Bitcoin’s meteoric rise of 7,887.77% is just a mere piffle of a profit compared to Saint Warren of Buffett’s stock, Berkshire Hathaway which has gained 50,796.23% and is currently trading (at the time of writing) at $539,500.00 per share. But those profit winners take a lower podium standing to the all time winner of the fastest stocks in the past 20 years to attain over 50,000% in rise in value.
That winner is Monster Beverage Corp (MNST), the energy drink maker, with a gain of 87,560%. That gain would have been hard to predict, because they were competing with Red Bull which is still privately owned and has no publicly listed stock figures and growth rates. However based on the fact that Red Bull has enough money to sponsor a Formula One team, as well as extreme sports around the world, including aeroplane racing, would tend to indicate that they are doing very well.
So who is the overall winner in technology? Microsoft of course. They have gained 323,600.00% since 1986. That’s 37 years of waiting. Apple comes second. They have gained 105,535.29% but you would have had to hold that stock for 40 years (since 1983) to get that return. Google and Meta/Facebook are far behind, with 4,713.84% and 654.05% respectively. The biggest tech gainer in the past 20 years is NVIDIA. They have gained a whopping 57,015.85%. A thousand dollar investment would be worth close to $600,000.00! With the next NVIDIA-type stock, you will have to wait only five years for performance like this.
The three NVIDIA founders Jensen Huang, a microprocessor designer at AMD, Chris Malachowsky, an engineer who worked at Sun Microsystems and Curtis Priem, a graphics chip designer at IBM and Sun Microsystems didn't found the company in a garage, but rather at a Denny's Diner in San Jose, California with $40,000 of their own money. They realized that video games were simultaneously one of the most computationally challenging problems and would have incredibly high sales volume. The graphics on video games at the time were about as good as your kid’s crayon drawings on your fridge. Fixing those graphics would be the royal road to riches. So they invented the GPU or Graphics Processing Unit. Then a funny thing happened on the way to the bank.
A bunch of artificial intelligence weenies needed a huge amount of computation to train neural networks with a large number of artificial neurons in a large number of layers of the network. It took for freaking ever. The reason is that every single weight and bias value in every single neuron in the AI machine has to be adjusted thousands, perhaps millions of times to be a little more correct every time a training epoch occurred, depending on the loss function or amount the error the machine had. It took almost as long as teaching a pig to dance. So the AI folks found a way to leverage the gaming GPUs to speed along AI training. That was the act that blasted NVIDIA stock into the atmosphere, to the point where they are now the dominant AI vendor globally. They are a fabless shop — meaning that they don’t build the actual chips, they just design them and farm out the silicon process to others. Some of their chips can cost tens of thousands of dollars.
A reasonable stock analyst could have figured that inventing a chip for gaming would be profitable, but they couldn’t predict that a GPU would be co-opted for AI. But as soon as the academic papers showed up using GPUs to accelerate AI learning, it would have been enough to trigger investment in NVIDIA for the discerning, savvy trend analyst.
So is it possible to foresee companies today that will match the investment performance of NVIDIA? You bet your sweet bippy. And I can guarantee you that it won’t take 20 years. More like 5 years or less.
The meteoric rise of ChatGPT out of nowhere, and its game-changing massive adoption creates equally massive opportunity. The opportunities are created with faster and faster fulfillment cycles. “Massive opportunities for what?” you might ask.
ChatGPT has changed the world. It has changed humanity. And the large language models that underpin the basis of Natural Language Processing using AI are proliferating like mushrooms after rain. There are many competing Large Language Models, and they are all pretty much as useful and as revolutionary as GPT (GPT stands for Generative PreTrained Transformer). But there are huge problems with LLMs. And as a result there are opportunities. Here are some of them that are not the biggie, yuge winner that I identified, but will still bring on the cash:
Large Language Models all hallucinate, or make stuff up. The genius who solves the hallucination problem will be hailed as the next Edison of AI and be as rich as Elon Musk.
Incorporating other data modalities. Can you imagine an LLM watching the news on a television screen and then automagically creating a screen summary, or doing one-shot learning from it? This is a biggie. It would make digital assistants like Siri or Alexa either extremely useful or extremely dangerous.
Large Language Models are trained in English. They operate on English language tokens. Other language functionalities just translate the prompt into English and translate the response back to the original language. Not all languages have the same structure and syntax in English. Translation and applying LLM transformer functions makes for awkward and sometimes nonsensical output. You need native language capability. Now that the population of India has surpassed that of China, the first company that comes up with a Hindi language LLM will be a unicorn.
Inventing an LLM that is faster and cheaper without the required tagging of billions of tokens using cheap 3rd world labour will be the next big winner in AI. There are one-shot learners, but they usually sit on top of larger LLMs which may be derivatives of even larger LLMs.
The guaranteed unicorn and stock market winner will be the company that ubiquitously puts LLMs in the mobile space.
The public company that develops a new transformer algorithm for LLMs will be a stock market winner. I would need to write a book to explain how the LLM algorithm works presently and how it could be improved so that the average Joe or Josephine can understand it, but the layman’s explanation is that a transformer in an LLM can take a sentence and analyze it with self attention. (For understanding, when referring to this kind of self attention or intention, it will be in bold italic). It means that transformers can determine the relationship of words in a sentence that are far away from each other instead of doing next word analysis. This is what is meant by attention. They do this through a sequencing matrix. To improve the architecture of the transformer requires a lengthy technical explanation, but I can explain it in one sentence using technical jargon. It is this:
“For the existing Transformer architecture, the complexity of attention is quadratic in sequence length and the complexity of an MLP (Multi Layer Perceptron — the neural network) is quadratic in model dimension. An architecture with subquadratic complexity would be more efficient.
Let’s leave it at that and move on to the next one. Someone will be sitting at a Five Guys burger joint and come up with the Eureka moment for this one while digging out the last of the fries out of the bag.
The person who develops an alternative to GPUs will be the next billionaire and the stock will hit record highs.
Design is king. Look at Apple. Whoever comes up with a new chat interface that doesn’t resemble a 2010 website will be a winner. It has to be multi-modal, mobile friendly and the coolest thing since a penguin’s bum.
Okay, so all of the above is semi-obvious boilerplate blah-blah-blah. It is very non-specific. And sure, all of the above points will all make money, but the really huge big winner isn’t mentioned above. And it will make most of the issues listed above go away faster than the Christmas spirit after the family turkey dinner. I promised you a specific real winner like the GPU. It will be something that will have the same effect as NVIDIA’s GPU but bigger. That company will be the first company that comes to market with something known as an ….
The company that makes this and first brings it to market will be the next NVIDIA beater! There you have it. If you want to catch the next wave of stock market unicornity, the company that does this will have their stock explode into clouds of cash. My web scrapers every night, are trawling the oceans of the web to find the company on the cusp of bringing birth to this next big thing in AI. And when it happens, don’t forget that you read it here first and didn’t do anything about it. So mote it be.
Thanks for reading and don’t forget …. this is not investment advice.
Great information Ken. You need to spin off a premium service that identifies these potential biggies early in their ascent. That would be of value to many.