Some Negative Takes on AI and Crypto

Hollis Robbins on LLM tells; The WSJ on Trump-crypto-China

Hollis Robbins writes,

it produces plausible-sounding text about the concept of red chairs without ever having seen one.

…So here’s a handy rule: if you can’t see anything, if nothing springs to mind, it’s probably AI.

Robbins points out that AI writing can be detected even by the models themselves! I would say that if the models were really thinking, then they could write differently. But the writing consists of word pattern-matching.

You see this in coding, as well. When I work with Claude, and there is a bug, it will suggest an approach for debugging that is clearly based on pattern-matching, not on logical thinking. I can see, based on logic, that the bug could not possibly be where Claude wants to start looking.

Robbins writes,

Another rule: look for formulations like “it’s not just X, but also Y” or “rather than A, we should focus on B.” This structure is a form of computational hedging. Because an LLM only knows the relationships between words, not between words and the world, it wants to avoid falsifiable claims. (I’m saying want here as a joke but it helps to see LLMs as wafflers.) By being all balance-y it can sound comprehensive without committing to anything.

I’m guessing that this is the result of reinforcement learning, where the humans train the LLM to offend as few people as possible.

In the WSJ, Angus Berwick and Patricia Kowsmann write,

The online trading platform, PancakeSwap, serves as an incubator of sorts, drumming up interest among traders to use coins issued by the Trump family’s main crypto company, World Liberty Financial.

The more World Liberty’s flagship coin, USD1, is used, the greater demand to increase its circulation, and the greater the profit for World Liberty and its owners, including the Trump family.

…employees inside Binance created PancakeSwap…

Binance’s majority owner and founder, Changpeng Zhao, spent four months in jail in the U.S. last year after Binance agreed to pay a $4.3 billion fine for becoming a global money-laundering hub for criminals, terrorists and sanctions evaders….

Zhao—considered the richest person in the crypto industry and worth over $70 billion, according to Forbes—recently hired Ches McDowell, a lobbyist and friend of Donald Trump Jr.

Suppose that you and I were a two-person crypto enterprise. How can we get rich? We can sell a coin back and forth for ever higher prices, but that only goes so far. At some point, we have to export coins or other services to the rest of the world.

So what are the crypto sector’s exports? Cross-border remittance services are the only legitimate one that I can think of. My hypothesis is that the fortunes being made in crypto come from criminal enterprises and Ponzi schemes.

If you get involved with crypto, you are entering the underworld. Do not expect the people you find there to be decent and honest.¹

substacks referenced above: @

How to Tell if Something is AI-Written

1. Just after I finished writing, I saw this Xtweet, courtesy of Tyler Cowen.