PLUS: Reid Hoffman's new AI startup
Hey folks, this is Ben's Bites your daily guilt-free fix of AI news. Or as it was described to me yesterday, a summary email for AI twitter.
When an ex-founder starts a daily newsletter, you have to understand that's a sign of great respect in their culture. 😂
Let's get to it.
🤌 Ben's Picks
Have you heard about Anthropic? Well, they're in a closed beta right now with a new AI assistant called Claude.
Don't feel bad, I haven't got access yet either 😡. It's supposed to be different than ChatGPT (does it get overwhelmed every 10 mins?!!). Claude knows more about its creators and ethical principles. And it can process more text too. They're using something called "Constitutional AI" to make sure it doesn't give harmful outputs. They even did a test where they compared Claude to ChatGPT, and Claude was more aware of its own limitations. Like, it knew it couldn't take the cube root of a 12-digit number (OMG same).
There are a bunch of examples comparing the outputs of ChatGPT and Claude which are worth a look. And it looks like it's a serious contender to ChatGPT.
Claude vs ChatGPT - Talents and limitations of ChatGPT's first real competitor. (link)
Now this isn't your average 2-hour YouTube watch. This is a lecture from Andrej Karpathy. In it, he goes through how to build a Generatively Pretrained Transformer (GPT) following the pivotal paper "Attention is All You Need".
Let's build GPT: from scratch, in code, spelled out. A tutorial form Andrej Karpathy. (link)
Reid Hoffman has done a lot of things; angel in Facebook, co-founder Linkedin, Investor in OpenAI, and on and on I could go! But I wont. He sits down with host, Elad Gil, to chat about AI, Big Tech and Startups. Reid's also working on an AI startup of his own 👀 .
Fireside Chat: Reid Hoffman, w/ Elad Gil on AI, big tech, & startups. (link)
🛠️ Cool Tools
- Reflect AI - Use GPT3 in your notes. (link)
- MagicForm - Train your very own AI salesperson in less than 3 minutes to sell your products. (link)
- PapersGPT - Feed any scientific papers into GPT to give it deep customised knowledge. (link)
- Arxiv Feed - The easiest way to keep up with AI research. (link)
- LangChain Chat - Chatbot grounded to answer questions about LangChain's documentation. (link)
- Kili: A platform to craft your own, personalised AI assistant with zero code. (link)
- Photoshop plugin for Stable Diffusion with Automatic1111 as backend. (link)
- Simplified - All-in-one design platform to create social posts, presentations and turn them to video. (link)
- Komo - Search engine powered by generative AI. (link)
- Cursor - A brand new IDE with LLMs, built from the ground up. (link)
- Getty Images is suing the creators of AI art tool Stable Diffusion for scraping its content. (link) Statement from Getty Images. (link)
- CNET's article-writing AI is already publishing very dumb errors. (link)
- Notion, AI, and me. (link)
- Hey Contigo - Our experiment building a real-time, interactive, agentic voice AI application. (link)
- Behind the scenes of the first NeRF in a TV commercial! (link)
- The state of generative AI. (link)
- How to build a GPT3-powered app on WordPress without code. (link)
- An unofficial PyTorch implementation of the audio LM VALL-E. (link)
- WorldCloner - An end-to-end trainable neuro-symbolic world model for rapid novelty adaptation. (link)
- OMMO dataset - A new large-scale outdoor multi-modal dataset for NeRF-based tasks. (link)
- A novel audio-driven approach to generate realistic lip-synchronised 3D facial animations from audio inputs. (link)
- Use of self-supervised learning pre-training on noisy web image-text paired data. (link)
- RxRx1 dataset - A biological dataset designed to study batch effect correction methods. (link)
- Laser: Latent set representations for 3D generative modelling. (link)
- Ankh: Optimised protein language model unlocks general-purpose modelling. (link)
- Infusing commonsense world models with graph knowledge. (link)
- T2M-GPT: Generating human motion from textual descriptions with discrete representations. (link)
- GLIGEN: Open-set grounded text-to-image generation. (link)
- Msanii: High fidelity music synthesis on a shoestring budget. (link)
- Diffusion-based generation, optimization, and planning in 3D scenes. (link)
- Hypothetical Document Embeddings~(HyDE) - Precise zero-shot dense retrieval without relevance labels. (link)
- Audio generation with diffusion. (link)
- Hassan builds impressive AI projects and writes a weekly newsletter breaking down how he built them.
- Know how to build with Zapier? I'm (me, Ben) looking to hire someone casually to help come up with ideas, build them and document how others can do the same. Building AI workflows as open-source experiments. Interested? Get in touch.
Want to reach over 20k subscribers? Get a short unclassified ad included here.
🖼 AI images of the day
🤗 Share Ben's Bites
Send this to 1 AI-curious friend and receive my AI project tracker database!
or copy/paste this link: https://www.bensbites.co/subscribe?ref=PLACEHOLDER
⭐️ How did we do?
How was today's email?