QQQ Rebalance & AI Model Collapse

QQQ – Nasdaq Rebalance

The Nasdaq 100 index is in the middle of a special rebalance due to the overweight of the largest positions. Microsoft, Apple, NVidia, Google, Amazon and Tesla collectively make up over 50% of the index. In order to maintain balance, Nasdaq will redistribute the weights by selling those positions down slightly and increasing the other 94 weights. Much of the 2023 performance has been concentrated in those top names. There are many names in the Nasdaq that are still well below their highs so this rebalance into other securities will likely benefit investors looking beyond the top mega cap global tech companies.

By this time next week the Fed will meet to adjust interest rates again. Markets anticipate a 99.8% chance of another .25% rate hike. An additional rate hike is entirely unnecessary as it comes at the additional expense of taxpayer dollars. Runaway interest expenses have eclipsed a $1T annual rate. Official inflation is at 3%, Truflation’s estimates are at a new low of 2.13%, but housing is still stuck near 7%. Housing costs are slow to move. As we said last time, these additional rate hikes do not make a difference other than to signal sentiment to market participants. The market has more or less moved on given that the rest of the rate curve stopped drifting upwards back in October.

AI Model Collapse – The Curse Of Recursion in AI

The first AI models were trained on original human generated content. Up until now that’s more or less all that has ever existed except for a handful of algorithmically generated news stories. As the volume of AI generated content explodes, the risk of accidentally training a model on AI generated data increases. When this happens the biases are reinforced and errors are compounded over subsequent AI generations. Futhermore, if AI generated data is undetected in early generations, it will be very difficult to nearly impossible to fix in subsequent trainings and iterations. Most art, and really most of everything we create, is derived from previous works. Most successful modern stories are iterations of historically successful character narratives dating back to Greek mythology. The line in the sand for what constitutes original content is blurry at best. Sarah Silverman has attracted attention on this topic for suing ChatGPT for what appears to be evidence that the model was trained on a pirated copy of her book. As brilliant as AI is, it does not fundamentally create truly original content and given the recursion issue described above, original content is becoming more, not less, important. Something for the writers and Hollywood to consider.