An OpenAI update

Ahhhh Ben's Bites, the smorgasbord of AI nibbles. We're inching closer to GPT-4, any bets on when we will see it happening? I think before end of year.

We're also seeing a lot of samey-samey products showing us that speed is becoming a diminishing advantage in the age of AI. Great products are still tough to build.

Let's dive in.

🤌 Ben's Picks

  • OpenAI Text-Davinci-003 has been released. OpenAI says that it produces higher-quality writing, more complex instruction and better long-form content. It is still trained on the same data as text-davinci-002 but tweaked differently. Is this a little taster for GPT-4? Will it be released at NeurIPS?!

    • A few observations on where it’s improved and not vs 002 (link).

    • It can write rhyming poems now. (link)

    • Here's a collab notebook to easily run comparisons against text-davinci-002.

  • Modal Podcast Transcriber - Transcribe any podcast episode in just 1-2 minutes. This tool is pretty incredible. It says what it does on the cover! (link)

  • GPT Index is a project consisting of a set of data structures that are created using GPT-3 and can be traversed using GPT-3 in order to answer queries. (link) A short thread on its capabilities. (link)

🛠️ Cool Tools

  • Craft - a document application, now has AI baked in. (link)

  • Pictoral - Effortlessly create graphics for your web applications. (link)

  • Storywizard - Create incredible children’s stories in no time using AI. (link)

  • - Create stories + illustrations in seconds with AI. (link) (yep, two similar products with one similar yesterday too!)

  • Personalized AI Summary by Glasp - Summarize what you learn with AI. (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.

🔬 Research

  • A technique for regional control in text-to-image generation called ReCo enables the generation of high-fidelity images with precise control over the content in a specific region. (link)

  • An efficient design of Transformer-based models for multivariate time series forecasting and self-supervised representation learning. (link)

  • Can conditional generative modelling be used to address the problem of sequential decision-making? (link)

  • Corgi, a novel method for text-to-image generation based on a shifted diffusion model. (link)

  • A framework for modelling categorical data with diffusion models that are continuous both in time and input space. (link)

  • A unified multimodal generation model that can conduct both the "modality translation" and "multi-modality generation" tasks. (link)

  • This article investigates the hypothesis that transformer-based in-context learners implicitly implement standard learning algorithms by encoding smaller models in their activations. (link)

  • Action-GPT - a plug-and-play framework for incorporating large language models into text-based action generation models. (link)

  • A fast frame-level model for anomaly detection in video, which uses knowledge from multiple object-level teacher models. (link)

🤓 Everything else

  • Hugging Face Inference Endpoints guide on how to deploy Stable Diffusion to generate images for a given input prompt. (link)

  • Stable Diffusion 2.0 and the importance of negative prompts for good results. (link)

  • GPT-4 predictions. TLDR: Paras thinks that GPT-4 will be able to combine language models and code, which could mean it will be capable of getting the answers it needs by generating code and executing it. Time will tell! (link)

  • How Hugging Face improved text generation performance with XLA. (link)

  • A LightningAI component that lets you run your own Stable Diffusion 2.0 server in ~5 min. (link)

  • Community models are now available on Stable Diffusion API. (link)

  • LangChain added the concept of a Memory Module. It loads things from memory before passing user input to the chain and saves things to memory after the chain is finished. (link)

  • Vox2 - a Voxel-ish Stable Diffusion model. (link)

  • AI invents millions of materials that don’t yet exist - ‘Transformative tool’ is already being used in the hunt for more energy-dense electrodes for lithium-ion batteries. (link)

  • Google partners with a med-tech company to develop AI breast cancer screening tools. (link)

🧑‍💻 Who's hiring in AI

VEED.IO - Simple Online Video Editing. VEED is hiring AI / ML engineers to level up its creative toolkit and make it more magical.

Buildspace - where builders, build! They're looking for an ML/AI instructor to build their new course.



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

or copy/paste this link:



How was today's email?

Login or Subscribe to participate in polls.


Join the conversation

or to participate.