😺 Bye-bye keyboards

Welcome, humans.

If you’re looking for the #1 reason why we shouldn’t have all software run by AI, this is probably it:

Nothing like waking up at 9am to a $30K bill…

This was definitely a glitch… the best theory as to what happened is that somebody added a few extra zeros to the cost of the text to speech API…

In all honesty, this was probably a human dev. But knowing OpenAI, would you really be THAT surprised to find out they’ve got some AI Agents running amok on the backend?

Here’s what you need to know about AI today:

• We break down the big news from OpenAI’s DevDay 2024.
• Anthropic hired ANOTHER former OpenAI co-founder.
• Pinterest launched AI tools for advertisers.
• NVIDIA released NVLM 1.0, a new open-source AI model family.

Advertise in The Neuron here.

OpenAI’s DevDay made one thing clear: the “computing interface” of the future will be voice.

OpenAI loves to time its drama around its DevDays. Or maybe it’s the other way around?

Either way, after a week of wild headlines for the soon to be* $150B startup, OpenAI held its DevDay 2024 on Tuesday, and, among many other things, it showed how you can program drones from scratch:

The drone literally streams video of the crowd, live. Here’s more about this.

The big news of the day is the company’s Realtime API, which lets people build products using the new technology that powers Advanced Voice Mode (the new talking ChatGPT that sounds more like a real human).

The real question is: how much does it charge for awkward silence?

Where does this go from here? Eventually, OpenAI will add vision and video. Which leads us to the OTHER big news of the day:

Introducing Copilot Vision

Microsoft announced "Copilot Vision" for Edge, allowing Copilot to “see” web pages on your screen and talk to you about them.

And that’s where this is all going: The end of the keyboard as the dominant way we communicate with computers.

Google has voice AI, Meta has voice AI, Microsoft has voice AI, and soon, every startup with an OpenAI account will have voice AI. It’s like the Oprah meme: You get a voice AI! You get a voice AI! Everyone gets a voice AI!

And once that happens: it’s going to become A LOT more common to talk instead of type.

FROM OUR PARTNERS

Quickly build, customize, and deploy generative AI models

AI applications trained on large language models (LLMs) can boost enterprise task efficiency, but they often struggle with real-time events and specialized knowledge, which can lead to inaccuracies.

See how you can deliver higher accuracy and more efficient responses for generative AI applications with retrieval-augmented generation (RAG).

Watch Now

Around the Horn.

Microsoft has around $108B worth of AI data centers in the pipeline …which is about $100B more than it had two years ago.

• Here’s a good recap of everything else OpenAI announced (NotebookLM version also available). The coolest things are probably…
  • Image fine-tuning, which lets you make a model to analyze ultrasounds, for example.
  • Fun fact: Jason Wei, an OpenAI researcher, says you should use o1-mini when doing math and coding, and use o1-preview to find inaccuracies in data or conduct hard science research and legal domain reasoning.
  • Chief Product Officer Kevin Weil said agents will become mainstream by 2025—and in a Q&A, Sam Altman said by 2030, we’ll “take for granted” having AI agents who work on tasks for a month at a time.
• NVIDIA released NVLM 1.0, an open-source AI model family that rivals GPT-4 in performance on visual and language tasks (check the 72B version here).
• Anthropic (maker of Claude) hired yet another OpenAI cofounder, Doug Kingma, after it hired John Schulman in August. Fun fact: Anthropic CEO Dario Amode was also formerly VP of research at OpenAI! *~The More You Know~*

Treats To Try.

1. *Dashboards & business insights, in seconds. Drowning in data? Zebra AI is your lifeline. This easy-to-use AI tool provides instant dashboards and insights from your most important data. No more endless spreadsheets or complicated software. Try Zebra AI Free – Get instant insights!
2. Bing Generative Search from Microsoft is now available in the Bing search engine—write “Bing generative search” to test it out.
3. Runware provides ultra-fast, low-cost images—you can generate up to 10K images for free (it generated us a cat eating ramen in < 1 second!).
4. VideoSDK helps you build and integrate characters that can interact with you in real-time through speech, vision, and task automation.
5. Graphite Reviewer gives you instant, codebase-aware AI feedback on every pull request to catch bugs and enforce best practices before human review.
6. Temperstack automates tedious tech tasks to help keep your software systems running smoothly.
7. Inbox Zero helps you clean up your inbox by automating email tasks and blocking cold emails.

See our top 51 AI Tools for Business here!

*This is sponsored content. Advertise in The Neuron here.

Where Do You #Neuron?!

We’re running low on entries for our “Where do you #Neuron” section!

If you want this part of The Neuron to stick around, fill out this form with a photo of where you read The Neuron, along with some witty commentary. You might just get featured next week in front of our nearly half a million readers!

Harshit from Spain: Just levelling up my AI knowledge

…This could be you, but you didn’t send us where you #Neuron!

A Cat's Commentary.