Hey folks, this is Ben's Bites - a little less bite than usual today as I've just had a tooth out 🦷. So I'm on liquids for a couple of days and I'll leave the chewing to you.

A few big AI dishes today, let's get to it.

p.s. I just launched a free AI job board⁸⁷⁴. Post any kind of gig you want on there.

Have a product, service, job, event, newsletter, app, book, movie, tool, or anything you'd like to share with over 26k subscribers? Check out our sponsor options²¹⁷.

Join Discord¹⁰³

🤌 Ben's Picks

1/

Google announce an upcoming chatbot search engine⁴³²

The Microsoft vs Google saga is seeeeriously heating up right now. Hot on the heels of the announcement of a ChatGPT-powered Bing, Google’s ‘code red’ has turned into code Bard - the name of their very own chatbot that could power their very own search engine. And it looks like it does everything that the former can!

But to add fuel to the fire, Microsoft has announced a SURPRISE media event today⁷²³ alongside Google’s on Wednesday. This is their chance to really give them a run: demos of the new Bing? Demos of AI-driven Teams meetings? AI-generated powerpoints? These could all be on the agenda… If I’ve got my timings right, it should be at 6pm London time, 10am Washington time - not too long after this hits your inbox!

Apple are also hosting an 'AI Summit' for staff according to this post.²¹⁰ It's notable as it's being held in the Steve Jobs Theater at the Apple HQ and streamed to employees also.

OH and ALSO, Baidu is launching 'Ernie Bot' in March and saw their share price increase 13%! (paywall removed²²⁹)

2/

Getty images sue Stability AI for stealing images³³²

Uh oh, there’s about to be some instability going on at Stability AI… Getty images have officially filed a lawsuit against them for using 12 million (yes, 12 million) images illegally to create their AI art machine. Big, scary phrases like “brazen infringement” and “staggering scale” are getting thrown around by lawyers.

This could be the start of a shift in the speed with which AIs get released going forward, where companies are gonna have to have more care for the contents of their training data, and probably spend more money on it too.

3/

Educational chatbot on Telegram^607†

I always thought Telegram was just for Bitcoin Bros, but these guys⁴⁰⁵ have done an awesome job of turning it into a pirate-themed educational tool using chatbot technology. (Don’t worry, you can learn about anything, not just pirates.)

Enter a topic of your choice and, after a few mins, a list of chapters gets displayed and you can work through them. After each chapter it offers quiz questions, you can go back and forth between sections, and it saves the output to a web page. I asked it to teach me about String Theory and it did a great job. (Although it still sounds like gibberish.)

Encord.com¹⁵⁴: Label, curate, and debug your training data. All in one platform (Sponsor)

Forget about fragmented tools and notebooks for creating your active learning pipelines. Encord is a single integrated platform that makes it quicker and easier to build production computer vision models using active learning pipelines. Encord helps you streamline your machine learning projects, giving you a single platform for labeling any visual data, managing annotators, improving training data quality and debugging your datasets and models.

Get in touch to arrange your free trial of Encord.⁹³ 

🛠️ Cool Tools

• DocuChat - Get answers from your documents with GPT-3 powered chatbots. (link⁷⁸⁸)

• Talently - Search, match & hire expert remote developers. (link³¹⁷)

• Private Mini Yohei - First LLM based chatbot with advanced capabilities like VC blog search, crunchbase search, drafting emails etc. (link⁴¹⁸)

• Clippy - Add AI-powered search and chat to your company's docs. (link⁴⁸⁸)

• EvenUp - Turn medical docs and case files into AI-driven demand packages for injury lawyers. (link²⁸¹)

• GPT Friends - Chat with the most human AIs ever created from GPT. (link⁴⁹⁰)

• Edward Teach Bot - Study any topic from a Telegram chat. (link^607†, co-founder’s blog here¹¹⁷)

• Invideo - Turn any content or idea into video. (link⁶⁵⁴)

• Gen-1 from Runway ML - A new AI model that uses language and images to generate new videos out of existing ones. (link⁵¹⁵, MIT tech review’s POV²⁵⁶) Go and watch their intro video!!

🤓 Miscellaneous

• Amid wave of layoffs, some tech giants are elevating AI. (link²²⁷)

• Exclusive: Bill Gates on advising OpenAI, Microsoft and why AI is the hottest topic of 2023. (link⁴²¹)

• Chat your data challenge from LangChain. (link^207†)

• CapitalG - Alphabet’s independent fund for growth stage companies. (link²⁰⁴)

• Latent Technology raises $2.1M to bring generative AI to video games. (link¹⁷⁴)

• ChatGPT + natural language + personal context = a more useful personal home device. (link³⁵⁵)

• Voice AI - General public licence violations with a side of DRM. (link¹⁴³)

• AI-generated ‘Seinfeld’ suspended on Twitch for AI-generated transphobic jokes. (link¹⁸⁷)

• Getty Images sues AI art generator Stable Diffusion in the US for copyright infringement. (link³³²)

• Baidu surges as hope mounts over Chinese answer to ChatGPT. (link⁹⁹)

• ChatGPT and generative AI in the enterprise. (link²⁶²)

• Stable Attribution: A new tool could ruin generative AI, or it could save it. (link²⁴⁵)

• How will productivity using ChatGPT evolve? (link³⁵⁸)

🎓 Learn

• How to use GPT embeddings for a second brain note-taking system? (link⁴⁹⁷)

• A primer on FOMO - Foundational model orchestration. (link¹⁴³)

• Storing OpenAI embeddings in Postgres with pgvector. (link¹⁹³)

👋 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.

🔬 Research

• AV-NeRF - Learning neural fields for real-world audio-visual scene synthesis. (link¹⁰⁷)

• The SSL Interplay - Augmentations, inductive bias, and generalization. (link⁷⁰)

• TR3D - Towards real-time indoor 3D object detection. (link⁹⁴)

• Multi-view masked world models for visual robotic manipulation. (link⁸³)

• AIM - Adapting image models for efficient video action recognition. (link⁷⁶)

• Languages are rewards - Hindsight fine tuning using human feedback. (link¹¹⁵)

• Grounding large language models in interactive environments with online reinforcement learning. (link⁷⁸)

• Representation deficiency in masked language modeling. (link⁶⁷)

• Measuring the impact of programming language distribution. (link⁹¹)

• Zero-shot image-to-image translation. (link¹¹³)

• Structure and content-guided video synthesis with diffusion models. (link⁹⁹)

• Mixture of diffusers for scene composition and high resolution image generation. (link⁸⁸)

• Eliminating prior bias for semantic image editing via dual-cycle diffusion. (link⁶⁷)

• Regulating ChatGPT and other large generative AI models. (link¹¹⁰)

• Design Booster - A text-guided diffusion model for image translation with spatial layout preservation. (link⁷⁷)

• SE(3) diffusion model with application to protein backbone generation. (link⁹⁴)

• Multi-source diffusion models for simultaneous music generation and separation. (link⁸⁸)

• Language models secretly perform gradient descent as meta-optimizers. (link⁶⁸)

• Hugginface demo of BioGPT-Large. (link¹⁵⁵)

📰 Unclassifieds

• Glasp, the social web highlighter, is hiring a Software Engineer. Learn more¹²⁰.

• Helping AI makers make better. Discover new tooling, see what others are building, and learn how to grow your product. Come make with us!¹⁵³

• AI/ML YouTube creator Siraj Raval will be building ChatGPT from scratch at the Founders, Inc. lab on Friday. Register here to drop by >>>¹³⁵

• Get ahead of the curve with Gold’s Guide¹⁵⁵: news + commentary on artificial intelligence.

Have a product, service, job, event, newsletter, app, book, movie, tool, or anything you'd like to share with over 26k subscribers? Check out our sponsor options²¹⁷.

🖼 AI images of the day