Applied AI Examples & Rates

Brief Update On Rates Expectations

The Fed’s interest rate policy is still the foot on the back of markets and the economy. 2023 will present us with the tail end of those hikes, but it’s not over yet. On March 22nd, the Fed will likely raise rates from 4.75% to 5%, a quarter-point (.25%) hike. There’s an increasing probability we will see a .5% hike, but I don’t think that’s likely. We’ve already slowed the pace from .5% to .25%, so accelerating back to .5% at this stage would be a big shift in momentum and quite negative.

The next CPI release is March 14th. The big moment will be when interest rates are higher than inflation. The last inflation reading was 6.4%, but real time attempts to estimate inflation put it closer to 5.2%. The following hike won’t be until May 3rd, over two months from now. That is plenty of time for lagging inflation data to find its way into CPI (specifically housing). Is there much point in raising rates in May if inflation is below 5%? I’m not sure there is. The first expectation of a rate cut is now December ‘23, moved back from November ‘23.

The 1st AI Employee

2023 marks the first time we’ve seen AI “employees”. I call them employees because anyone can “hire” ChatGPT or applications built on OpenAI’s APIs to do all types of generalized work for them. For example, Spotify integrated with OpenAI to power a new AI radio DJ, which provides spoken-generated commentary between songs. It’s never going to replace a DJ, but maybe it makes the Spotify experience better.

The downside to testing AI models is low (unless you want to avoid being insulted by Bing’s AI model). When they work well, LLMs (large language models) can produce incredible results. If you’re on a team, it’s not a bad idea to try handing some tasks to emerging AI models to see what they can do and how they can potentially do it better. Bain just announced an alliance with OpenAI to promote its use among its corporate clients; CocaCola is among the first to sign up. I imagine CocaCola’s multi-billion dollar marketing department sees incredible potential. Meanwhile, JPMorgan has banned ChatGPT in its offices, which makes sense for a large organization primarily focused on risk mitigation.

There is a need for content moderator AI models. IEEE Spectrum has an article on this. ****Eventually we will have AI models working in and around each other. If you think of each model as an employee, you could give the work to your five best models and possibly even use different models to double check the work.

Generative PowerPoint and Excel would dramatically change the lives of many professionals. While Microsoft looks to address that, a generative AI competitor to Powerpoint called Tome just raised a second round by many notable investors, including Eric Schmidt. They reached 1 million users in less than 5 months from launch in September. Tome is also powered by OpenAI.

LLMs can act as a user interface. Here is a simple proof of concept of ChatGPT used to control a robot. In this example, the user tells ChatGPT to generate the code to “throw the blue block in the green bin” and ChatGPT generates the appropriate code to carry out the task. It also asks the robot to stack blocks and create the Microsoft logo with available color options.

Two Blockchain Articles

For those who want to keep up with the blockchain world, I recommend reading this story about the Bitcoin core maintainers and this blog post on private blockchains by a new company, Brale. Private blockchains seem counter to the typically libertarian values of the blockchain community, but it’s a necessary next step in the evolution. Many of the private blockchains are open source copies of existing protocols.