Welcome back to the AI For Humans newsletter!

Have you heard that AI might be a bubble? And that this particular bubble, like many bubbles before it, might bring the entire economy down with it?

I have. Again. And again. And again. 

Honestly, if you haven’t heard this, I commend you because it means you’re enjoying a mostly off-line life, unencumbered by algorithms.

For the rest of us, I thought it might be worth parsing the arguments for a bubble to start with and give you some of my reading & I’ll follow up with a second newsletter this week that gets into the arguments against being in an AI bubble.

The goal here is to get you some clarification as to what both sides are saying & why but more importantly for you, the human, to understand what’s going on.

Let’s First Define What Bubble Means Here…

Before we even start to talk about whether AI is or is not in a bubble, it’s really important to figure out what we’re talking about.

Some people think the AI bubble is about AI capabilities, whether or not AI will continue to improve. Essentially, those who believe we’re not going to get to artificial general intelligence (at least not soon) and those who do.

I’m comfortable saying that yes, AI will get better as more resources and talent pour into it. We’re seeing some plateauing of capabilities from the SOTA models. Overall though? AI IS getting better.

But the AI bubble that everyone’s talking about right now? It’s about money.

The ‘Hell Yes, We’re In A Bubble’ Side…

The dominant position here is that, yes, we are in a bubble. I mean, good lord, even Starbucks is pivoting to AI.

Here are the basic arguments for WHY we’re in a bubble…

1. Enormous Spending & Investment Due to “Unprecedented Demand”

The amount of money being invested in AI is a whole heck of a lot. Interestingly, it’s not just the valuations of the companies themselves (which, let’s be clear, are sky high) but more so the investments in AI infrastructure like data centers & energy companies.

The basic pro-bubble argument here is that like the first internet boom, we’re building based on what is a perceived level of future demand (see the ‘Dark Fiber’ argument). In the early days of that boom, the all-in POV was that the biggest winners were going to be the companies that provide the pipes to power all that new fangled stuff.

But, people warn, we’re repeating history by building out sizable infrastructure for AGI & everything that’s to come even if it’s not fully working economically just yet.

If you’re not familiar with how that ended, google ‘pets dot com’.

tl;dr Too much money is being invested into things which don’t make money… yet.

2. The Circular & Semi-Shady Way This Money Is Getting Invested

Have you seen this chart floating around the internet?

A LOT of the money that’s flowing around in AI right now is coming from the same slush fund. NVIDIA, the world’s largest company by market cap, is providing a lot of it because, well, nearly all of the AI companies rely on its chips to function. Oracle, meanwhile, gets to host a lot of these models & has a ton of interest in seeing all these companies grow in capacity.

On top of all this are the new “Special Purpose Vehicles”, or SPVs, that are being created to fund some of this infrastructure, which, to be honest are confusing as hell to me too. OpenAI and Anthropic have specifically gone out of their way to say don’t assume that any of these SPVs has direct access to investment in them.

Circular investments? Extreme overvaluations? Where have we heard that before?

I don’t think this is the Housing Bubble situation but, still, it all feels shady and weird.

tl;dr These companies are basically trading money back and forth and inventing new ways to get more money. Kind of classic bubble behavior.

3. Sora 2 & Other Recent AI Products Show Us That The Promise of AI Is Just More Crap For Us To Consume

This is the most cynical of the bubble arguments but it’s starting to gain quite a bit of traction so it’s important to understand it.

With the wave of MLK Sora 2 videos hitting social media (and Dr. King’s daughter clearly being unhappy & working with OAI to remove him) and Sam Altman’s recent tweet saying ChatGPT will soon allow users to generate erotica, there’s been a growing sense that perhaps there is no there there when it comes to the amazing stuff we were promised with AI.

This argument is most clearly illustrated in the Adam Conover video I linked above that I believe everyone should watch, even if it’s a bit much. I really loved Adam’s show “Adam Ruins Everything” but this version of him is a much more caustic which, to be clear, I know a lot of his audience wants.

More so, some of these people argue that there never was a there there and that all of this has been another way (like, say, Web3) to scam the American public again into buying the empty promises of Silicon Valley as they laugh all the way to the bank.

In addition, these people point directly to the story from MIT that says that 95% of businesses use cases for GenAI are failing (even if there’s a whole second side to that argument).

In my opinion, stories like Google’s recent advancement in AI-dervived cancer therapies show that real things ARE happening but Sora videos travel.

tl;dr The bubble is caused by false prophets and it’s all just a hype machine for $$$

Further Reading / Listening

Here are three quick links to better understand the “Hell yes, it’s a bubble!” argument that I’ve personally consumed over the past few weeks…

• Derek Thompson + Paul Kredosky chatted on Derek’s podcast 'Plain English’ and laid out an understandable version of how all this could blow up.

• Hank Green (an incredible explainer of complicated stuff) breaks down Bloomberg’s circular AI investment charts.

• I really like the Baron’s Streewise podcast, it’s business focused but they tell stories from a human-first lens and this is a great AI bubble breakdown.

That’s it for today. See you on Wednesday for the other side!

- Gavin

In this week’s AI For Humans: ChatGPT To Allow… More Freedom. Plus, Sora 2 Update👇

We 💛 This: Sora 2 Fake Ad Prompts

If you’re anything like me, you’re a sucker for a fake commerical. I’m not sure if it’s my background in comedy (SNL ad parodies are just part of my DNA) but I love making a good fake ad.

And now, thanks to Sora 2, the dream of the single shot fake ad is here. I saw this prompt form one of our favorite AI creators FoFR and tried it myself.

As with all things AI, it’s not perfect but it does point to a future where prompt-to-completed-thing feels possible & I’m not entirely sure how I feel about that.