Practical AI, Rates

All eyes on rates

For the first time since Nov 2021, we are seeing the first real reprieve from the persistent downward pressure on markets. The overwhelming strength of the dollar peaked in October when the GBP hit 1.06, the EUR hit .95 and the Yen hit 150. A weaker dollar will help earnings going forward. As expected, this week the Fed raised interest rates to 4.75%. The absence of any surprises was enough to continue the recovery momentum present since the start of 2023.

All eyes are on rates because they are falling everywhere except at the Fed. Mortgage rates are down, futures markets expect rate cuts as early as November. In all likelihood, we will see another .25% hike on March 22nd, but that matters less than it used to. A 4.75% risk-free rate is nice while you wait, but soon that will fall and it’s better to get moving before that happens. Please reach out if you would like to discuss where we plan to allocate new capital at this phase.

Just about all of the relevant trends have reversed and held. Whether we’re in a sidewise market from here or a new bull market matters less because, barring an exogenous shock, we’re at the end of this Fed rate hike chapter. “Don’t fight the Fed” works in both directions.

Practical AI

The visible portion of AI’s evolution began last year with the announcement of Open AI’s Dall-E 2. ChatGPT was first released two months ago and has now surpassed 100 million users, the fastest accumulation of customers of any business in history. AI tools are becoming meaningfully useful to those who take the time to integrate them. The GPT3 API is now part of my daily workflow in unexpectedly helpful ways.

I recently came across LogoAI.com, which instantly generates iterations of logos, fonts, color schemes and brand assets on demand. Generative AI art is a neat trick, but generative brands and marketing materials are extremely valuable for all businesses, especially young ones looking to save on startup costs.

The cost to train AI models will come down over time and the limited data sets will expand. At some point you will train your own models on your own datasets. Someone recently trained a model on all of Tesla’s earnings calls, for example. It was neat, but not practical. The evolution of practical AI applications for consumers and businesses will continue to multiply at warp speed.

At a higher level, researchers at Salesforce’s AI division developed a biotechnology AI tool called ProGen. ProGen allows researchers to explore the “grammar and semantics” of protein design. The model was trained on 280 million protein sequences. While AI drugs have been created and tested in the past, I think this is more of an impressive stepping stone for researchers than a destination for actual drug development and rollout.

For those interested in all things AI, I recommend reading “The Age of AI” by Henry Kissinger, Eric Schmidt and Daniel Huttenlocher.