Commodities, AI, Automation

Commodities Falling

Commodity market prices have dropped swiftly over the past few weeks, which should help ease inflation worries. Oil is almost back where it was pre-Ukraine invasion, in the mid $ 90’s. Copper is down -30% from its high in February. Natural gas is down -40% over the past month, which translates to lower prices for fertilizer and electricity. Wheat and corn are down -30% from the highs. It appears increasingly likely that the shortages of the past year will be followed by excess supply.

We mentioned last week that the Fed does not use market prices in its decision to raise rates. They target the CPI (consumer price index) print published every month. Wednesday, July 13th, is the next print. These falling commodity prices will take some time to make it through the system before they end up in the CPI, certainly not as soon as July 13th. We’re unlikely to see an end to this bear market until we see that official inflation print change which seems to be another month or two down the road. Aside from small investments in March, we continue to wait to deploy any new capital, although many companies and categories are enticing.

Transformer AI

Initially published by Google in 2017, an artificial intelligence training tool called a ‘transformer’ has dramatically improved AI capabilities over the past few years. Transformers process collections of data in any order while simultaneously noting unique relationships. It’s like reading all the words in a book simultaneously while storing any and all unique patterns. It’s also like looking at an image, every pixel at once, and organizing it into distinctive sections. Nvidia CEO Jensen Huang said in March ’22 that transformers have allowed AI to “jump to warp speed.”

The results vary but can often produce shocking levels of quality. Transformers are credited with DeepMind’s AlphaFold breakthrough which solved the protein folding problem, OpenAI’s GPT-3 model which can write realistic generated text, and OpenAI’s Dall-E 2 which generates images using text prompts. Since Dall-E 2 was announced a few months ago, copycats have quickly emerged. The Economist used the MidJourney for their magazine’s cover art in June. Google created Imagen and HuggingFace published a lightweight open version called Craiyon.

It’s breakthroughs like transformers that allow for dramatic leaps in progress, and it appears that many more AI applications will be coming soon. While many of these AI breakthroughs are open source, realistically, they require serious engineering talent and resources to unlock their full potential. Technology companies with dedicated AI teams and access to significant processing power will benefit the most from the coming commercial rollout of these AI tools. AI can be a vague term, but one way to think of it is that it allows for new levels of precision previously unattainable for computing. For example, FIFA has been testing AI-based referees for the past few years and has already allowed it to issue penalties.

Automation adoption

There’s a robotic arm in Portland, Oregon, called Jarvis that can make precise espresso drinks, including latte art, using traditional coffee machinery. Jarvis is produced by Universal Robots, which sells these robotic arms for roughly $20,000, far less than the annual recurring cost of a barista. However, just because we can automate a task does not mean companies or customers will opt in, regardless of the cost-efficiency. As we saw with COVID, companies often wait until they are forced to adopt newer and better technology. For example, FedEx finally set plans to migrate from their servers to the cloud, a move that most companies have made over the past decade. FedEx estimates the savings will be greater than $400m annually. Given customer preferences, it seems unlikely that robots will replace baristas beyond just the novelty factor. However, as living wages and inflation continue to pressure the economics of all businesses, we will likely start to see increases in automation adoption. What would happen to coffee shop economics if the cost of that robotic arm eventually dropped to just $2,000? At what point can you not afford to ignore the alternatives?