_{Brought to you by:}

Crypto may need a productive crash

Is AI winter coming?

With the introduction of ChatGPT 18 months ago, a robot takeover seemed an imminent threat to humanity — remember when a prominent expert proposed bombing datacenters training large language models? 

That already feels like a different era, because despite massive investments in such datacenters (and an irresponsible lack of airstrikes), large language models (LLMs) haven't gotten much smarter since.

Instead, a consensus seems to be forming that, far from being a threat to humanity, LLMs won’t progress much beyond mimicking what humans post on the internet (see here, here, and here, for example).

This checks out with my personal experience.

Eighteen months ago, I thought LLMs would soon be writing this newsletter (I was hoping for a sweet spot of paid time off between the LLMs starting to do all the work and my employer realizing it). 

Now, though, I find myself using the chatbots less, not more — they are pretty good for brainstorming ideas, yes, but all of the hallucinated answers can turn them into major time-wasters. 

Anecdotally, it seems like a lot of people are having similar experiences. 

Aside from GitHub’s co-pilot, are there any mainstream use cases for LLMs?

There doesn’t seem to be — and that could turn into a problem for investors, as the handful of stocks that have carried the S&P 500 to new highs may not get much of a return for the hundreds of billions they’ve been investing in datacenters and GPUs.

If history is any guide, though, a stock market bust may be exactly what we should be hoping for.

LLMs don’t seem to be the next big thing, but with luck, the massive capex spend to build LLMs will be what enables the next big thing.

Chances are there will be an AI-related crash in between and (by making the infrastructure ultra cheap) it’s often the crash that does the enabling.

This is the Carlota Perez theory of productive bubbles — which suggests crypto should be hoping for a crash, too.

Not all heroes wear capes

The forgotten heroes of the internet revolution are the investors who made the irrational decision to pile into telecom stocks like Global Crossing during the dotcom bubble.

Global Crossing raised $399 million in its 1998 IPO and billions more in debt in the four short years before it filed for bankruptcy in 2002 (history’s fourth largest bankruptcy at the time).

It only seems irrational in hindsight.

Shortly after its IPO, Barron’s opined that Global Crossing “appears to be a sane way to play the internet” — and that was very nearly correct.

Global Crossing turned out to be a lousy way to “play” the internet, but a great way to build it.

The ambitious company used investors’ money to build a sprawling fiber-optic network connecting four continents by submarine cable. 

Underwater cable guys like the one pictured at the top do not come cheap, so it’s unlikely that network would ever have been funded by investors acting rationally.

Much of that infrastructure continues to be used today — in 2024, you may well be reading this newsletter courtesy of a fiber-optic cable installed by a telecom company that paid for it by selling stock to the writer of this newsletter in 1999.

This, of course, is why the much-derided dotcom bubble was a good thing for the world.

“Asset price inflation was good in the late 90s,” according to Carlota Perez, “when there was a need for overinvestment in telecoms and internet even before they could produce profits or dividends. Everybody was willing to put their money into the new high tech companies because they could get short-term capital gains [...] That is how the capital was gathered to ‘fiber-wire’ the whole world, to get everybody to learn the new paradigm and to test every imaginable dotcom idea.”

So, you’re welcome.

This time around, however, I’ll be selfishly keeping my money in Berkshire Hathaway instead of selflessly funding the AI boom. 

But if everyone did that, nothing important would happen.

Let’s hope they don’t.

LLMs may not be the thing that ultimately changes the world, but the investment craze they’ve inspired will allow companies to “test every imaginable [AI] idea.” 

I’m certain a few of them will pan out.

The problem for investors, as Perez explains, is that there are two phases of every technological revolution: the “installation phase” when the infrastructure is built and the “deployment phase” when the technology goes mainstream — in between is a “turning point” that’s partly enabled by a financial bust.

So, if AI is as important as we think it is, there’s probably a bust coming.

Let’s hope it’s as productive as the dotcom one.

Finding purpose

Dare we hope the same for crypto?

In some ways, Perez’s pattern seems to apply: Memecoins are currently funding a massive overbuild of blockspace.

In other ways, it doesn’t: Crypto has had multiple crashes and none of them have been particularly productive.

But maybe we just haven’t had the crash.

The crypto industry continues to funnel its capital (and developer talent) overwhelmingly into infrastructure and that might have been the right thing to do — it may be that blockspace is only now getting cheap enough to enable mainstream adoption. 

But, even if so, the Perez playbook would suggest that adoption will come too slowly for investors to make a return on all the infrastructure they’ve financed.

We may still need token prices to crash — perhaps to force both investors and developers to shift their efforts towards building some applications that people might use. 

These things are unpredictable, of course. 

Bubbles can last far longer than anyone would guess and when they do bust, most don’t produce anything useful — Perez says the productive ones come along only once every 40 to 60 years.

She also thinks crypto is unlikely to qualify.

To prove Perez wrong (at her own theory), crypto, like AI, will have to find a greater purpose.

Language models are probably not the productive purpose of AI — and we can only hope that memecoins aren’t crypto’s most productive purpose, either.

— Byron Gilliam

Brought to you by:

MANTRA, the world’s leading RWA L1 blockchain, has launched its incubator aimed at fueling the next era of Web3 innovation. The MANTRA Incubator program kicks off at the prestigious Dubai World Trade Centre (DWTC) and offers 5 promising projects a seed investment of $100,000. 

Reasons you should apply: 

• Seed funding: Each project will receive $100,000 in seed investment, personally funded by the CEO & Co-Founder, John Patrick Mullin. 

• Global exposure: Participants experience Dubai, Hong Kong and San Francisco ecosystems, while networking with experts, developers, VCs globally. 

• End-to-end concept to market support: Receive expert support in development, marketing, operations, and more to effectively go-to-market. 

Ready to take the leap? Apply now to transform your ideas into a successful venture.

Harnessing the Untapped Potential of Bitcoin Staking

BabylonChain co-founder David Tse joins us to dive into the transformative impact of Bitcoin staking, the innovative role of Babylon, and the broader technological and cultural shifts in the Bitcoin ecosystem.

Watch or listen to Bell Curve on YouTube, Spotify, or Apple.

Thank you to our sponsor:

The PayPal stablecoin PYUSD is live on Solana. PayPal has over two decades of payments experience, and is now making low-cost, high-throughput web3 payments possible through the PYUSD on Solana launch.

Learn about PYUSD on Solana and start building the future of payments.

PayPal, Inc. is licensed to engage in virtual currency business activity by the New York State Department of Financial Services.