Meta's computer vision tool

PLUS: Open AI's stance on safety, India opts out of regulation, 3D videos

Hey folks, yesterday I said I was making an announcement and here it is. Quick version: I left Zapier, and I’m focusing on AI; Ben’s Bites (yay) and investing (I’ve become an a16z scout).

Ben’s Bites has been a little team of 4 for a few months now: Keshav, Olly and Dani - all have been key to helping organise, write, edit, and community manage everything. Extremely grateful to them, and all your lovely thank you’s deservedly go to them.

But the biggest news is my wife and I are having twins, coming July 2023!! Truly, thank you all for your support—it encouraged me to take the leap of faith.

Right, today we’ve got OpenAI’s stance on safety, Meta launches a state-of-the-art computer vision tool, Replit partners with Chroma, India opts out of AI regulation and creating 3D images+videos with ONE photo.

Let’s get to it.

From our sponsor:

Always wanted to speak a second language?

Really speak it, in real life with real people? Babbel is packed with lessons, games, podcasts, even live classes - everything you need to learn a new language and have real-world conversations. Gonna read those AI papers in Chinese now.

For a limited time, get 60% off when you sign up for a Babbel subscription!

🤌 Our Picks

Highlights if you've only got 2 minutes

1/

They’ve remained eerily quiet during the whole “pause-AI-petition” charade, but the guys who ultimately got the whole debate started by dominating AI development have responded. OpenAI have released a post outlining their approach to AI safety as the World moves forwards - it’s not a direct response per se, but we all know it is really. Front-and-centre is their defence of the ideology that AI can’t be fully aligned in the lab, use-cases can’t be foreseen, and thus real-world experimentation needs to be iteratively conducted. Release, release, release, slowly but surely. Also featuring strongly on their radar is the protection of children in their training processes (thankfully…), as well as swiftly addressing data privacy concerns, even embedding them directly into the training procedure - take note Italy 😉

2/

Are Meta about to edge their way back into the AI race? I don’t know a great deal about image segmentation and computer vision, but rumours are doing the rounds that its “GPT-3 moment” has officially landed, and it’s come from Meta AI’s research team. It’s called SAM (Segment Anything Model), and it looks like it literally segments any object in any photo or video - with high edge detail too! Just check out that video on their announcement tweet! They’ve even gone and fully open-sourced the dataset, 400x larger than any other ones out in the open.

According to this guy, this is the culmination of years and years of research, and people seem to be getting damn excited. Could this be the foundational development that things like self-driving cars have been looking for?

3/

AI development is literally moving at breakneck speeds these days. (And I mean “literally”, I’ve just pulled a muscle in mine whilst writing these damn picks - I think I’m getting old, sad times .) But anyway, with such fast development has come the need for speedy implementations of business ideas, and a name that’s been popping up time and again is Vercel. We’re frequently hyping up popular new start-ups that are frequently building on the platform, so if you don’t know then get to know! This thread goes through some of the most exciting ones: It features Tome - a company I love; writing assistant companies; autopilot developer bots; chat-with-any-pdf tools, and way more. Read the thread and check Vercel out - maybe your next big idea will be built around it…

🛠️ Cool Tools

Product launches, updates and demos

  • Krea’s AI canvas tool beta is now open for everyone. (link)

  • Audio2Doc - An audio player with a semantic search. (link)

  • DoNotPay’s spam troll - Troll scam and marketing email/text messages by engaging them in an endless A.l. conversation. (link)

  • LLMParser - An OS library that makes it dead simple to classify and extract text with LLMs. (link)

  • Vowel shipped their AI meeting summaries 2.0 and it seems promising. (link)

  • Dalai - The simplest way to run LLaMA on your local machine gets some updates. (link)

  • Ermine - 100% local recording and transcription. (link)

  • Quadratic - The spreadsheet with Python, SQL, and formulas. (link)

  • Flow Coder - ML to source code compiler. Available on request. (link)

  • Cards Against ChatGPT - Negative stories about ChatGPT. Love the design on this one. (link)

  • June AI - Product analytics and SQL queries in simple English. (link)

  • Horace - LLM chatbot server with ChatGPT plugins. (link)

  • Banter AI - Call up your favourite celebrities and talk with realistic voice clones. (link)

  • Whimsy works - Get your kids to love reading with choose-your-own adventures. (link)

🤓 Miscellaneous

News, podcasts, videos, blogs etc

  • Thinking companion, companion for thinking. (link)

  • GPT-4 has its own compression language. (link)

  • How CEOs build with AI - Sean Devine of XBE. (link)

  • Updating iOS for the AI era. Apple has two choices: adapt or die. (link)

  • Replit + Chroma - AI for the next billion software creators. (link)

  • Two AI agents role-play to create a Gradio application by themselves. (link)

  • Behind the curtain - What it feels like to work in AI right now. (link)

  • A new browser that simulates your fantasies of exploring the internet. (link)

  • Latent video diffusion models for high-fidelity long video generation from Tencent. (link)

  • Kandinsky 2.1 - A new open source image generation model with higher benchmarks than SD 2.1 (link)

  • ChatArena - Python library of multi-agent language game environments for collaboration between multiple LLMs. (link)

  • India opts against AI regulation. (link)

  • What is it like to be ChatGPT? (link)

  • GeNVS: Generative Novel View Synthesis with 3D-Aware Diffusion Models. (HUH?!—basically, a computer model that can create realistic 3D pictures from just one photo, and it can show the same scene from different angles or even make a 3D video.) (link)

🎓 Learn

How-to’s and resources

  • LinkedIn now has a ChatGPT for business mini course in their learning section. (link)

  • Build your own ChatGPT AI assistant on Telegram. (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.

📰 Unclassifieds

Short, sponsored links 

  • Unleash efficiency🌪️ using Bitskout! Add no-code data extraction from docs & emails to Asana, Monday, Zapier, and other workflows. #ProductivityUnlocked🔓 (link)

  • Free AI Investing Book - Revealing the top 7 AI companies to invest in for long-term returns. (link) pss: use your discretion while investing.

  • Langchain is hosting a webinar on Wednesday, 12th April, to talk about autonomous code agents and experiments from top builders. (link)

  • Chroma beats us to the first-ever ChatGPT Plugins hackathon😅. Happening this weekend in SF. (link)

Have a product, service, job, event, newsletter, app, book, movie, tool, or anything you'd like to share with over 60k subscribers? 
Advertise

🖼 AI Images of the day

Funny memes and pics from around the web

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

How was today's email?

Login or Subscribe to participate in polls.

Reply

or to participate.