😺Google + Meta + Apple New AI

PLUS: You won't believe THIS industry is losing jobs to AI...

Welcome, humans.

There are years when nothing happens, and weeks when years happen. This is definitely one of those weeks.

See, there's an unwritten rule in the AI industry: when one company announces a major AI update, everyone else follows. Wednesday proved this, with Google and Apple launching massive AI upgrades within hours of each other (more on that below).

And if that wasn't enough, Meta launched Llama 3.3 just days ago, while OpenAI packed so many releases into five days that it actually crashed their own system, sending their status page viral. Now that ChatGPT has 300M weekly users, any downtime is a big deal. Especially for students cramming for finals…

We'll see what Day 6 brings soon—check here at 10am PST. In a week where every day feels like a year in AI progress, let's hope whatever gets launched doesn't sink OpenAI’s Shipmas...

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

Google announced Gemini 2.0 and Apple launched iOS 18.2 with ChatGPT.
SORA recreated near-replica game footage.
VCs flocked to a space AI datacenter startup.
Harvard shared 1M book dataset for AI training

Advertise in The Neuron here.

Google, Apple, and Meta Just Released Their Biggest AI Updates of 2024

We already gave ya the preamble in the blurb, so let’s get right to it.

First up, Google introduced Gemini 2.0, a series of experimental new projects:

Introducing Gemini 2.0 | Our most capable AI model yet

The TL;DR of this launch is that Google's going all-in on AI agents that can actually do things, not just chat.

Example: the experimental Flash 2.0 version available today, and it comes with some serious multimodal upgrades: native image generation, voice output, and built-in tool use for things like Google Search and coding.

Apparently, Flash is shockingly better and cheaper than Claude 3.5 for coding AND it has a native 128k context window. Here’s a roundup of how ppl are already using it.

What else Google announced:

Deep Research: New feature that can explore topics and compile reports.
Project Astra: AI agent that can use Google Search, Lens, and Maps, understands multiple languages, and has 10-minute conversation memory.
Project Mariner: A browser control agent that can complete web tasks (83.5% success rate on WebVoyager benchmark). Read more about Mariner here.
Agents to help you play (or test) video games.

Apple Intelligence also launched in full yesterday with iOS 18.2, bringing AI features to iPhone, iPad, and Mac. If you have an iPhone 15 Pro/16 or newer, iPad with A17 Pro/M1+, or Mac with M1+, you can use it. If you’re in the EU, rollout’s in April 2025.

Here’s what we get:

Image Playground: Built-in DALL-E style image generator (works in Messages, Notes, everywhere).
Genmoji: Custom AI emoji generation from text descriptions or photos.
Image Wand: Turns sketches into polished art, generates images from note context.
Visual Intelligence: Point camera at anything for instant analysis (for iPhone 16 only).
Writing Tools: New “Describe Your Change” feature for text editing.

The big news of this launch is the ChatGPT integration. Here’s the official announcement video from ChatGPT where they demonstrates how this all works:

Say “Ask ChatGPT” or double-tap Command key.
No account needed + complete privacy (hidden IPs, no data storage).
Camera integration for visual questions.
Full document/image understanding.
One-tap continue in ChatGPT app.

OpenAI and Apple demo’d the integration to:

Analyze a 49-page system card doc instantly.
Generate a pie chart from technical analysis.
Use the camera for live object analysis.
And how to continue every interaction in the ChatGPT app.

Last thing: do NOT sleep on Meta Llama 3.3 70B.

Meta's Llama 3.3 70B just launched too, and it's basically GPT-4 that runs on your laptop. Currently #19 on LiveBench, beating Claude 3 Opus.

Simon Willison got it working on his 64GB MacBook Pro M2, though he advises against using less RAM.

The model takes up about 42GB of storage space, but the results are impressive—it can write complete web apps, generate SVG graphics, and handle complex reasoning tasks right on your machine

Here's how to run it:

Download LM Studio from LM Studio.
Follow the onboarding, then go to Discover, and search for “Llama 3.3 70B” in models.
Click download (heads up, it’s 42GB).
Hit "Chat” to start using.

If you’ve never used Llama before, this intro + prompt template will probably be helpful.

Whew! That was indeed a lot. What a week for AI… can you believe it’s not even over?!

FROM OUR PARTNERS

How to solve the three biggest roadblocks to scaling your business.

Here are three problems we bet you didn’t know you have:

The “We Can't Scale Globally Because Language” problem.
The "Our Marketing Team Needs Better Customer Stories" problem.
The "Our Reps Don't Know Everything (Yet)” problem.

Attention eliminates these three major sales roadblocks by breaking through language barriers, turning calls into case studies, and making every rep your best rep yet.

Bonus: It does all this while saving your team hours of manual work. (Just ask Gladia—they eliminated 90% of their low-value tasks). See how → here.

Treats To Try.

*Leaders from Microsoft, Reddit and beyond dive into AI's future with Reid Hoffman on Masters of Scale, supported by premier brand partner Capital One Business. Listen Now.
AISmartCube builds custom tools w/ a drag and drop interface that’s like Lego blocks—connecting chatbots, voice tools, and data scrapers.
Shortcut turns your voice into polished written content at 3x speed - just speak and watch it create emails and messages that match your tone.
Remention monitors online discussions and suggests natural places to mention your product across social platforms without spamming.
Bricks turns your spreadsheet tasks into simple conversations - just tell it what you want to do and it handles the formulas, charts, and data analysis.
Speaking of agents…SmythOS helps you create agents by describing what you want, then it automatically builds them using the best tools and services.
Tavus lets you video chat with an AI Santa anytime.

See our top 51 AI Tools for Business here!

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

Around the Horn.

At least someone’s able to use it!

SORA can apparently recreate gameplay footage pretty faithfully (with promtps that workaround filters, like “italian plumber game”, indicating it was definitely trained on game footage (probably because there’s a lot of that on Youtube).
200 VCs wanted in on the seed round a startup called Lumen Orbit, who plans to build data centers in space to power AI (The math? $140M in electricity on Earth vs $10M to launch a satellite).
Harvard announced it will release a dataset of 1M public domain books for anyone to train their language models on—that’s 5x the size of the book dataset used to train Llama.
Humans getting replaced by AI in action: OnlyFans models are now using AI to chat with the humans in their DMs—a job that used to be done by other humans called “Chatters” the models would previously hire to chat for them.

FROM OUR PARTNERS

💼 Want to become an AI Consultant?

The AI consulting market is projected to explode from $6.9B to $54.7B by 2032. The AI Consultancy Project has helped 300+ students build successful consulting businesses, with a 6-month roadmap to building a 6-figure business and methods to land clients in 3 days.

Take advantage of early access—request your spot ASAP!

Under the Hood

Check out Gemini’s Multimodal Live API: real-time audio/video streaming with tool integration (official docs here).
Rememberall helps you maintain continuous conversations with your GPTs by storing and recalling context from past chats when you mention @rememberall.

Thursday Trivia

One is a real, and one is AI. Which is which? (vote below!)

Which is AI?

The answer is below, but place your vote to see how your guess compares to everyone else (no cheating now!)

Here are the results from last week’s trivia (A was AI):

That’s 4 - human, 2 - robots. Easy win for the humans this time.

Here’s what you said:

Q.S: “Just looks too clean, too good.”
C.H: “the girl w a pearl earring only has one pearl - the three pearls is a computer mistake.
S.H: “B looks more like an earnest attempt at matching the original painting.”