Weekend Reading #302

This is the three-hundredth-and-first weekly edition of our newsletter, Weekend Reading, sent out on Saturday 15th February 2025

To receive a copy each week directly into your inbox, sign up here.

*****

What we're thinking.

It’s happening! With all the attention on AI (and maybe even because of AI who knows) China is really breaking out. None of this pretend stuff we saw over the past year or so. Commodities are going nuts; Gold is soaring but the real one to watch is Dr Copper which is closing in on $5. The copper stocks have begun to move too. The trashing of the dollar appears to be in motion after a week or two of flirting. What does this all mean? It means the real fun is happening outside of America for the first time in donkeys’ years. And hold on, an end to the war in Ukraine? Everyone seems to have forgotten about that! To be honest apart from all the politics, it would simply be fun to trade Russia again should we ever get to that.
 

What we're reading.

This week I finished reading The Immortal Mind by Ervin László. László describes himself as a philosopher and systems scientist. In English that means he specialises in the nature of consciousness, my pet topic. Much of this book builds on other things I have been reading and learning about this topic including research about near-death experiences, reincarnation and what exactly is the Akashic field. His conclusion is that consciousness is primary versus the current mainstream scientific approach that the brain generates consciousness. Lots to dig into here and I recommend for those as curious about this as me!
 
This is a really thoughtful thread by Arjun Balaji on meme coins. It aligns with much of how we think about it. The prolific proliferation (am I allowed to say it like that?) of meme coins has diluted capital across more and more of these speculative instruments, rendering them mostly a place where capital goes to die a slow and petered out death. The Trump meme coin has thus far proved to be a crescendo for this. Could we be about to see a leadership change inside crypto from Solana (meme coin casino coin) to Ethereum (future of finance coin)? DC
 
Going further down the AI rabbit hole, and the question of what values are expressed and proliferated by your favourite LLM, I came across this paper called “Utility Engineering”. It of course stands to reason that the training of the LLM, and the tweaking of weight values by humans inadvertently passes (and amplifies) those values on to the modelled outcomes – remember an LLM is really a big inference formula. With the implementation of reinforcement learning processes into systems (an approach made popular by DeepSeek R1), the optimisation functions that govern the inference of a prompt into a “most likely to be correct” output recursively run on themselves. The result is they optimise for optimisation and doing so means they become even more entrenched in their values. And it turns out that this may be a recursively slippery slope that makes itself simultaneously steeper and more slippery. Consider what that means in terms of “values” that are expressed, and ultimately adopted as truth by the masses? So, what, you may ask? Here’s one to think about – the implied valuation of different types of human life by GPT-4o:
 

 
Having an exchange rate on different human lives? That’s something we definitely don’t need the world to go back to.
 
The link to the full paper is here: https://arxiv.org/pdf/2502.08640
 
Bottom line: as the complacency around assuming your favourite AI is always correct because it “seems” clever (it isn’t, not in that sense) grows, and people start taking the outputs as gospel truth, we might find some really troubling values being disseminated. If TVs got a reputation for being an “idiot box”, then falling into full reliance on LLMs could well be the equivalent of having one’s brain sucked out and replaced with indolent dependence. The only way around it is awareness and a minimum level of actual competence to know what’s outright wrong. EL

 

What we're watching.

I have been following this company in the US called Boom Supersonic which is bringing back supersonic flight. The amazing thing is that we had supersonic flight in the form of the Concorde as long ago as 1976! The Concorde stopped flying in 2000 after its famous crash. 25 years later, Boom (a decade old startup) is reinventing supersonic flight and last week completed the final test flight of its XB-1 plane, which is a testing plane 1/3 the size of what its upcoming passenger plane, Overture, will hopefully be. If you think about it simply: Why shouldn’t there be supersonic flight? If the safety is good, why wouldn’t we want to collapse travel times? It’s actually quite unbelievable that there hasn’t been more evolution of businesses like these. This technology was commercialised 50 years ago! Here’s to hoping we get it back in the coming years.
 
Curtis Yarvin is someone I have heard referenced quite a bit in recent times but only last week did I dig into his musings and his blog. I had read how he has influenced many of the tech titans in Trump’s administration but even I was surprised at how obvious it actually was! Have a read of his newest blog, his older one (stopped in 2007) and this Youtube video gives a good intro into his philosophies (as his alter-ego, Mencius Moldbug). DC
 
 

What we're listening to.

I’ve heard a lot about Graham Duncan and listened to his Invest Like the Best appearance some years ago but when I listened to the most recent one I appreciated his perspectives about finding talent a lot more. This guy goes deep into understanding people, and it really is worth listening to this one. Really deep and insightful in a podcast world where that is becoming more and more rare. DC