Have you heard about the Perceptron? It is not a Transformer, but it could transform your crypto investing strategy. Tune in to learn more.
Do you have a question you would like me to answer on the podcast? Follow me on IG: @indestructiblewealth and send a message, or visit me at www.myindestructiblewealth.com for more resources.
About Indestructible Wealth: I’m Jack Gibson. I’m your wealth strategist and I’m here to help you make some money. The Indestructible Wealth Podcast is for young entrepreneurs who want to make, keep and grow wealth to enjoy now, and for years to come.
Episode #77 – Juice Your Crypto Returns with Artificial Intelligence
Podcast Episode Transcripts:
Disclaimer: Transcripts were generated automatically and may contain inaccuracies and errors.
The human brain has long been the inspiration for artificial intelligence (AI). It’s an extraordinary organ comprised of 100 billion neurons connected in a complex and largely unknown web of interconnections. Our brains are capable of incredible feats in a split second that the most powerful supercomputers struggle with today. And the brain can do this with about the same amount of energy that is required to power a small lightbulb.
It’s no surprise that scientists have long thought to create an artificial neural network (ANN) based on the structure and connectivity of our neurons. In fact, the earliest attempts to do so go as far back as 1943, before the world had invented the first semiconductor. The earliest model of a neuron however was unable to “learn,” and in that regard, it didn’t hold much promise. But by the 1950s, things got far more interesting thanks to a Cornell psychologist by the name of Frank Rosenblatt. Rosenblatt invented something called the Perceptron. The novel idea behind a perceptron was to be able to apply weights to the connections between neurons. It’s the weighting of a connection between two neurons that enables an AI to “learn” over time.
As an artificial neural network trains on a set of data, it compares inputs and connections to the optimal outputs. Through the process, the neural network recognizes which connections consistently produce more accurate results. When the neural network finds these strong connections, it applies a higher weight to those connections. In contrast, it applies lower weights to connections that produce less accurate results. And with each iteration, all weights are changed and the process repeats. We can think of this as a living, evolving system. As more data is made available, the neural network will incorporate that data and update those weights as conditions change.
Perhaps an easier way to think about these weights is to use an example for our own lives. As we build our network of friends, family, colleagues, and advisors over the course of our lives, we learn whose opinions we can trust and whose we should dismiss. Through our own experiences, we figure out which people from our networks consistently give good advice or information versus those who tend to do the opposite. And obviously, depending on the people, the value of that information can both improve or lessen in value over time.
Our brains essentially apply a weight to each individual person. We of course do this both consciously and subconsciously. It weighs heavily in our own decision-making process as we learn what to trust and distrust. What if a machine could simulate this complex process? What if a machine could “learn”?
Our mission here is to use a proprietary neural network to find high-conviction trades in the world’s most exciting markets: cryptocurrencies and digital assets. I can confidently say that Neural Net Profits is unlike any investment research service in the world. This took years and hundreds of thousands of dollars to perfecting this neural network. And for reasons I’ll explain in a moment, this technology will not be found on Wall Street or in any hedge fund in the world.
The solution is a proprietary neural network that was developed, a technology called the Perceptron. But before I show you how the Crypto Perceptron works, it’s important we understand this technology. While a perceptron might sound like science fiction, it is a very real technology with a storied history.
The first perceptron was run on an IBM704 computer. Compared to what we are used to today, it was a series of systems that filled an entire room. The system was archaic by today’s standards, and it couldn’t do much. For decades, it was the lack of computing power that held back progress in developing neural networks. The theory of an artificial neural network was sound, but the computing power simply didn’t exist yet. This was frustrating for the industry. After all, while our brains are capable of incredible things, they are inherently limited. There is only so much information that we can store in our short-term memory ready to use at a moment’s notice.
This is where the power of seemingly unlimited data storage and computing power comes into play. The recent breakthroughs in semiconductor technology over the last five years have brought the use of neural networks back to life after decades of languishing. It is the reason why our smartphones are millions of times more powerful than all the computing systems used for the Apollo missions to the moon. But the real explosion in growth, the massive shift, happened in 2014.
That was when bleeding-edge semiconductors had around 5.7 billion transistors. And not surprisingly, it was around that time that the acceleration of progress in neural networks absolutely took off with a mind-boggling 2.6 trillion transistors. That’s about 27 times the number of neurons in the human brain. Our brains are still better at doing certain tasks, but for most complex problems, an artificial neural network running on this kind of hardware is far more capable of outperforming the brain.
These recent developments in neural networks and semiconductor technology have enabled us to use the most advanced semiconductors and computing systems, combined with the most advanced forms of AI, to create an AI-powered trading system that can predict the movements in asset prices before they happen. The Perceptron is designed to manage any number of data inputs, many hidden layers which explore all possible outcomes, and of course, a predictive output that provides strong signals about which assets the Perceptron expects to move higher in the days and weeks ahead. Put even more simply, we are “unleashing” the power of a neural network to predict which assets will move sharply higher. And the asset class this machine is focused on is the cryptocurrency market. I can’t think of a more exciting sector to apply bleeding-edge artificial intelligence to a trading system.
Cryptocurrencies are a fantastic market for a neural network like this because of how large and liquid the market is, and how quickly it is changing. As we can see in the chart at right, the total cryptocurrency market exploded from around $190 billion in January of 2020 to more than $2 trillion today. That’s more than 10X in just two years. We’ve never seen an asset class achieve this kind of scale in such a short period of time. And that’s across more than 16,000 different cryptocurrency projects.
no matter how involved in the industry they are, no one knows it all. We all acknowledge that there is simply too much happening at the same time to keep track of it all. It is literally impossible for any one person, or team, to be able to grasp everything that is happening, in real-time, for such an active asset class that is growing so quickly.
And that’s exactly the purpose of the Perceptron. It has human-like intelligence that has the ability to recognize patterns in the way that humans do, but on a scale, unlike anything even a team of the smartest people could possibly do. There are simply too many interrelated variables to track by one person, all changing in a dynamic system every single day. They trained the Perceptron on the last five years of data from around 6,800 cryptocurrencies. Neural networks in particular benefit from larger data sets. That’s what helps it learn and identify the patterns that result in strong buying and selling signals.
So How does The Perceptron Work? Every day, the Perceptron processes 4 gigabytes of data representing 200 different data points for each cryptocurrency. This includes everything from price and volume, to liquidity, concentration risk, and even social media activity related to the underlying assets. The Perceptron considers more than 200 data points, 4 gigabytes of data, every day.
The Perceptron “crunches” this data with one objective: Find the assets poised for a sharp move higher. And here’s what may come as a surprise: We don’t and can’t know exactly how the Perceptron “decides.” That’s the nature of deep learning and neural networks. All we know is that it works. The mean trading return from a strong signal from the Perceptron is 93.4% in our testing. This compares to a 33.7% return from the benchmark performance of bitcoin (BTC) over the same time period. It’s hard to overstate what an extraordinary performance difference that is. It is nearly 6,000 basis points, or 60 percentage points better than the benchmark. In the world of hedge funds, outperforming a benchmark by 600 basis points is thought to be an outstanding performance. The Perceptron is generating returns 10 times that.
Through backtesting, it was determined that the strongest signals are from a 60 day time horizon. That means that we know that we can expect the big moves in the underlying asset within 60 days of receiving the signal. Again, while we might have an idea of why the system is recommending the trade because of our extensive research focus on cryptocurrencies, we won’t know the exact reasons why the Perceptron made the choice. And this is to our advantage. Using a 60-day time horizon for trading, we won’t be competing with the high-frequency trading systems that dominate the cryptocurrency markets. It is estimated that as much as 80–90% of all cryptocurrency trades are made by computers running algorithms.
We don’t want to compete with the hedge funds running these systems as they are just trying to eke out fractions of a percent of profit on each trade… usually at the expense of normal investors. I can guess what we’re probably thinking: With this type of outperformance, why hasn’t Wall Street used their own perceptron already? It’s a fair question. And here’s the answer…
Imagine you are a data scientist working at a storied hedge fund on Wall Street. You have developed your own perceptron, a neural network that spots high-conviction trades.
One morning, your perceptron delivers an incredible signal for a particular stock. You are floored. You know this AI works, that it finds winning trades again and again. And this is the strongest signal you have ever seen. Excited by your discovery, you run to your manager down the hall. You share your discovery. You are adamant. We must buy this stock. And we must buy as much as we can right now. You explain how you’ve never seen a reading this strong before, that it’s practically a sure thing. And this would be the first question you would be asked: “Well, why did the perceptron pick the stock?” And the answer – as we know – is “I don’t really know…”
This is the roadblock to the adoption of this technology in traditional money management. After all, hedge funds have limited partners they are accountable to. And there is no hedge fund manager in the world that will deploy hundreds of millions of dollars into a stock without knowing why he or she is buying it. The reason is that they must be able to provide an explanation to their investors for such a large trade. Inevitably, this results in hedge funds reverting to machine learning-based more on statistical analysis rather than taking advantage of the dark arts of deep learning. The results for these strategies can be “explained” and are therefore acceptable.
But there is an element of mystery to neural networks and how they “decide” on an outcome. For this reason, perceptrons have not been adopted by institutional investors. And this is our advantage. As intelligent, self directed investors, we are empowered to make our own decisions. We are empowered to use this technology for our gain, even as Wall Street misses out.
We’ll wait for the strong signals from the Perceptron, build our position in the cryptocurrency trading on large digital asset exchanges, and wait for the price action to follow. If and when the signal starts to weaken, we’ll close out our position and take the profits off the table. In backtesting of the Perceptron, they spent thousands of hours of high-performance computing systems to refine the performance. The results are extraordinary. Shown above is the performance of the Perceptron in blue. We compared this to the performance of bitcoin (BTC) as a proxy for the cryptocurrency market.
the Perceptron has consistently outperformed both the broad cryptocurrency market and BTC by a wide margin. If we look at the April 2021 peak performance of bitcoin around 450% and compare that to the peak of the Perceptron around that time of 1050%, we see an incredible difference in performance. Even during times when bitcoin is in a downtrend like it was last summer, the Perceptron significantly outperformed.
The results demonstrated that the system was able to deliver a positive return 85% of the time. This kind of performance is unheard of, and I can’t be more excited to make it available to my subscribers. We’re going to run circles around the best performing hedge funds in the world and deliver incredible annualized returns. Every month, subscribers should expect one or two perceptron trades to act on. We can’t know exactly which days the recommendations will be made as we’ll only make recommendations when the Perceptron delivers strong, high probability signals for the underlying assets. When we receive a strong signal, you’ll get an immediate alert with the following information: Crypto ticker Project Name Entry Price Market Capitalization Exchanges that the asset trades on Category of the cryptocurrency
when the signal weakens, we’ll send out a quick alert letting subscribers know that it’s time to close out the position and take profits off the table.