🍎Apple’s making their own AI chip

ALSO: AI is scanning medical images + AI Army Guide Part 3

Estimated Read Time: 5 minutes

A year ago Nvidia was worth $1 trillion. A lot of money. Now it’s worth $2 trillion thanks to the AI boom. Except apple doesn’t like that. They don’t want to rely on Nvidia, so they’re creating their own chip. Oh, and we have the next part of our guide to creating your personal AI army (for peaceful purposes of course).

• 🍎Apple’s making their own silicon AI chips

• 📲AI on your phone is the future

• 🧐New AI finds uncertainty in medical images.

• 🤖Create your AI army part 3!

🍎Apple’s making their own silicon AI chips

What happened: Apple is developing its own silicon AI server processor. The chip is designed for large data centers. This is a strategy for weening off of Nvidia’s chips. Apple has a history of controlling their production lines, so this comes as no surprise.

The details: 

• Apple is designing this processor specifically for AI workloads, aiming to optimize performance and efficiency in its data centers.

• The move is meant to make Apple self sufficient rather than relying on Nvidia and Intel.

• This will allow Apple to manage costs more efficiently.

• Apple claims their proprietary chips will enhance user privacy.

Why it matters: This could be a pretty big deal if Apple makes good on it. Nvidia is a huge player in AI specifically because of their chips. Apple is cutting out the middleman. If others follow suit this could lead to serious trouble for Nvidia and other chip makers.

As of April 25 market close.

💰 Nvidia is acquiring Run:ai for $700M. Run:ai is a company based in Tel Aviv that specializes in managing AI hardware infrastructure. Talk about expensive 🤑. Apparently initial discussions suggested $1B. Run:ai had previously raised $118 million and had attracted a significant customer base including Fortune 500 companies. This move marks one of Nvidia’s significant acquisitions following the $6.9 billion Mellanox purchase in 2019.

🐇 The new AI device, Rabbit R1, is out. And people actually like this one. While it’s not perfect, reviewers say the R1 is fun and quirky little device for $200. It’s cited as being between a fidget spinner and an MP3 player. Add salt to Humane AI’s wound, will you?

📲AI on your phone is the future

What happened: From Microsoft releasing Phi 3, to Apple’s acquisitions and AMD’s APU chips. Everyone gearing up for AI to become available locally. What does this mean for the AI industry and you as an individual?

The details:

• Recent releases and acquisitions point us to a clear conclusion: AI will be made available locally on PCs and phones.

• This will not get rid of cloud computing, rather they will work together.

• Smaller tasks and assistant features will likely remain on local devices, while heavy processing takes place in the cloud.

• A large reason for this is the desire to reduce energy costs for AI infrastructure. Which can rack into the billions.

• IOS 18 is expected to release with local AI features this year. It will be a turning point for public AI use.

🧐New AI finds uncertainty in medical images

What happened: MIT researchers have introduced Tyche, an AI tool that generates multiple plausible segmentations of medical images. What does that do? It addresses the inherent ambiguity in tasks like identifying disease markers.

The details: 

• Basically it’s a second eye looking for diseases in images that doctors may have missed.

• Tyche provides multiple different points of view by itself, so it acts as a collective rather than an individual. This helps it be objective.

• Tyche can adapt to new tasks with just a few example images and without the need for extensive retraining.

• Tyche operates efficiently on standard computing systems, so it doesn’t need a massive LLM server.

Why this matters: Medicine is the area I’m most excited for AI. A cure for cancer and destroying diseases. That’s what I want. AI gives us the best chance for that. But right now it’s all about optimization and checking things like images. If this helps even one person prevent an untimely death, then it’d be worth it.

🤖Create an AI army part 3

Part 3 of our guide:

Today we’re going to build a copywriting GPT. 

To do this we’ll need to train our GPT on copywriting books. 

We can also apply a style/tone of voice to the copywriting. 

Let’s get started with the following prompt: 

Create a copywriting GPT which will create sales copy for my business. 

Use the uploaded knowledge to create the GPT. 

Make note of the style, format, and tone of voice of the copywriting on display in the uploaded files. 

Use them to dictate the style and voice of the copywriting GPT. 

This is information about my target audience: (enter info)

This is information about my product: (enter info)

Use this background information and the uploaded books to create the copywriting GPT. 

Make sure you upload files that are more than just books. 

For example, you can upload multiple files to the GPT. I recommend uploading some classic copywriting books, such as the Adweek Copywriting Handbook. Use these books as a reference for learning copywriting. These are lessons on how to write sales copy. 

Then provide examples of good copywriting. Examples in this case would be a pdf of actual sales copy done in the style/author you want. Just pure copywriting. You can even enter the sales copy directly into the GPT itself. Provide as many examples as possible. A whole book’s worth if possible. 

This is important. Most people just upload a few basic files and call it a day. 

But by providing learning material and examples together, you can create a GPT that is much more capable. 

(Next week we’ll learn about applying a specific tone of voice to a GPT by using another GPT to grade the tone of voice. If that sounds confusing… it’s not. You’re just using two GPTs together.)

Obviously this technique is applicable to any skill you want the GPT to learn, not just copywriting. You can apply this to just about anything. 

Need business advice? Upload your favorite business books. Just use the techniques here as a guide. 

We are using the knowledge from the book to train the GPT just as you would train yourself or an employee. 

This is also why you need to consider what books you upload. In the case of copywriting, some authors advocate for specific styles of copywriting. 

For example, I personally love the format of Great Leads. So I upload this to the GPT as part of the learning material, and I use examples that work with that system/process. Don’t use examples that contradict the very philosophy/process the training material covered. 

Play around with the GPT a little. Alter the format or style after previewing the GPT for a bit.

🚊Toy trains galore

Source: Midjourney

Give us your feedback!

Got feedback? Rate today’s newsletter by clicking below! 

Got more personal feedback? Fill in the box when you rate us and give us your criticism and feedback.

Thank you for reading! 

❤️Share the (ARTIFICIAL) Love!

Got a friend who wants to appear as smart as you? An AI fanatic or someone shaking in their boots of getting replaced? Share us by clicking the link below and you’ll get our love ;)