On Thursday, Sep 3, 2020, there was a massive sellout of tech stocks. The NASDAQ dropped 5% on Thursday. Apple was down 8 per cent; Amazon was down 4.5 per cent; Microsoft was down 6%. The losses were further extended on Friday, Sep 4, 2020.
Tech is the rage now and may possibly be the future – these are the companies that brought about new technologies which change the life.
In this article, I would take about investing in tech.
Wait, Isn’t The Title About Technical and Fundamental Analysis?
Yes, when I said investing in tech, I meant investing yourself in some tech knowledge. Who knows it might come handy in your future interview with a tech company. Or maybe you would like to use new technology to upgrade your products and services.
The hot trend now with tech is AI and ML, just put those acronyms in anything, and you can market it as something brilliant and cool.
AI
Back in 1999, the trend was adding a ‘dot com’ to boost your company’s brand – it makes it seem high tech. The current trend now just adds ‘AI’ to your product name, description or features and that differentiates you from your other ‘dumb’ competitors.
So what is AI?
AI stands for artificial intelligence. It basically means teaching machines to think like humans and solve problems like humans.
In the early days of AI, someone would have to program the machine/computer with a set of logical rules – if some input then performs this action etc.
Deep blue was a chess computer that beat the Russian grandmaster Garry Kasparov in 1997. It is a classic example of traditional AI where the algorithm calculates and evaluates all possible positions on the board and chooses the most optimal next step.
Algorithm?
An algorithm is a sequence of steps or instructions to perform a calculation or solve a problem. It is like an instruction manual for the computer to follow. A classic example is a sort algorithm which is an algorithm to perform sorting operations (You may encounter these questions for software eng/tech/coding interviews).
So, deep blue is not exactly intelligent, in a sense, compared to humans. It has enough computing power that it can brute force calculate all possible moves and choose the best. So, if you can already see the future, you just have to select the best outcome.
For humans, we don’t have that much brainpower to simulate 10,000 combinations at once. But what makes humans intelligent is our ability to be creative and to make decisions with limited data or imperfect information.
Recently, with the explosion of computing power and data, machine learning became more popular.
ML
Machine learning (ML) is a subset of AI.
ML is about teaching machines to find patterns, to ‘learn’ from data.
Unlike traditional AI, humans do not explicitly program a set of rules. Instead, humans feed the machines large datasets to train the machines to find a pattern.
So, suppose you want a machine to sort the names of your students. In that case, you can explicitly program it to look at the first letter of the name and sort according to the alphabetical order.
However, if you are given a picture of a dog, how would you program a machine to tell if it is a dog? You cannot just say to the computer to look for two ears, one nose and a snout.
Hence, ML is used to solve such problems without implicitly telling the machine what a dog looks like. You just have to show the machine a picture of a dog, and with a large enough sample (millions) the ML model will begin to learn what a dog looks like. That mimics the way humans learn. If you keep showing a kid pictures dogs, once the kid sees a dog on the street, he will cry out: ‘dog! dog!’. But humans are better in learning because you probably only need to show a kid a few pictures for it to learn how to identify a dog. Whereas an ML model probably needs hundreds to thousands to be able to identify a dog.
Typical ML usages are handwriting recognition, speech recognition or image recognition. These are challenging problems to program but easy to solve via ML (if you have a large dataset).
Deep Learning
Another cool buzzword is deep learning, which is a subset of ML. Deep learning mimics the human brain using neural networks. When you hear neural processing engines, it basically means hardware designed to run neural networks.
Neural networks basically try to imitate networks of neurons in your brain, and they can be trained repeatedly to learn and perform tasks.
Alphago was considered an achievement when it has beaten a grandmaster in go. Unlike chess, go has infinitely more possibilities and variations, and it is simply not possible to brute force calculate every position of the board. (Maybe with quantum computers we might be able to brute force but that is for another topic).
Alphago is an example of an ML model that uses neural networks. It had developed more human-like intelligence by learning from experience.
Indeed, the next frontier would be developing general artificial intelligence. So far, when we teach a machine to solve a problem, it only knows how to solve that given problem. General AI is when it can solve any problem even if it was not programmed to solve it specifically. It could learn like a human to perform a completely new task.
General AI is also referred to as strong AI. Weak AI refers to a particular implementation of AI and does not have any intelligence (at least at the human level of intelligence). Again, another topic and definitely not suited for this blog.
AI and ML In a Nutshell
If you want to program a computer to read words out loud, here is how traditional AI and ML approach would differ:
In traditional AI, you would program the computer to learn how each alphabet sounds. Then you would teach the computer to apply phonic rules when letters are combined together to make the sound of the words. It is explicitly teaching the rules of how to read.
In ML, you would just collate thousands of words and record how each word is pronounced. Then you feed all these data into the ML model without explicitly teaching it the alphabet or any rules. It just learns the sounds of words from the data.
You are Still Not Talking About Fundamental Analysis and Technical Analysis
Ok. I digressed.
Fundamental analysis and technical analysis are methods which analysts use to determine the price of a stock.
Fundamental Analysis
For fundamental analysis, you will look at the financial statements and other economic, market factors to calculate or estimate the intrinsic value of a company. If the value of the company is higher than the stock price, the analyst would recommend buying the stock because it is undervalued. Vice versa, if the intrinsic value is lower than its share price, then the analyst would suggest selling it.
Fundamental analysis is similar to the good old AI – you will need to understand the problems and the rules and constraints governing it.
In the example of deep blue, it understands the game of chess and abides by the rules of the game. Deep blue will employ typical human strategies (eliminating more pieces of the enemy, trapping the enemy into disadvantageous positions, etc.). Still, with its supercomputing capability, it can simulate more possibilities and choose the best strategy to secure victory.
Technical Analysis
Technical analysis, on the other hand, only looks at the price and volume of the shares traded. Technical analysts believe that all information about the company should already be reflected in the price of the security. What technical analysis does is to look out for trends or patterns and try to predict where the price might go.
In a sense, it is like machine learning. In machine learning, the only information you need is the label for your dataset. For the example of the dog recognition ML model, the label will be whether the picture is a ‘dog’ or is ‘not a dog’. The ML model will just look at thousands of data and try to find similarities or patterns of pictures, which are labelled ‘dog’.
In Technical analysis, you don’t really need to understand the company, analyze its board of directors or even know the business.
Machine learning models also do not need to understand the fundamental concepts. They just need to find patterns and reduce their prediction errors.
Summary:
| Fundamental Analysis | Technical Analysis | |
|---|---|---|
| Approach | Reviews fundamentals of a company – financial statements, economic factors | Assumes fundamentals are reflected in price of the company. Only interested in looking at price and volume to identify patterns and trends |
| Understanding the company | Yes, you will need to understand how the company operates, its business model etc. | No a requirement. Only price movements in the stock market matters |
| Output | Calculates intrinsic value of the company | Identify patterns and trends to predict its future price |
Conclusion
I merely wanted to highlight the key principles behind fundamental analysis and technical analysis and draw parallels with AI and ML.
In fundamental analysis, understanding the core business, its strengths and weaknesses are essential – that is how they value the company. It is similar to how traditional AI is build – write rules and program a machine to understand and solve the problem.
In technical analysis, finding patterns and trends are essential – only the price and volume matters. You don’t really need to know the business to do technical analysis well. It is similar to how machine learning operates – finding patterns within large datasets. An ML model does not need to understand the inherent problem to solve it.
I know I made a long digression into talking about AI and ML, but I thought it might be educational.
It might be a stretch to compare AI, ML to fundamental and technical analysis.
AI is a broad term which can mean anything. A simple heuristic rule to solve a specific problem can be AI (weak AI); A complicated deep learning model can also be AI.
So next time, when you see a fund or a roboadvisor that uses the name AI, it does not mean anything. You should peel the surface and peek into the algorithms or models used by the fund or company.
Fatty's Finance 