• Ben's Bites
  • Posts
  • ChatGPT API is here and it's 10X cheaper

ChatGPT API is here and it's 10X cheaper

PLUS: Youtube to add AI features

Hey folks, today we’ve got YouTube entering the AI space, OpenAI’s APIs, behind the scenes of OpenAI’s CTO, and famous paintings with real characters.

Let’s get to it.

🤌 Our Picks

Highlights if you've only got 2 minutes

Week by week we update you with a new big tech company wading into the AI scene, and today’s turn is YouTube. Given that quite recently a young chap called me an “old man” for watching YouTube shorts, perhaps this is just the move they need… New CEO Neal Mohan has announced that in the coming months, generative AI will be available to content creators, where they’ll be able to do things like swap outfits and put themselves into AI-generated fantastical settings. With ad revenue dropping 8% vs last year, this will be a welcome development for parent company Alphabet.

For YouTube updates going beyond AI, check out the link above.

🚨OpenAI technology update 🚨 

Don’t worry, they haven’t updated their tech to bring us a step closer to an AGI that’s going to imminently end the world, as Eliezer Yudkowsky and Gary Marcus might be concerned about. Instead, they’ve put their heads together to reduce costs by 90%, and passed it straight to the end-user at a 10x cost reduction. The big move involves easy-to-use (I think?) APIs for both ChatGPT and Whisper (for those who don’t know, Whisper is their speech-to-text model). This is a move more in-line with OpenAI’s founding principles of openness and accessibility, working on optimization to reduce costs rather than leaping towards the next best chatbot.

I’m also giving $1000 each for the most impressive thing built with the new ChatGPT and Whisper API (two separate prizes). Submit your project to [email protected] by Sunday, 5th March, Midnight PT.

When things are going really well, there’s often someone behind the scenes, taking less of the spotlight, but nonetheless driving things forward. In this case, let’s take a moment to recognise the contributions of Mira Murati, CTO of OpenAI.

She’s been at the forefront of developments in OpenAI’s technology, bringing up questions about the trade-offs between creativity and accuracy. She holds the view that a GPT-4 release will be a significant step for AI, but a little less hype around it would probably be a good thing. Take some time to read the article - it’s a good read! Unpaywalled version here.

I always love it when there’s something fun and light-hearted to fill the fourth pick with. This guy has taken some of society’s best known paintings and seen what generative AI can muster up by using the title as a prompt. And he’s really brought them to life - it’s pretty cool. For me, the funniest is “Self-portrait, 1889 by Vincent Van Gogh” - it’s so real and just so damn normal!

(Sponsor)

Hey Zoë, show me subscription revenue by marketing channel in the state that Lebron James was born, in the quarter that Apple released iOS update 14.5.”

When everyone on your team can ask questions and get the answers they need to do their job, better things will happen for your brand.

Zenlytic combines the power of Open AI’s GPT-3 with the accuracy of their own proprietary data model to empower humans with data.

🛠️ Cool Tools

Product launches, updates and demos
  • Dashworks - The AI search assistant for work, now in Slack. (link)

  • Plan my trip ASAP - Get a custom itinerary for your next trip in minutes. (link)

  • Converse by B7 Labs - Enter a webpage URL to summarise. (link)

  • Rubberduck - ChatGPT for Visual Studio Code. (link)

  • Startup Jobs introduces sample interview questions and answers to help you make a great impression. (link)

  • Chunky - Train your own smart assistant by simply using Google Sheets. (link)

  • AuralByte - Audio to video conversion tool to expand their audience on short video platforms. (link)

  • LucidPic - Generate quality stock photos of people that don't exist, in seconds. (link)

  • Goldie - AI to execute sales plays on your behalf, so you can close more deals in less time. (link)

  • Demotime - Magic highlights after every demo. (link)

  • Rubbrband - CLI that enables training of the latest ML models in a single line of code. (link)

  • Outset - Create a slackbot trained on your company info in less than a minute. (link)

  • ChatGPT wrapper to use with GPT index. (link)

  • Voice Design from Eleven Labs - The first generative AI for voice creation. (link)

  • What on Earth - Search engine powered by ChatGPT. (link)

🤓 Miscellaneous

News, podcasts, videos, blogs etc
  • Could ChatGPT help “Citizens Advice Scotland”? (link)

  • D-ID unveils new chat API to enable face-to-face conversations with an AI digital human. (link)

  • YouTube’s new leader teases AI tools that can swap creators’ outfits and locations virtually. (link)

  • Marc Andreessen joining Character AI board + other hot AI fundraising rounds afoot. (link)

  • OpenAI is now everything it promised not to be - corporate, closed-source, and for-profit. (link)

  • Don’t look up - The AI edition. (link)

  • Everything OpenAI shipped yesterday. (link)

🎓 Learn

How-to’s and resources
  • Question & answer using LangChain JS. (link)

  • Training your own custom Cohere models without knowing ML. (link)

  • How to turn a voice recording into a to-do list. (link)

  • Full tutorial to use ChatGPT API with Colab. (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

Published research papers
  • Teaching old labels new tricks in heterogeneous graphs. (link)

  • Diffusion probabilistic fields from Apple. (link)

  • Collage Diffusion creates globally harmonised images from complex compositions of several objects. (link)

  • Mask shift restoration - Dealing with images of arbitrary sizes during diffusion. (link)

  • StraIT from Google - Non-autoregressive generation with stratified image transformer. (link)

  • Reward design with language models. (link)

  • StrucTexTv2 from Baidu - Masked visual-textual prediction for document image pre-training. (link)

  • Are more layers beneficial to graph transformers? (link)

  • MuAViC from Meta - A multilingual audio-visual corpus for robust speech recognition and robust speech-to-text translation. (link)

  • Efficient and explicit modelling of image hierarchies for image restoration. (link)

  • S-NeRF - Neural radiance fields for street views. (link)

  • WhisperX - Time-accurate speech transcription of long-form audio. (link)

📰 Unclassifieds

Short, sponsored links 
  • ChatGPT hackathon in SF on Sunday, 5th March, sponsored by Replit. RSVP here >>

  • Research opportunity: Outset is offering a $10 gift card to founders, engineers, and product managers actively integrating LLMs into their product in exchange for a 25-minute conversation.

  • Clickup is hiring a staff AI & ML engineer. Learn more »

  • Bandalab Technologies is hiring a music production operator.

Have a product, service, job, event, newsletter, app, book, movie, tool, or anything you'd like to share with over 37k subscribers? 
Advertise with us | Job board | Join Community

🖼 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.

Join the conversation

or to participate.