Building the world's biggest computer🌎

Your web browser gets smarter, the race to build the biggest computer, and how to explain anything with AI

Jan 26, 2024

Quick Definitions (skip if you already know them!!)

LLM – Large language model, this is the general term for AI models like ChatGPT, which in practice, are incredibly smart and capable chatbots. For a more in depth definition, go here.

1 cool thing

Chrome gets some AI upgrades

Abstract illustration with photos showing the northern lights and windows showing the “Help me write” and Tab Organizer features. — Image from Google

Google Chrome is getting two sweet new AI power-ups (and one that’s useless).

First: AI-enabled writing in your browser

If you are a good ape, and you have at least tried ChatGPT, Copilot, or any of the other big LLMs, then you know AI is great at writing just about anything super fast (great as in it’ll do it fairly decent in about 2 seconds, not great as in this is A+ level writing).

Well this Chrome update is SUPER COOL because it integrates that awesome AI writing ability directly into your browser.

What does this mean?

It means that all you will have to do is right-click on a search bar in Amazon or a Facebook comment or whatever and Google’s AI will give you custom suggestions based off of the text / content on the page. See below!

Next up: Smart Tab Organizing -> How to have 100+ tabs open & it NOT be a problem

I use a lot of tabs, so I’m extra excited about this one. This new feature enables smart grouping of tabs by AI (or manually if the AI messes up). In the above example, it automatically grouped together a couple of hotel websites and Expedia as a ‘travel’ group (with a nice plane emoji too, love that), a few weightlifting tabs as fitness, and then my apes tabs.

You can quickly condense the tab groups as well! I probably have on average 50-70 tabs open at work when I’m really getting into some projects so this is going to be huge!

You can try out these new features in Chrome RIGHT NOW on Apple and Windows computers. Once you are signed into Chrome it’s super easy:

Select “Settings” from the three-dot menu in the top right corner
Navigate to the “Experimental AI” page / tab on the left side
Toggle the bar to enable!

Can’t find the Experimental AI section under settings?

You might have to update Chrome (I did)

Once you’ve got it working play away, and never get roasted for having 25 tabs open again!

The big deal

The biggest computer the world’s ever seen.

Server farm image made with AI. For a look at a real supercomputer, check out Tesla’s Brain.

Organizations, governments, and companies across across the world are all racing to build the world’s largest computer, at a pace completely unprecedented.

Why?

Compute.

Compute?

To get more compute. Compute (the AI industry’s latest new lingo) is the amount of computing power that a company or individual has access to. It’s the horsepower of the AI / computing world. Except we’re not talking about comparing a 150 horsepower Prius to a 500 HP Corvette. We’re talking nuclear-powered gigaton aircraft carrier levels of computing power.

But why? Why do companies need more compute?

Simple.

AI.

The smartest AI model out there is still GPT-4, created by OpenAI almost one and a half years ago now. In order to turn GPT-4 into the smart AI that it is, OpenAI had to teach it a whole lot. To teach one of these big generative AIs / LLMs, you need a whole truckload of guess what?

Compute.

In fact, you need so much compute that only a company like Microsoft or Google or Facebook can even afford to actually do the training, because we’re not just talking about internet usage or your electricity bill here.

To train GPT-4, and to train future smarter AIs, you need multiple buildings full of the most advanced computer chips in the world (called H100s, and they cost 40k a pop).

[Insert graphic of building full of computers / server farm]

This brings us back to why a number of separate groups are all racing to build the biggest computer farms in the world.

build the biggest computer -> have the most compute -> train the smartest AI model -> rule the world

That’s about all there is to it!

This is why OpenAI needs Microsoft’s money. To buy computer chips to train their AI models. And it’s why Sam Altman, CEO of OpenAI is working to start a new billion dollar chip venture to rapidly ramp up the production of the chips best suited to train these AI models. (Currently Sam, and most other companies, are VERY reliant on NVIDIA for these chips, and yes that is why NVIDIA’s market value as a company has increased by nearly a TRILLION dollars in the last year) And one company can only make so many chips. So Sam wants to get some Middle East Mula💰 and go make his own chips too.

Guys, if I were to write the script for an AI movie, I mean this would literally be it. We’re in the lovey-dovey phase of ChatGPT is making everyone’s life easy and great, meanwhile Sam Altman is the totally-entranced prophet of AI building a massive AI supercomputer on some island that will turn around and take over the world. Just a thought.

Make life EASY

Explaining complex concepts simply

One of the things I have recently become most passionate about is communicating in a way people understand. It’s often not easy to do, but it’s really important to getting your message heard and just teaching people / communicating effectively in general, so it’s a good thing we’ve got AI to help us!

3 Quick Techniques to get AI to explain anything well:

Define the age range
1. “Explain physics to a 5 yr old”
Make it specific / applicable
1. “Explain the stock market to a 10 yr old who is really into Monopoly / board games”
2. “Explain why brushing your teeth is important with a baseball analogy”
Define the length / organization
1. “Give me 5 bullet points a college student should know before getting a credit card”

And finally, iterate! The 5 yr old explanation was too dumb? Say, “hey chat, that was dumb, explain it to an 8yr old”. Continue to iterate until you get what you’re looking for.

You can try this out right now in Copilot! (Microsoft’s AI, powered by GPT-4)

That’s it for today folks. Hopefully you learned a little, laughed a little, and live today like there’s no tomorrow!

Smell ya later,

Joe

Enjoyed today’s breakdown? Share it with a friend who might benefit from some free AI labor!

Thanks for reading AI for Apes! Subscribe for free to receive new posts every week.

P.S. I had trouble finding an inside look at one of the latest and greatest Microsoft / Google data centers, but this video of a Bitcoin Mining facility in Texas should give you the idea of what a building full of computers actually looks like! Spoiler: it’s insane!

AI for Apes