THE ONE IMPORTANT STORY

🤖ChatGPT’s biggest update since GPT-4

Sam Altman took the spotlight at OpenAI's Developer Day in San Francisco with a significant announcement: the introduction of GPT-4 Turbo. This new model can take an input as large as a 300-page book in one prompt. It's even bigger than competitor Claude 2’s context window. Summarizing long documents will be a breeze from now on.

And there's more. Turbo's got its knowledge dialed up to April 2023, leaving the old September 2021 cutoff in the dust. It's a big leap, for sure.

Unlike before, when you had to jump between modules like DALL-E 3 or searching with Bing, this new model gets what you're aiming for straight away making ChatGPT truly multi-modal.

GPT the GPTing GPTers

Yep they did it. If it wasn’t confusing enough to call their chatbot ChatGPT and the models underneath GPT 3.5, GPT 4, GPT-4 Turbo etc., now ChatGPT has a feature called GPTs.

GPTs are tailored versions of ChatGPT that you can build by programming it in natural language, aka by talking to it. With the new massive context window, you can really tune it to your needs.

GPTs are meant to be helpful at specific tasks, in my case I will probably use it to help me create this newsletter, by automatically checking the grammar and correcting mistakes while staying in line with my style. Other examples include: fitness coach, tech advisor, sticker generator, etc. Check out the video in the tweet about how it works:

Developers will be able to connect GPTs to databases, plug them into emails, or make them your shopping assistant as well.

GPTs can also take small actions. In a live demo they presented a usecase where the GPT looked through a calendar to find conflicting events, and sent a message in Slack.

You can build a GPT, publish it in the brand new GPT store (coming later this month) and make money selling it. Plus and Enterprise users can start creating GPTs this week at http://chatgpt.com/create.

New tools to build apps on

On the developer's front, OpenAI has busted out some shiny new tools. They're handing over the keys to their text-to-speech model that's basically the MVP of AI voices and also to Vision and DALL-E 3, which means app developers are about to get real crafty.

Whisper V3 is also stepping into the ring, putting iPhone dictation to shame with its ability to understand your mumbles and transform them into text. Sam Altman claims it has been dramatically improved in many languages. This is essential for apps that aim to translate speech in real time. Babel fish soon?

Assistant API

One of the most awe-inspiring part of the keynote was a presentation of how ChatGPT can be integrated into websites now through Assistant API. This is a first step from OpenAI towards AI agents, that can perform complex tasks on their own. Worth watching it, to see how websites of tomorrow might look like.

With the new model they were able to slash the API prices as well, on average it’ll be 2.75x cheaper to use it for third party apps than before.

In classic OpenAI style, they're not just walking the AI walk; they're out there breaking the sound barrier. With their focus on making AI useful and a whole lot easier to weave into our daily life, their building a strong user and developer base that competitors will struggle to compete with. What’s your move Google?

ONE MORE THING

Check out this early test of the new GPT function

Click on the tweet for the video.

⌚ If you have one more minute

• I’ve Worked With Generative AI for Nearly a Year. Here’s What I’ve Learned. - The WSJ

• This AI Bot Fills Out Job Applications for You While You Sleep

AI Art of the Day 🎨

After the latest OpenAI announcements the memes already started rolling in about how OpenAI wipes out parts of the AI startup industry by releasing new features, like reading 300 page documents, and customizable chatbots.

🏄🌊🏄🌊🏄🌊🏄🌊🏄🌊🏄🌊🏄🌊🏄🌊🏄🌊🏄🌊🏄🌊🏄

That's all for today, folks!

If you enjoyed this, please share this hand-crafted newsletter with a friend. It would make this writer's day!Huge Update