Aside from the physical heft of data centers seen from highways and the fiber optic cables crawling into homes and offices, the digital world mostly exists in our imagination. That imagination is shaped by the people selling services that rely on that infrastructure.

Large language models like those offered by OpenAI and Google famously require vast troves of training data to work. The latest versions of these models have already scoured much of the existing internet which has led some to fear there may not be enough new data left to train future iterations.

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

Is this the world's first AI-generated documentary? Alan Warburton was commissioned by the ODI's Data as Culture programme to bring us 'The Wizard…

Responding to the 1 millionth time that people have said "the left has failed men" and somehow... kinda agreeing with Shoe on Head

2:00 Y'all are overreacting to that table 7:40 The Manosphere isn't new it's just more visible 11:08 Algorithms make this an uphill battle 15:14 The online left does have a weak analysis of masculinity 17:57 Stop whining about mean feminist and do the work 19:40 The real problems facing men

Classic debunks of various manosphere topics MuneCat debunks the manosphere - https://www.youtube.com/watch?v=BgO25FTwfRI&pp=ygUHbXVuZWNhdA%3D%3D My Black Manosphere video - https://www.youtube.com/watch?v=upt_ks61_70&t=1466s&pp=ygUQYmxhY2sgbWFub3NwaGVyZQ%3D%3D Sissyphus trying to make sense of the manosphere - https://www.youtube.com/watch?v=lSq3bcyrJY0&pp=ygUOdGhlIG1hbm9zcGhlcmU%3D Frank Laundry explaining Fresh and Fit - https://www.youtube.com/watch?v=WBsCmYCShqI&pp=ygUbZnJhbmsgbGF1bmRyeSBmcmVzaCBhbmQgZml0 Zatzman on the Manosphere as a content farm - https://www.youtube.com/watch?v=pvZITKjaffw&pp=ygUOdGhlIG1hbm9zcGhlcmU%3D

Other male content creators to check out

@Salari @FinntasticMrFox @ForeignManinaForeignLand @NoahSamsen @Sisyphus55 @TheZatzman @COLORMIND.mp4 @victorythecreator @lilbilliam @ThinkpieceTribe @JawnLouis

Monitoring the future study 2022 - https://monitoringthefuture.org/wp-content/uploads/2022/12/mtf2022.pdf Political ideology data from 2021 - https://monitoringthefuture.org/wp-content/uploads/2023/08/PoliticalBeliefTrends12thGraders1976-2021.xlsx CDC Deaths of despair study - https://www.cdc.gov/surveillance/blogs-stories/deaths-of-dispair.html Meta study mentioned - https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9566538/

You have full access to this article via your institution. An artificial intelligence (AI) system based on Google DeepMind’s AlphaZero AI created algorithms that, when translated into the standard programming language C++, can sort data up to three times as fast as human-generated versions.

Workers hired via crowdsource services like Amazon Mechanical Turk are using large language models to complete their tasks – which could have negative knock-on effects on AI models in the future. Data is critical to AI.

Critical Topics: AI Images is an undergraduate class delivered for Bradley University in Spring 2023. It is meant to provide an overview of the context of AI art making tools and connects media studies, new media art, and data ethics with current events and debates in AI and generative art.

A Primer on AI in/from the Majority World is a curated collection of over 160 thematic works that serve as pathways to explore the presence of artificial intelligence and technology in the geographic regions that are home to the majority of the human population.

A Primer on AI in/from the Majority World is a curated collection of over 160 thematic works that serve as pathways to explore the presence of artificial intelligence and technology in the geographic regions that are home to the majority of the human population.

Data & Society is an independent nonprofit research organization. We believe that empirical evidence should directly inform the development and governance of new technology.

When I was studying for my doctorate, in the late 1960s, we budding anthropologists read a book called Ideas on Human Evolution, a collection of then-recent papers in the field. With typical graduate-student arrogance, I pronounced it “too many ideas chasing too little data.

When I was studying for my doctorate, in the late 1960s, we budding anthropologists read a book called Ideas on Human Evolution, a collection of then-recent papers in the field. With typical graduate-student arrogance, I pronounced it “too many ideas chasing too little data.

"Datamoshing is a technique of damaging video clips to create a glitch effect wherein frames that should change don't. It's most noticeable between cuts and across motion. Datamoshing is the process of corrupting, removing or replacing I-frames, causing P-frames to be applied to the wrong picture."

This write-up re-introduces the concept of entropy from different perspectives with a focus on its importance in machine learning, probabilistic programming, and information theory.

This write-up re-introduces the concept of entropy from different perspectives with a focus on its importance in machine learning, probabilistic programming, and information theory.

Art takes a unique combination of skill, creativity the very human element of aesthetic taste. But what if the visual arts also fall to AI. In this episode we discuss Dall-E 2, a powerful text to image generator that's set to shake things up.

A.I. Playlist: https://youtube.com/playlist?list=PL0iVR8sl9TiXmZUgZSZOpGFouuFNieqgP

My new Spotify album: https://open.spotify.com/album/2mQkQEgmLxCJC8JpcsiF2T?si=560AO6QiQ3ShO3qKBWnRhw

Sign up link here: https://labs.openai.com/waitlist

--- About ColdFusion --- ColdFusion is an Australian based online media company independently run by Dagogo Altraide since 2009. Topics cover anything in science, technology, history and business in a calm and relaxed environment.

» ColdFusion Discord: https://discord.gg/coldfusion » Podcast I Co-host: https://www.youtube.com/channel/UC6jKUaNXSnuW52CxexLcOJg » Twitter | @ColdFusion_TV » Instagram | coldfusiontv » Facebook | https://www.facebook.com/ColdFusioncollective » Podcast Version of Videos: https://open.spotify.com/show/3dj6YGjgK3eA4Ti6G2Il8H https://podcasts.apple.com/us/podcast/coldfusion/id1467404358

ColdFusion Music Channel: https://www.youtube.com/channel/UCGkpFfEMF0eMJlh9xXj2lMw

ColdFusion Merch: INTERNATIONAL: https://store.coldfusioncollective.com/ AUSTRALIA: https://shop.coldfusioncollective.com/

If you enjoy my content, please consider subscribing! I'm also on Patreon: https://www.patreon.com/ColdFusion_TV Bitcoin address: 13SjyCXPB9o3iN4LitYQ2wYKeqYTShPub8

--- "New Thinking" written by Dagogo Altraide --- This book was rated the 9th best technology history book by book authority. In the book you’ll learn the stories of those who invented the things we use everyday and how it all fits together to form our modern world. Get the book on Amazon: http://bit.ly/NewThinkingbook Get the book on Google Play: http://bit.ly/NewThinkingGooglePlay https://newthinkingbook.squarespace.com/about/

Sources:

https://cdn.openai.com/papers/dall-e-2.pdf

https://fortune.com/2022/04/06/openai-dall-e-2-photorealistic-images-from-text-descriptions/

https://www.engadget.com/open-a-is-dall-e-2-produces-fantastical-images-of-most-anything-you-can-imagine-170056814.html

https://www.engadget.com/the-morning-after-open-ai-dall-e-2-images-111840375.html

https://towardsdatascience.com/generating-images-from-prompts-using-clip-and-stylegan-1f9ed495ddda

My Music Channel: https://www.youtube.com/channel/UCGkpFfEMF0eMJlh9xXj2lMw

//Soundtrack//

Burn Water - Nostalgia Dreams

no spirit - leaves covered by snow

Burn Water - Does it Get Easier (clean)

Nils Frahm - You (Teen Daze Rework)

Jon Hopkins - The Wider Sun

Ben Böhmer - Flug & Fall

Kidnap Kid - Moments (feat. Leo Stannard)

Hammock - Wasted We Stared at the Ceiling

Burn Water - Fate

» Music I produce | http://burnwater.bandcamp.com or » http://www.soundcloud.com/burnwater » https://www.patreon.com/ColdFusion_TV » Collection of music used in videos: https://www.youtube.com/watch?v=YOrJJKW31OA

Producer: Dagogo Altraide

Charles Darwin thought of evolution as an incremental process, like the patient creep of glaciers or the march of continental plates.

Anthropologist Steven Gonzalez Monserrate draws on five years of research and ethnographic fieldwork in server farms to illustrate some of the diverse environmental impacts of data storage. The Cloud is not only material, but is also an ecological force.

Earth’s Black Box will record every step we take towards this catastrophe. Hundreds of data sets, measurements and interactions relating to the health of our planet will be continuously collected and safely store for future generations.