Join institutional investors, analysts and strategists from the world's largest banks: Subscribe today
Already have an account? log in
S&P On Brink Of Bull Market Brings Out The Bears
The bears are out.
With the S&P on the brink of a bull market, many Wall Street strategists remain convinced that 2023's top-heavy, A.I.-inspired trek towards last summer's highs is inherently unsustainable.
Measuring from the October lows, the S&P came into the week a rounding error away from a 20% gain.
It was around these levels late last August when Jerome Powell snuffed out the rally with a terse Jackson Hole speech, setting the stage for a new leg lower in equities and all man
The current AI boom is driven almost exclusively by one technique: LLMs using the transformer architecture. Their capabilities are impressive and ingenious new use cases are discovered daily, often by chaining them together to process each other’s output. The fact that LLMs can be programmed in plain English is even more remarkable.
Every technology has its limits. An LLM is basically a device for predicting the next word in a sequence. Although their potential is huge, it cannot be generalized to solve all problems everywhere.
Just as calls for humanity’s impending doom at the hands of the machines are overblown, I believe the economic impact of this technology is being overestimated simply because the machines suddenly SEEM to have mastered language. In fact, we’ve merely discovered a new way to use statistics to reflect our own skill with language back to us. There is no intention or motivation to these things; they’re just the first technology that comes close to passing the Turing test, and so they feel like artificial general intelligence to us.
Vis a vis markets, there’s going to be a prodigious wave of investment, but any product or service that merely calls other people’s models will be swiftly commoditized by those who control the models and provide the capital-intensive compute platforms on which these models depend. It’s certainly not conducive to creating new behemoths of the future.
Likewise, numerous problems just can’t be solved by predicting the next word. We may have stimulated a wave of discovery in other areas of machine learning, but these will take time to research and develop.
TL;DR the AI revolution is not here now. I’m not sure that markets are pricing the timeline accurately. They do seem to understand who will benefit, though.