5 min read

Crypto's Bitcoin betrayal

Crypto has betrayed its origins; a new AI breakthrough; something's brewing in China; and an ideology-free plan B for energy.

Please note that due to travel obligations it's unlikely that we'll publish anything for the remainder of this week . See you next Monday!

1—Crypto's Bitcoin betrayal

Sam Bankman-Fried was a crypto billionaire, but he had little in common with the cypherpunks who were engaged in decades of discussion and development that eventually led to the creation of Bitcoin:

"This was an eclectic group of individuals concerned with issues of privacy in the digital age, and the way the digital world requires accounting ledgers to keep an electronic record of transactions.
...
To introduce money with the characteristics that the cypherpunks desired would require something that was resistant to censorship. Creation of a new form of money would also have to deal with the incentives of the issuer, since someone who can issue their own money could potentially manipulate the supply to benefit themselves at the expense of others.

Bitcoin solved both of these problems. Anyone can download the Bitcoin software and operate a node on the network. The decentralized network maintains a digital ledger called a blockchain that keeps track of balances of the cryptocurrency known as bitcoin.

The crypto industry today has moved in a completely different direction, where blockchain is treated "like another thing for the tech industry to tinker with... at the expense of principles like decentralisation and censorship resistance that are central to Bitcoin":

"This past year, the scheming has only gotten worse. The first domino to fall was a project called TerraUSD, launched on the Terra Network. TerraUSD was purportedly designed to be a stable coin, or a token that trades one-for-one with the U.S. dollar. The creators came up with a convoluted scheme to trade TerraUSD with another cryptocurrency to make sure that 1 TerraUSD always had a price equal to $1. As one might imagine,  trading one worthless asset for another is not a sustainable strategy.

Nonetheless, the project became very popular because of a promise that investors could earn a 20% interest rate on their assets. This promise was nothing more than a Ponzi scheme. The current value of TerraUSD is worth mere pennies, which means that anyone who continued to hold it has lost almost all of their money."

You can read plenty more about the origins and direction of crypto from economist Joshua Hendrickson here (~5 minute read), which concludes with "Whatever this crypto industry is, most Bitcoin and Bitcoiners want no part of it."


2—How does it work?

Facebook owner Meta's new AI, Cicero, recently beat a bunch of humans at the game blitz Diplomacy (it ranked in the top 10% amongst a mixed group of pros and amateurs). But is Cicero really "a breakthrough toward building AI that has mastered the skills of negotiation, persuasion, and cooperation with people", as claimed by Meta? That was one of the questions asked by Gary Marcus and Ernest Davis:

"In AI, it's always hard to answer questions about implications without first looking at a system's architecture. As it turns out, the architecture of Cicero differs profoundly from most of what's been talked about in recent years in AI.

The first thing to realise is that Cicero is a very complex system. Its high-level structure is considerably more complex than systems like AlphaZero, which mastered Go and chess, or GPT-3 which focuses purely on sequences of words. Some of that complexity is immediately apparent in the flowchart; whereas a lot of recent models are something like data-in, action out, with some kind of unified system (say a Transformer) in between, Cicero is heavily prestructured, in advance of any learning or training, with a carefully-designed bespoke architecture that is divided into multiple modules and streams, each with their own specialisation."

There is far too much detail for us to cover here but Marcus and Ernest were impressed that Meta's team managed to "combine game-theoretic strategizing with natural language technology that has been developed for purely linguistic tasks", which would have been no easy feat. However, they doubt that Cicero will be able to be adapted to "complex interaction with people outside the closed and limited world of Diplomacy", meaning this kind of machine learning model is probably most useful when "embedded in highly structured systems":

"Even within the world of Diplomacy, the scope is somewhat limited. Human players for example can probably cope well when with an alternative board (the map of Europe as of 1400, say) or with slightly revised rules for actions (e.g. forces that could travel by air rather than just land or sea ). In Cicero, there is no simple way to 'present' any such rule or map changes, and its training is very heavily tied bound to the language describing actions specifics of the standard Diplomacy board; how much would carry over is unclear. Our best guess is that if you played Diplomacy with alternative rules, the system would like to have retrained almost from scratch."

You can read Marcus and Ernest's full essay here (~10 minute read).


3—Something's brewing

"People are chanting 'down with the Chinese Communist Party' and 'step down Xi'. This is happening in cities across the country."

4—An ideology-free plan B

Australia has an energy problem. It's a problem largely of its own making, with decades of bipartisan complacency and poor advice leading to a "critical national infrastructure that (has) deep reliance on non-dispatchable wind and solar":

"While it is true that Australia's Liberal and National parties took too long to accept the mainstream science on climate change, it is now true that the Labor Party is taking far too long to accept mainstream science on next-generation nuclear power.

[Clean-energy expert Adi] Paterson shares the view of many clean-energy advocates all over the world, including billionaire philanthropist Bill Gates, that nuclear energy is an essential component of any clean energy future.
...
International experience shows us a plan B is necessary. Germany's current energy debacle is the product of two decades of flawed decision-making that is only now being fully realised. Paterson argues that the three errors in Germany's hardcore push to renewable energy consisted of a 'naive' over-reliance on wind and solar, a failure to recognise the importance of a secure gas supply, and the premature closure of nuclear power plants.

Not only are Germans now being told to ration energy, they still emit 8.09 tonnes of CO2 emissions per capita compared with France's 4.74 tonnes. This is despite tens of billions of euros being invested in the renewable energy sector every year for the past 20 years."

Do read the full write-up by Claire Lehmann here (~5 minute read), which cautions against dismissing new-generation nuclear technologies such as small modular reactors, which should at least "be on the table".


5—Further reading...

🙊 Xi Jinping's self-inflicted dilemma: "How do you tell citizens to live with Covid when you've spent the best part of three years aggressively locking down at the first sign of infection and stating that the virus is the biggest threat facing China?"

📉 According to Nomura any stimulus in China is unlikely to be very effective, "as we believe the real hurdle for the economy lies in local officials' more zealous implementation of Covid restrictions rather than insufficient loanable funds".

🕐 Why the leap second is set to go into hibernation until at least 2135.