24/7 Markets, Immigration, Quantum Primer

24/7 Markets

This week, there was a big push in the media by various parties to convert the stock market into a 24 hour market. Crypto, forex and certain futures markets already operate on a 24-hour trading schedule. Markets often move more when they are technically “closed”. Earnings are the prime example of this, where companies wait for the market to close to release their results only to have the stock pop or drop in the “after-hours” trading sessions. It makes you wonder, what is really the point of closing only part of the market and allowing a select group to continue trading in off hours? Nearly everyone has access to after-hours trading, it’s not an exclusive function. Robinhood already offers 24/7 trading for select companies which they administer internally.

Gamestop in 2021 is a good example of how markets in their current configuration are not ready to handle 24/7 trading. A two week delay in short-selling reporting allowed hedge fund clients to short more shares of Gamestop than existed in the market, roughly 120% of outstanding shares. A 24/7 market needs 24/7 continuous reporting, not a two-week or even a one-day lag, otherwise, something breaks. Markets are always changing and adapting, and 24/7 markets are an inevitable future. The only limitations left at this point are a handful of policies and some legacy systems waiting to be upgraded. This is a blind prediction based on gut feeling, but I put the timeline to 24/7 equity trading four years from today, not that far out.

Immigration supporting the economy

I liked this quote from this recent article on immigration. “Ernie Tedeschi, a research scholar at Yale Law School, estimates that the labor force would have decreased by about 1.2 million people without immigration from 2019 to the end of 2023 because of population aging, but that immigration has instead allowed it to grow by two million.”

Positive net immigration is a critical input to continued economic growth in the US. Every developed country in the world is experiencing a shrinking population with birth rates well below replacement rates with zero signs of reversing. The US economy would undoubtedly shrink without immigration, which leads to economic recessions and even deep depressions, and no one wants that.

Some predict certain developed countries will start to compete for immigrants by offering economic incentives. South Korea and Japan have some of the worst local demographic profiles with birth rates below death rates, already aging populations and little to no immigration. South Korea’s population today is 50 million and without a dramatic change its own government anticipates its population will drop to 30 million by the 2070’s. Japan’s government predicts its population will drop by 30% over the same period.

Quantum Primer

AI hype has me wondering what breakthroughs are coming next that will allow the momentum to continue. The scale of the current AI operation is hitting some upper bounds of what’s physically possible. Zuckerberg has acquired as many NVidia H100s as he can, and Sam Altman declared that the world does not produce enough electricity to support his vision. NVidia’s newest chips help address efficiencies, sometimes 2-3x better, but not orders of magnitude better. Quantum computing seems to be a potential emerging solution for reasons I will outline below.

Richard Feynman is one of the founding fathers of quantum computing and the subject of one of my favorite biographies, Surely You’re Joking, Mr. Feynman. You can access his original paper on quantum computing from 1982 here: “Simulating Physics With Computers.” It’s written in both simple and technical language. Feynman is famous for saying, “if you can’t explain it in simple terms, you don’t understand it.” He’s been known to publicly call out entire educational institutions for repeating definitions without demonstrating understanding.

The word quantum refers to the smallest possible “quantity” of physical property within an atom. Feynman said, “nature isn’t classical, if you want to simulate nature you better make it quantum.” He is referring to classical computing based on 1s and 0s (bits).

In classical computing, all data is stored in a binary format as either a 1 or 0, exactly like a light switch on or off. Quantum computers use qubits that can handle exponentially more data with fewer operations, but I still don’t understand exactly how or why. It has something to do with storing, linking and interacting with probabilities. I’m clearly not alone as even practitioners admit a vague wholistic understanding of quantum physics. One analogy provided by Cloudflare is that classical computing will read a book page by page to find a particular answer, while quantum computers can first read the table of contents and then skip to the relevant chapter.

Computers are all about scale, so imagine how valuable this is when taken to its extreme. Cryptographic security exists beyond the limits of what classical computing can handle, making it safe and secure by definition. Algorithms that would take thousands of years for a classical computer to solve can theoretically be solved in minutes with a quantum computer.

The increase in complexity at the bit level from 1s and 0s to qubits improves certain capabilities by orders of magnitude. NVidia’s GPUs are orders of magnitude faster than CPUs at machine learning tasks which allowed machine learning to thrive. It stands to reason that quantum machine learning algorithms could also be orders of magnitude better than current GPUs. Thanks to AI hype, there is now a massive incentive to develop quantum machine learning algorithms.

Evidence of tangible quantum progress can be seen in how quickly top global security teams are taking corrective action. Apple rolled out post-quantum encryption for iMessage in February. Google Chrome rolled out post-quantum encryption by default last week. Since that rollout by Google, 14% of all web traffic on the highest security standard (TLS 1.3) is now post-quantum secure. That may be as high as 35% next week. The WEF with IBM published quantum ethics and policy guidelines in January. People don’t act without incentives, so this tells me that quantum has recently made real progress.

IBM has one of the top quantum computing teams globally. Their machines are named after birds. IBM Heron is a working 133-qubit machine with the highest performance level. IBM Condor has 1000 working qubits but 5-times the error rate, which makes it lower performance and not ready for public use. You can rent out actual quantum cloud computing from IBM Heron here, at the rate of 10 minutes per month for free or $1.60/second paid. I signed up for a free account. It appears to be optimized for researchers at this stage. Of course, the major cloud companies have their own quantum teams. Amazon holds a stake in the publicly traded quantum stock IONQ (not advice).

Training new machine learning models today can require hundreds of millions in electricity costs and can take months. Quantum computing cannot be used to train AI models today, but it’s a fair bet that quantum machine learning is coming. I like this phrase, “the likelihood of doubt being correct diminishes with time.” I’ll add that Feynman’s original interest in quantum computing was its ability to simulate physics, which could unlock some amazing opportunities for discovering new materials.