Welcome to another edition of the best damn newsletter in AI.

As a reminder, we’ve opened up Reader Questions. Our mission is to make using AI accessible to every business, and a huge part of that is making sure you’re not stuck and that we continue answering diverse questions. So, submit your question here. We’ll review every single one and cover answers here in the newsletter.

Here's what we're covering today:

• Our NEW!! course with LinkedIn Learning (access details below!)

• A breakdown of why AI will keep getting better and costs getting lower

• What’s happening in AI from the big guys to the startups and law suits… it’s spicy down there

Let’s get to it! 👇

ANNOUNCEMENT

New LinkedIn Learning Course: Using AI to Boost Productivity in Any Role 🤘 

Our team just launched a new course in partnership with LinkedIn learning! In it, we cover:

• How to get the most out of AI

• Learning to speak the language of LLMs (our best prompting tricks!)

• Finding the best AI for the job

• How to create once, use multiple times

• How to supercharge your research with AI

• Using AI as an on-the-go thought partner

• Improving your writing & presentation skills with AI

• Applying any framework with AI

• Organizing your knowledge

• & Rapid fire tips

Access the course here, and hit reply to let us know what you think!

TODAY’S PERSPECTIVE

How AI will continue getting better and cheaper + how to plan for it

If you’ve been mostly using ChatGPT Plus or one of the subscription apps, you might not have noticed a major trend happening over the last 18 months.

While the quality of AI is raising, the cost is dropping.

Both major AI players - OpenAI and Anthropic - have cut prices by an order of magnitude, multiple times over the past 18 months. Meanwhile, the quality of their AI models is getting better and better.

How is this possible?

The costs that go into building AI models comes from two factors:

• Fixed costs: The upfront cost to develop and train the model

• Variable costs: the cost to process your request to the AI model each time you send it a message

With these large language models, the fixed costs of training are astronomically high… usually 7 figures+ and there are rumors that training ChatGPT-5 is coming close to 9 figures+.

Versus the variable costs are really quite small, estimated to be around or slightly under the costs that many of these AI model providers are charging customers for.

Will the cost continue to go down?

Or is this the Uber strategy (dominate the market, then raise prices)?

Experts debate both sides.

Obviously the 9 figures+ of cash required to train one of these massive models has to come from somewhere, but based on the revenue numbers OpenAI continues to report — it appears they’re recouping their investments.

Today’s costs also account for very little cost optimization done by the smart AI teams at each of these companies. From our experience, there’s a lot of performance gains and cost savings to be made with these types of systems — it just takes time for teams to experiment and unlock them.

Our advice: Plan for the “cost” of AI going towards zero

Don’t limit yourself or be afraid of the costs even as they exist today.

LINKS

For your reading list 📚

It seems this week’s theme is controversy...

• AI music generation tools are facing lawsuits, and Perplexity is in the same boat. The problem of where data is coming from remains a hot topic in AI.

• Toys R Us debuted the first AI generated commercial using OpenAI’s Sora, and critics are not happy

Meanwhile the big guys continue pushing forward…

• Apple is in talks to join OpenAI’s board as an observer

• Microsoft published more of their results on using graph architecture to help with adding context into LLMs for answering questions (it’s nerdy, but very interesting!)

And startups are continuing to push the envelope…

• Perplexity unveiled a new version of Pro Search

• Startup product leaders shared counterintuitive advice for building with AI

• And more startup thought leaders weigh in on the different levels of AI agents

Have a burning AI question? Or stuck trying to help your team adopt AI?

We read every single question submitted and use these to inform the topics we go deep on. 👇️ Submit your question below!

That's all!

We'll see you again soon! Thoughts, feedback and questions are much appreciated - respond here or shoot us a note at [email protected].

Cheers,

🪄 The AI Exchange Team