😸 No AI returns before 2026?

PLUS: New research on why AI can't train on AI...

Welcome, humans.

ANNOUNCEMENT: Imagine your team achieving twice as much in half the time. We've seen this impact firsthand on our own business using tools like ChatGPT and Claude.

Now, we're helping you achieve the same. We’re very excited to announce that we’re partnering with Exec to launch a ChatGPT Bootcamp to help teams get the most out of ChatGPT.

The program includes:

Introductory workshops to make your entire team proficient with ChatGPT.
A "hackathon" where participants build custom GPT models for their tasks.
Ongoing support with weekly tutorials, prompts, and assignments.

This is the #1 opportunity to accelerate your company into the AI era and gain a leg up on your competitors. Plus, it's taught by experts from Airtable, LinkedIn, and Google who really know their stuff.

$10,000 for up to 50 team members. Limited spots available. Book a consultation here.

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

Google's AI spending spree shows that tech's costly gamble on future profits.
New research showed that AI models stagnate when trained on more AI-generated data.
The New York Times is using an AI tool to match advertisers with stories.
Bing launched its own AI search tool to rival Google’s.

The biggest names in AI are losing money with not a lot to show for it.

Source: FT

Ever heard of the prisoner's dilemma? It's that tricky situation where two people have to decide whether to help or betray each other without knowing what the other will do. You know, like you’re both in a holding cell, and the cops want one of you to rat on the other one?

That’s Big Tech CEOs and AI spending right now. Let us explain:

Tech giants and VCs are pouring unprecedented amounts of $$$ into AI.

Alphabet's Q2 earnings report yesterday revealed a 91% surge in capital expenditures to $13.2B, with $2.2B spent on AI development alone.

Despite these substantial investments, the returns weren’t anywhere close to a positive ROI.

One analyst even said we’re unlikely to see “material” AI revenue until 2025 or even 2026.

An analysis of OpenAI's finances paints a similar picture:

Operating costs could reach $8.5B.
Annual revenues estimated at $3.5-$4.5B.
Projected losses up to $5B this year.

Anthropic's situation is potentially even more dire, reportedly making 1/5 to 1/10 of OpenAI's revenue while burning $2.5B in compute costs alone.

Is this sustainable?

Zuckerberg and Alphabet CEO Sundar Pichai say it doesn’t matter—it’s necessary. Both argued that while their companies might be spending too much today, they're making an informed decision for tomorrow.

And this brings us to the prisoner’s dilemma: do nothing now = risk falling behind the other guy when the payday DOES arrive.

Why does this matter? AI hype is propping up the tech sector, and by extension, the entire friggin economy. If returns don't materialize soon, we could see a significant market correction. Case in point: Nvidia had a bad three-day stock plunge, and it wiped out more value than 96% of the S&P 500 (y’know, the top 500 biggest companies on the U.S. Stock Market).

OH, and goes without saying, but this ain’t investment advice :)

FROM OUR PARTNERS

Build your next big AI product idea.

At AE Studio, they’ve done some crazy things. One time, they taught AI to brew beer, and then market it — it sold out.

They want to help you build crazy things, too. But first, you need a crazy idea. Their AI Ideas Generator takes your business problem and gives you a report of AI-powered solutions that can help you address it.

Give it a try for free.

Around the Horn.

ChatGPT-4o mini has entered the LLM Arena competition, tying for first place with ChatGPT-4o while being 20x cheaper.

Listen to Noah talk about The Neuron and future of AI on the awesome AI for Humans podcast (interview starts around 45-minute mark)!
A new study found that AI models plateau in ability when trained on AI-generated data.
The New York Times used AI (BrandMatch) to better match advertisers with relevant stories—and it’s performing better than non-AI techniques.
Bing released its rival to Google’s AI-powered search called Bing generative search; it’s only available to a small % of users to start.
Colin Kaepernick launched a new AI startup to give creators tools for skills they don’t have—e.g., comic writers can create graphics, illustrators can create text, etc.

Treats To Try.

*Struggling to Attract Niche B2B Leads? Tely autonomously creates expert B2B content for your niche and captures leads—all on its own. Get traffic in 3 weeks.
Mistral dropped a new open-source model (Mistral Large 2) that’s competitive with Llama 3.1 and therefore ChatGPT-4 (Try it on Mistral or HuggingFace).
Noded AI automatically organizes your notes, creating insights and shareable documents (raised $4M).
Stability AI released Stable Video 4D, an AI model that turns single object videos into multiple videos with eight different angles/views (HuggingFace link).
Reworkd, who built AgentGPT (free AI agent creator), has a new tool to build AI agents that can extract structure data from the public web. MrScraper is another option.
Mermaid AI from Mermaid Chart uses prompts to generate diagrams to create professional flowcharts.
Hemingway Editor Plus, the writing app that shows you how to write like Ernest Hemingway, now has AI writing features that help you write faster w/ more clarity.
FilmAI.tech offers a catalog of 6,000 Hollywood films for AI companies to scan, aiming to help businesses avoid copyright infringement lawsuits with licensed content.