Microsoft Teams + ChatGPT for $7

Hey! Welcome to Ben's bites - coming in fresh from Austin - it's time to laugh, cringe, and enlighten yourself with today's top drawer in AI.

Have a product, service, job, event, newsletter, app, book, movie, tool, or anything you'd like to share with over 20k subscribers? Check out our sponsor options.

Join Discord

🤌 Ben's Picks

1/

Microsoft releases Teams Premium using ChatGPT technology

ChatGPT has officially just swooped into the mainstream with Microsoft’s release of Teams Premium, with the AI technology being fully embedded! Now I’ve just gotta convince my boss to fork out $7… The tech comes with automatically generated meeting notes, recommended tasks, and personalised highlights, and looks pretty damn slick to boot.

2/

ChatGPT releases $20/month subscription plan

Getting “servers busy” messages on ChatGPT when you most want to use it? Well, wave that frustration goodbye for $20—gotta make that $10bn back for investors. (But yeah, you’re right, the $42 pro fee that was doing the rounds indeed looks like some kind of weird error.)

Besides, this seems like a pretty reasonable fee for something that can help you with basically anything and only lies sometimes. Sam Altman has even hinted at upcoming, potentially cheaper, pro plans. It’s a shame that I’m based in England, and it’s only available in the US…

3/

AI chatbots can actually make us more skilled

With chatter circulating that over-use of chatbots can dumb you down, this article actually makes a pretty compelling case for why using AI can make you smarter. (And no, sitting there like a potato all day letting ChatGPT do your dirty work is not how it works.)

They show a pretty awesome example of getting a chatbot to create scenarios to respond to, giving feedback on their responses, and then discussing how they can do better. Real, mistakes-driven practice—which is how anyone can get better at anything.

4/

Git commands in plain English

For some of you seasoned developers, it may come as a surprise that a lot of people really struggle with using git from the command line. Personally, I remember a time when it drove me absolutely crazy, and would have loved nothing more than this tool. Type what you want to do into the command line in plain old English, and out comes the git command. Wicked!

🛠️ Cool Tools

• Video Summarization by Cloudinary Labs - Automatically summarise long videos to engage your audience with bitesize clips. (link)

• LinkOut - Generate a personalised LinkedIn cold message with a single click. (link)

• No Code Tutorial Search - Find the best no code education on YouTube. (link)

• GraphGPT - Extract knowledge graphs from unstructured text and visualise them. (link)

• Latitude - Write faster SQL with AI. (link)

• Gradio QandA - Answer your Gradio questions accurately along with links. (link)

• GPTwitter - The first AI generated personalised social media platform. (link, blog here)

• Symbiotic FYI - Your AI knowledge assistant. (link)

• Trigger - Effortless automation built for developers. (link)

🤓 Miscellaneous

• Gaslighting and reality in AI. (link)

• Flan-T5 beats all public models for instruction tuning. Data and methods now released by Google. (link, author’s thread)

🎓 Learn

• How to run GTR T5 LLM offline on a Macbook? (link)

• Introduction to multi-agents and AI vs AI challenge. (link)

• How to build a simple semantic search engine. (link)

🔬 Research

• FROMAGe - A model that efficiently bootstraps frozen LLMs for images. (link)

• Epic-Sounds - A large-scale dataset of actions that sound. (link)

• Jointist from ByteDance - An instrument-aware framework capable of transcribing, recognizing, and separating multiple musical instruments. (link)

• Improving few-shot generalisation by exploring and exploiting auxiliary data. (link)

• Clinical Decision Transformer - Intended treatment recommendation through goal prompting. (link)

• mPLUG-2 - A new unified paradigm with modularized design for multi-modal pretraining. (link)

• Learning universal policies via text-guided video generation. (link)

• Transformers meet directed graphs - from Deepmind. (link)

• Synthetic Prompting by Microsoft - Generating chain-of-thought demonstrations for large language models. (link)

• Large language models can be easily distracted by irrelevant context. (link)

• In-context retrieval-augmented language models. (link)

• Debiasing vision-language models via biassed prompts. (link)

• The power of external memory in increasing predictive model capacity - from Google. (link)

• CFM - Conditional flow matching, a simulation-free training objective for CNFs. (link)

• Two for One by Microsoft - Diffusion models and force fields for coarse-grained molecular dynamics. (link)

📰 Unclassifieds

• Twofer Goofer: a new daily word game. A piece of AI-created art celebrates each day’s whimsical rhyming pair. Solve your first puzzle now>>

Have a product, service, job, event, newsletter, app, book, movie, tool, or anything you'd like to share with over 20k subscribers? Check out our sponsor options.

🖼 AI images of the day

Thomas the tank

🤗 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

⭐️ How did we do?

⭐️ REAL REVIEWS