process’ and a ‘special machine’ for ‘reducing and enlarging objects of “art and industry”’.

A new study strongly suggests that at least some memories are stored in genetic code, and that genetic code can act like memory soup. Suck it out of one animal and stick the code in a second animal, and that second animal can remember things that only the first animal knew.