Gm. If you spend any amount of time on TikTok, then you are probably familiar with the question sweeping the internet: how often do you think about the Roman Empire? Personally, I think about it at least once a week, which made my girlfriend stare at me in confusion for a solid 30 seconds. Make sure to respond to today’s poll so that I can show her I’m not alone here.

— Stephen Flanders

🏠 AI

Gemini Grows Closer

We’ve spent a lot of time in recent months covering Google’s Gemini AI model, and for good reason:

It could very well exceed GPT-4 in performance due to Google’s treasure trove of advanced GPUs.
It combines the text powers of a GPT-4 with the image powers of a Midjourney or Stable Diffusion.
It is the next shot fired in the increasingly entertaining and competitive AI wars.

Thankfully, some very exciting news is coming out of The Information: Gemini is close. So close that Google is now allowing a small number of companies to give it a spin.

And the early returns for Gemini appear to be positive, with one early tester saying that Gemini generates fewer incorrect answers than GPT-4 due to Google’s disgusting amounts of proprietary data (read: data from Search).

This is nothing but good news for Google, which has a lot riding on Gemini’s success:

They’ve poured a metric ton of resources into developing Gemini.
They are banking on it to power all of its AI products, including the chatbot Bard and a GitHub Copilot competitor.
They also want to use it to bring AI to “regular” products, like Google Cloud.

If you didn’t already know, now you do: the AI race is fully underway.

The Big Picture: Gemini is a big moment for Google, but competition is fierce with GPT-4, Claude, and Llama all lurking about.