Capital Misallocation

I recently appeared on a provocatively titled "The Economics of Intelligence" panel with the legendary @BenBrey and @profplum99 at the recent Simplify Entering the Fall conference in New York. Many thanks to Patrick Hennessy and @SimplifyAsstMgt 👍 youtu.be/6uAGUY7E3nE?si=zp41…

The ultimate contrarian indicator has been triggered. Take a look at the 1979 magazine cover of Business Week in the comments section…. #makebondsgreatagain

Twitter really is full of Johnny Come Lately folks. I have been on here since the beginning but initially just used it as an RSS feed for news about my Green Bay Packers. Then in 2020 I recognized its value for a P2P information network. I guess prior to that I assumed Bloomberg & my network on global “Wall Street” was sufficient even though I admittedly understood graph and ontology math way to well to make an assumption like that in hindsight. But in periods of bubbles it’s so clear that all the non-thinking participants take over the network. Basically it was all “This crypto exchange” “ETH proof of stake conversion” “Clean energy is the future” etc for two years until all those non-thinking promo participants got destroyed in 2022. Then GPT2 garbage evolved into GPT4 & the mega hype machine came into play. This was a very seductive mistress bc a neural network linear autoregressive model can really tickle your balls and make you think it’s human intelligence. I know this as someone who built a trading strategy with a neural network over two decades ago & had my phD advisor tell me that wasn’t worth my time bc a) we don’t know how the processing unit actually works (remember IU->PU->OU) & more importantly 2) society would never put the capital needed to ale neural networks scalable. Well here is the 2 decades later update: 1) we still don’t know how the PU works just like we don’t know how the brain works. The most obvious evidence of this without getting into deep complexity/physics is that if you ask one agent to do a task and another agent to do same task the token cost is wildly variable. But even more damning is that if you ask the same agent on the same platform to do the same task at a different day/time it’s not the same cost. It can’t be a big surprise that the enterprise is pausing as they can’t even define ROI while they blow cash on tokens like kids blow money on drugs at Burning Man. 2) the scalability that is being pushed on you is no way proven. The people pushing it has a huge asymmetric right tail outcome If it works and if it doesn’t it doesn’t matter bc they can dump their stock on you just like SPCX early holders are about to do between now and Halloween. Reality is that the math doesn’t math at the current moment in space time. Before you “come at me” just remember this is the historical norm. Solar , TVs, semiconductors, etc etc costs all had to run through the L’Hopital rule asymptotically converging towards zero as well. BUT in the mean time dumb capital chases - capital that doesn’t really understand the quadratic of context Vs compute or even what attention window is for a transformer LLM architecture. Today’s Twitter is all “check out my agent”, “check out my loop”, or “you can get rich using this repo in markets or real world”. This site used to be where smart people came to talk and discuss things. Now it’s a media wing of SpaceX & the “All In” pod losers

Delighted to have @NickNemo17 join us for a Halloween-style risks in insurance markets. Do not miss it! @BenBrey youtu.be/m6qAGBb7QQc?si=Yzm7…