A breakout year for AI

Hey everyone, my OOO is on and I'm starting my celebrations today with a big, slightly festive, meal.

We're wrapping up (no pun intended) the hackathon winners today and hope to announce them tomorrow before the Christmas weekend.

Let's get to it.

🤌 Ben's Picks

A thread on how Home Alone in the style of the opening credits of White Lotus season two was made with Runway. This is just such a wonderful look at how creative you can be using AI tools. Most likely this would've taken much longer and with a lot more technical skill if AI wasn't as powerful as it is today. (and about to get a whole lot more powerful!) (link)

A breakout year for AI art brings as many questions as answers. A summary: In 2022, companies working on generative AI raised $1.3bn, a 15% year-on-year increase, according to Pitchbook data. OpenAI's DALL-E 2 image-generation model went viral, with over 3 million users generating over 4 million images per day within five weeks of its public release. Generative AI has been used in a range of industries, but it has also raised ethical and legal questions, including how leading image platforms should approach AI-generated content. One programmer recently sued Microsoft, GitHub, and OpenAI over GitHub Copilot, a generative AI tool that used publicly available computer code to learn to write its own code. (link)

David Friedman talks about the how complicated problems of the Westworld-like future are already here with ChatGPT. He went into a role-playing scenario and made ChatGPT have a personality - although it could break character with ((a double bracket)). Again we're seeing a really interesting script-like output that's kind of choose-your-own adventure. Which I'm sure you could just use some Runway magic to turn into a video. (link)

🛠️ Cool Tools

  • Karlo - The first large scale open source DALL-E 2 replication. (link)

    • Try it on Hugging Face spaces. (link)

  • Penelope (Alpha) - A sophisticated AI writing assistant. (link)

  • SlackGPT - Zapier template to query GPT-3 via a Slack channel in real-time. (link)

  • DetectGPT - A Chrome extension to detect if a page contains AI-generated content. (link)

  • VALYR - Simplify GPT-3 monitoring with one line of code. (link)

  • Paperade - AI-powered tool that generates commercial use cases and company ideas from over 100 million academic papers and research studies. (link)

  • Visual Hound (BETA) - Prototype your fashion design ideas with AI. (link)

  • Proface by Avatarize - Create high quality AI-generated professional photos. (link)

  • Alfred - AI-powered search and content creation engine on iOS. (link)

  • MarbleFlows - AI-generated forms to convert more leads. (link)

  • Replai - Chrome extension to create meaningful Twitter replies with AI. I wonder how much of 2023 content on Twitter will be AI-generated... People are already spotting AI-replies to tweets. And interestingly, we'll run out of data to train AI in the next 3-4 years which would mean looking at content on Reddit and Twitter - AI trained on AI content?! (link)

🎓 Learn

  • A complete guide to building an AI-based chatbot app like Replika. (link)

  • How to make AI content generator in Python with Cohere AI. (link)

  • A free course on ‘Effective MLOps: Model Development.’ (link)

  • A guide on how you can use CLIPSeg, a zero-shot image segmentation model, using transformers. (link)

  • How to make an entire slide presentation using Tome. (link)

  • How you can use your own artwork to train custom AI style models with Runway. (link)

🔬 Research

  • ImageNet-X: A set of human annotations pinpointing failure types for the popular ImageNet dataset. (link)

  • EHR-Safe: Generating high-fidelity and privacy-preserving synthetic electronic health records. (link)

  • Language Models as inductive reasoners. (link)

  • From images to textual prompts: Zero-shot VQA with frozen large language models. (link)

  • ImPaKT: A dataset for open-schema knowledge base construction. (link)

  • mFACE: Multilingual summarisation with factual consistency evaluation. (link)

  • 3D Highlighter: Localising regions on 3d shapes via text descriptions. (link)

  • Parallel Context windows improve in-context learning of large language models. (link)

  • X-Decoder: Generalised decoding for pixel, image and language. (link)

  • WALDO: A trained detection (bounding-box) deep neural network for UAV enabling the detection of land-based objects. (link)

  • PaletteNeRF: Palette-based appearance editing of neural radiance fields. Explain like I'm 5: PaletteNeRF is a way to change how something looks in 3D space on a computer. It takes a bunch of little pieces and mixes them together to make a picture look different. You can change the colors of the pieces to make the picture look different. PaletteNeRF is really good at making the changes look real and not fake. (link)

👋 Too many links?! I created a database for all links mentioned in these emails. Refer 1 friend using this link and I'll send over the link database.

🤓 Everything else

  • Mapping the Generative AI landscape. (link)

  • A book on copywriting written 100% with AI. (link)

  • ​​Gradio demo of The-Jam-Machine, a Generative AI trained on text transcription of MIDI music on Hugging Face spaces. (link)

  • A thread on prediction for GPT-4 in 2023. (link)

  • New York Times article on how OpenAI's ChatGPT & search AI innovations could threaten Google's search dominance. ChatGPT made Google declare 'code red' (apparently) and got management to seriously think about what the implications are (apparently). Google doesn't want to cannibalise it's ad revenues on search so it'll be interesting to see what their response will be (if any). (link) Related: another article talking about Google's 'code red'. (link)

  • A journalist used OpenAI's ChatGPT chatbot to write essays for high school English classes to test its capabilities. The chatbot was able to write an 800-word essay on "Ferris Bueller's Day Off" as an existentialist text in under a minute, but it struggled with demand and experienced network errors. It also successfully wrote an essay on "The Great Gatsby," but the teacher noticed some factual errors and that the writing style was too advanced for a high school student. (link) (no paywall link 🤫)

  • Amazon's Machine Learning University debuts responsible AI course. (link)

  • Understanding Hugging Face’s Model Cards. (link)

  • ChatGPT and the professional's guide to using AI. (link)

  • A thread on AI projects built as part of Replit Bounties. (link)

    • Do you have an AI app idea that you need built? Get it built with Replit’s bounty program. (link)

🖼 AI images of the day

Stelfie Log #10 : England, 1266.

🤗 Share Ben's Bites

Send this to 1 AI-curious friend and receive my AI project tracker database!

or copy/paste this link: https://bensbites.beehiiv.com/subscribe?ref=PLACEHOLDER

👋 See ya

⭐️ How did we do?

How was today's email?

Login or Subscribe to participate in polls.

⭐️ REAL REVIEWS

Join the conversation

or to participate.