Direct Lending & AI

Direct lending and private credit

Historically, banks dominated corporate lending, providing the majority of debt financing to both large and smaller businesses. However, the great financial crisis severely impacted banking lending capacity. This was due to tighter regulatory restrictions, capital constraints, and a shift towards more conservative balance sheet management. This void in lending capability paved the way for non-bank financial institutions, such as direct lending platforms, to step in and fill the funding gap. These non-bank institutions pair investor capital with companies seeking capital. Banks like JPMorgan sometimes refer to this as “unregulated shadow-banking,” but the choice of that terminology is due to the direct competition with their bank lending activities.

This is a relatively new investment asset class that we will be continuing to cover in these articles. Unless banks loosen their lending requirements, which seems unlikely, this asset class will continue to grow as it fulfills a necessary gap in our financial system. We are discussing this asset with each one of our clients. If you are interested in learning more, please reach out.

AI Trump v AI Biden

A 24-hour AI-generated Trump vs. Biden debate on Twitch has been running for about four days now. This project was created by a 40-person volunteer group working for SingularityGroup.net. The project’s founder mentioned that it’s a very expensive project to operate. The live stream was banned once, so it’s unclear if it will continue. This starts the long road of AI-generated political content unfolding this election cycle.

AI & Oppenheimer

Luca Dellanna is a fairly well-known complex systems researcher and writer. He recently shared some thoughts on AI risk that were particularly well articulated.

• “The #1 AI risk isn’t some AI becoming a mad dictator and killing humans, but powerful AI tools making it easier than ever for a human dictator to harm us. Because of this, AI risk is not a good reason to pause the development of AI: we cannot afford falling behind in AI development compared to our enemies.”

Given that the Oppenheimer movie is top of mind, this is reminiscent of the logic behind the Manhattan Project. Luca believes that managing AI risk is not a computer science problem but a social governance one. In other words, from a risk management perspective, we should be more concerned with keeping psychopaths out of power than with controlling AI tools.

WorldCoin

Sam Altman from OpenAI launched WorldCoin this week. This indicates that there is some financial silliness back in the crypto market. The WLD token is thinly traded on Binance and KuCoin at an implied market cap of $22 billion, two of the least regulated international markets. There are a few dozen eyeball scanners globally and in the US that will give you WLD coins upon registration.

The purpose of the project is to address universal basic income by creating a global identification to eventually distribute some form of UBI (universal basic income). The biggest practical issue with Bitcoin is the self-custody wallet management system. WLD attempts to attach wallets to biometric information. It’s a slight step in the right direction to solve this problem of simply losing your wallet or getting it hacked. I’ve argued in the past that the FaceID on the iPhone would serve as a great self-custody management tool. FaceID is serialized to the phone itself and cannot be swapped out. Or our banking system can simply get better, as it slowly is with the release of FedNow this summer. I’ve long repeated that the function of crypto might still end up being to create a source of competition for our existing systems. Bitcoin remains the only global money system not controlled by any government, which is why people continue to follow it.

Cuneiform translators & AI

Researchers in Tel Aviv were able to create a Cuneiform translation model with a sufficient accuracy score to translate the tens of thousands of Cuneiform tablets preserved in digital form. The 3,000 year old language was previously translated manually by a handful of experts. The one drawback is that the data set is actually relatively small in the context of training and tuning accurate AI models. The article covering the project can be found here.