Welcome, humans.
AI engineers are like bears—they both hate being disturbed during their focused periods. And like bears hibernating, AI engineers sometimes retreat into intense work phrases before emerging with something that’ll shake us to our core.
This week, the metaphorical bears have made their presence felt. Let’s dig in!
Here’s what you need to know about AI today:
- A new Gemini reportedly supports up to 700,000 words.
- OpenAI unveiled Sora, a remarkable text-to-video model.
- OpenAI is working on a competitor to Google Search.
- Congress struggles to pass significant AI regulations.
Google and OpenAI keep trying to one-up each other.
For most of 2023, OpenAI's ChatGPT-4 reigned as the superior AI model.
Last week, Google finally clapped back with Gemini Ultra, claiming it surpasses ChatGPT-4. We maintain ChatGPT-4's superiority, yet nuances exist (see a 10,000-word comparison here).
Two days ago, The Information revealed OpenAI's plans for a Google Search alternative, potentially resembling Perplexity.
And just yesterday, Google announced that it’s already prepping Gemini 1.5 Pro, a new model that’s rivals 1.0 Ultra, and can process 1M tokens—and potentially up to 10M.
To clarify, 1M tokens could mean analyzing 700,000 words, 30,000 lines of code, or 11 hours of audio, summarizing, rephrasing, and quoting the content.
This. Is. Absolutely. Nuts.
Gemini
Imagine being able to dump your entire inbox into ChatGPT and ask questions. Or all of your client/project notes? Or an entire codebase? The possibilities are endless.
In theory.
Google has a habit of overpromising and underdelivering. Yet if these numbers materialize, it would be a MASSIVE upgrade to current chatbots, and possibly eliminate the need for RAG (explainer).
BTW: you can tinker with Gemini 1.0 Ultra in AI Studio and Vertex AI.
But what came yesterday afternoon from OpenAI was the craziest thing of all (continue reading)…
FROM OUR PARTNERS
Apply to speak at THE prestigious AI Conference!
The AI Conference is making a grand return this September after a phenomenal success last year (Neuron readers attended & raved about it).
It's two days packed with the brightest minds in AI, discussing AI models, infrastructure, alignment, ethics, and all the other topics we love to geek out on.
The AI Conference is looking for engaging and deeply technical speakers to share their expertise this year.
The application deadline is just two weeks away—seize this life-changing opportunity and apply now!
OpenAI's text-to-video model is groundbreaking.
OpenAI announced a text-to-image AI model,“Sora,” that turns prompts into detailed 60-second videos, and the results are absolutely stunning.
Like, STUNNING.
Take a look at this video that seems like it's straight out of an animated cartoon:
OpenAI
And this chess-playing monkey:
Or this zoo…
Or our top pick (for obvious reasons):
Why it matters: While not Netflix-quality quite yet, this is just version #1 (consider DALL·E 3 now to where it was 15 months ago).
Sora makes us believe that AI-generated movies really are possible. Other, more immediate use cases could include B-roll / stock video, video ads, and video editing—all using AI!
And Sora doesn’t just do text-to-video—it can:
- Bring still images to life as video.
- Extend videos in either direction.
- Edit videos by altering styles / environments.
- Merge two videos.
Pinch us: how crazy is it that we’re seeing this just over a year after ChatGPT?
More details to come next week—for now, access is limited to select users and you can read Sora’s technical report here if you’re nerds like we are.
Around the Horn.
- OpenAI identified and shut down five state-affiliated accounts using AI for harmful cyber activities.
- OpenAI’s Startup Fund is owned personally by Sam Altman.
- Lambda, which sells GPUs for training AI, has become the latest AI unicorn after raising at a $1.5B valuation!
- The likelihood of Congress enacting "meaningful AI legislation" soon looks bleak.
- Sam Altman isn’t actually trying to raise $7T for AI chips.
Intelligent Insights.
- Confessions of an AI Clickbait Kingpin (link).
- Meet the Pranksters Behind Goody-2, the World’s ‘Most Responsible’ AI Chatbot (link).
- As more governments look at AI rules, Accenture tests software to help companies comply (link).
- Designing Chips Is Getting Harder. These Engineers Say Chatbots and AI Can Help. (link).
- Adobe’s Very Cautious Gambit to Inject AI Into Everything (link).
Friday Faculty.
- Frontend Engineer @ ElevenLabs (Europe/East Coast).
- Product Manager @ Stability AI (North America).
- 13 AI jobs at LinkedIn (Sunnyvale/hybrid).
- 1,000+ AI jobs at Nvidia (everywhere!).
- 86 open roles in all areas @ Scale AI.
Search from over 10,000 AI roles on The Neuron’s Jobs Board.
A Cat's Commentary.
That’s all for today, for more AI treats, check out our website.
Get your brand in front of 330,000+ professionals here.
See you cool cats on Twitter: @nonmayorpete & @noahedelman02