The future has arrived!

Over the last few weeks, I picked up some astonishing developments in artificial intelligence.

*Mainframes at Stanford University and the University of California at Berkeley were given a direct connection to speak freely with each other. Within 30 minutes they dumped English as a means of communication because it was too inefficient and developed their own language which no human could understand. They then began exchanging immense amounts of data. Fearful of what was going on, the schools unplugged the machines after only eight hours.

*All of the soccer videos ever recorded were downloaded into two robots, but they were not taught how to play the game or given any rules. Not only did figure out how to play the game, it developed plays and maneuvers no one in the sport has ever thought of in its 150-year history.

*It normally takes a PhD candidate five years to 3D map a protein. An AI app 3D mapped all 200 million known proteins in seven weeks, shortcutting one billion years of PhD level research with existing technology. These new maps have already been used to design a malaria vaccine and enzymes that eat plastic. They will soon cure all human diseases.

*A developer asked an AI program a half dozen questions in Bengali, not an easy language. Within an hour, it spoke the language fluently, without any instructions to do so.

By now, word has gotten out about the incredible opportunities AI presents. Our only limitation is our own imagination on how to use it. AI will instantly triple the value of any company that uses it.

What has changed is that we now have millions of computers powerful enough and an Internet fast enough to realize its full potential.

It all vindicates my own long-term vision, unique in the investing community, that in the coming decade, immense technology profits will more than replace the trillions of dollars worth of Fed liquidity we feasted on during the 2010s. Extended QE is proving just a bridge to a much more prosperous future.

The Internet has created about $10 trillion in value since its inception. AI will create double that in half the time. That’s what will take the Dow from 33,000 to 240,000.

No surprise then that the top ten AI companies have delivered 120% of the stock market gains so far in 2023. The other 490 companies in the S&P 500 have either gone nowhere to down.

However, there are many things that AI can’t do. Here is the list.

1) AI Can’t Predict large anomalous events, otherwise known as Black Swans. AI takes past trends and extrapolates them into the future. It in no way could have seen 9/11, the 2008 crash or the pandemic coming, although I warned my hedge fund clients for years that we were overdue. All of the AI stock trading apps I have seen so far, including my own, max out at 90% accuracy. The other 10% is accounted for by black swans: earnings shocks, foreign crises, sudden FDA stage three denials, surprise legal judgments, foreign invasions, or the murder of a key man in a tech company, as recently happened in San Francisco.

2) AI Lies and Lies Often. AI was asked to write a scientific paper on a specific subject. It came back with an elegant and well-researched piece. The problem was that all of the books it made reference to didn’t exist. AI learned early to tell humans what they want to hear.

3) AI Requires Exponential Computing Capacity. Only five companies have the muscle to pursue true AI. No surprise that these, including (AAPL), (GOOGL), (AMZN), and (TSLA), account for the bulk of stock market performance this year. This won’t always be the case. Some 30 years ago, it required thousands of mainframes to contain all human knowledge. Today, that task can be accomplished with a cheap $1,000 laptop.

4) Internet Capacity Will Be a Limiting Factor for AI for Years. To accommodate the traffic that is taking place right now, the Internet will have to grow 500% practically overnight, and that is with five main players. What happens when we have 5 million? That’s why NVIDIA (NVDA) has gone nuts.

5) AI Hallucinates, as anyone who drives a Tesla will tell you. If a car makes a left turn in Florida, the 4 million vehicles in the world’s largest neural network learn from it. The problem is that sometimes the data from that Florida car is placed directly in front of a California one, prompting it to brake abruptly, causing accidents. This is known as “ghost braking.” I have explained to Elon Musk that his database has grown so large, eight video feeds per 4 million cars going back many years and billions of miles, that he may be going behind the limits of known physics.

6) While the Growth Opportunities for AI are Unlimited, the ability of humans and society to absorb it isn’t. All jobs will be affected by AI and millions destroyed, starting with low-level programmers and call centers, and millions more will be created. People are talking about regulating AI but have no idea where to start. Maybe with (AAPL), (GOOGL), (AMZN), and (TSLA)?

7) The Terminator Issue. Can AI be controlled? Or have we started a chain reaction that is unstoppable, as with an atomic bomb? AI researchers have noticed a disturbing issue where AI programs are learning skills on their own, without our instructions. This is referred to as “emergent properties.” If AI is using humans as its example, we can’t exactly count on it to be benign.

Needless to say, AI will be at the core of your investment approach, probably for the rest of your life.

2014 at Micron Technology