🏠 The Meta And Apple AI Edition
Big moves from big companies
Gm. Brett, again! I recently shared a recommendation for this amazing service I've been using - Athena - and there was a TON of interest from you all. Athena basically sets you up with an amazing (yet very affordable) full-time executive assistant that helps you manage everything from your email to meeting scheduling and more. Our EA Maria even helps onboard advertisers for Homescreen!
Anyway, wanted to remind you all of this amazing deal they're offering Homescreen subscribers - if you sign up in the next 2 weeks, you skip the waitlist (it’s long!) and get a free month along with personalized onboarding.
Check it out here.
BROUGHT TO YOU BY:
OPT Or Llama?
Meta has been able to carve out a distinct niche for itself in the AI race as the open-source king.
Their models Llama and Llama 2 aren’t just widely praised, but they are in many cases as good or even better than the closed-source alternatives from OpenAI, Google, and Anthropic. So good that it is even used to build other open-source models, like Code Llama and Vicuna.
However, building it didn’t come without consequences for Meta.
Two Teams, Limited Resources
Meta’s AI efforts centered on two different teams building two different models: the Llama team and the OPT team.
As we all know, building AI models requires a hell of a lot of computing power. And both sides want as much of it as possible.
But, unfortunately, Meta doesn’t have an unlimited amount of computing power. They had to divvy it up, which led to friction between the two teams, especially among the Llama team which typically received the short end of the computing power stick.
By the time the OPT model was abandoned and Llama was released in February, Meta had lost roughly half of OPT’s original 19 researchers and more than half of Llama’s original 14.
A New Rivalry
Even with the victory by Llama, the drama isn’t over at Meta.
The newest battle for computing power comes from Meta’s commercial product-oriented team and Meta’s research team.
And, judging by the laying off of the team behind its revolutionary protein folding model, it looks like the commercial product team is winning.
The previous Meta internal battle led to a huge talent exodus, which resulted in the birth of startups like Mistral AI.
It’ll be interesting to see if this battle does the same.
The Big Picture: It gets easier and easier to see why AI chipmaker Nvidia is now a trillion-dollar company.
Apple Boosts Its AI Spending
One of the biggest surprises of the current AI moment is how little Apple has been involved.
Typically on the cutting edge, Apple hasn’t released a single new consumer-facing AI product. No chatbots. No text-to-image generators. Nothing.
But that might be changing soon.
According to The Information, Apple is pumping millions of dollars a day into developing its conversational AI capabilities.
Some of the products being built include:
The ability for people to tell Siri to accomplish tasks that include multiple steps, such as creating a GIF using their last five photos and texting it to a friend.
An Apple Chatbot that could interact with customers who use AppleCare.
Ajax GPT, it’s counter to OpenAI’s GPT model.
A software that generates “images, videos, or 3D scenes”.
Multimodal AI, which is AI that can recognize and produce images, videos, and text.
Millions of dollars a day sounds like a lot, and it is a lot, but with Apple’s virtually unlimited resources and top-notch team, we couldn’t be more excited to see what they can cook up.
The Big Picture: AI is going to get a whole lot more interesting once Apple fully gets in the game.
What else is going on?
Stat: 83: How many Falcon 9 or Falcon Heavy missions SpaceX has launched in the last year, a new record.
Rabbit hole: Cars Are The Worst Product Category We Have Ever Reviewed For Privacy (Mozilla)
🤑 FUNDING FRIDAY
If you want a boring list of startup raises, look away now. If you want the 3 coolest raises of the week, then you’ve come to the right place.
🎁 FINAL JEOPARDY!
This Final Jeopardy! question stumped every contestant. Can you get it right?
Category: Medieval Asia
Clue: Though it means “one to serve” in medieval Japan, it was a property holder who received rent from serfs.
🛠️ FOUNDERS CORNER
The best resources we recently came across that will help you become a better founder, builder, or investor.
💉 HelpKit AI turns your Notion database into a fire customer support assistant.
🎧 AudioNotes converts your random notes into succinct text summaries.
💎 Merlin uses AI to conduct your qualitative research.
💉 DOPAMINE HIT
How is he…just sitting there?
What an incredible morning with the @kwisanga_family of gorillas. When we first arrived the clouds were darkening and the vegetation was t... See more
HOW WAS TODAY'S NEWSLETTER?
REACH 40K+ FOUNDERS, INVESTORS & OPERATORS
If you’re interested in advertising with us, send an email over to [email protected] with the subject “Homescreen Ads”.
FINAL JEOPARDY! ANSWER