Buck Jordan here to talk about one of my personal passions – artificial intelligence (AI).
Now, I’ve written to you a few times before about AI as it relates to the AI investment boom, the global competition to be an AI superpower, and, most recently, the controversy about Google’s LaMDA AI program. But I think I need to spend a little more time walking through the nuts and bolts of AI and all the different pieces of the AI puzzle.
Let’s start with a dictionary definition.
“Artificial Intelligence” is the simulation of human intelligence processes by machines.
In plain English, AI is the ability for machines to learn and make decisions at the same level as, or better than, humans.
We’ve been dreaming about AI for hundreds of years and started really getting excited about the idea with the dawn of the information technology (IT) revolution starting in the 50s and 60s.
For decades, the advancement of AI was held back because it requires tremendous computing power. For perspective, an iPhone today has 100,000X more processing power than the computer that landed Apollo 11 on the moon in 1969.
AI requires far, far more computing power than a smartphone can offer. It is only with the dawn of cloud computing throughout the last two decades that progress in AI has been unlocked, and things are accelerating every year.
Learning… Not Coding
The first important component of AI is the ability of machines to learn – this subsection of AI is known as “machine learning.”
Unlike traditional computer programs, AI isn’t based on rules-based systems and if-else statements. In other words, AI systems can’t be hardcoded with instructions on what to do and when to do it. One of the core theses of AI is that an AI system can learn, adapt, and improve its decision making over time.
This is where machine learning comes in.
So how does machine learning work?
Well, the key is having vast amounts of data. The next step is to train the AI system on this data to teach it to do something specific.
Let’s look at an example.
If you want to train an AI system to recognize dogs versus cats, the first thing you need is a large dataset of cat and dog pictures. What is important is that this data set is labelled. So, for every image of a cat, you want to have it labeled as a cat, and for every image of a dog, you want it labeled as a dog.
Once you have this dataset, you run it through the AI system. In that process, the AI learns to identify certain characteristics of cats and dogs in images and is “trained” to distinguish between the two. The labels on the data help guide the learning process and provide a feedback loop for the algorithm.
That is machine learning in a nutshell – large labeled datasets and training an AI system with feedback loops. Now that you understand the mechanics of machine learning, I’m sure you can imagine dozens of different use cases for this.
Algorithms for Everything
The next component of AI is algorithms. Again, let’s start with a dictionary definition.
An “algorithm” is a process or set of rules to be followed in calculations or other problem-solving operations, especially by a computer. So, an algorithm is a methodology and set of rules to solve calculation problems.
Different algorithms solve the same calculations in different ways, sometimes with better or worse results. In addition, different algorithms are also designed to solve different problems. The kind of algorithms that can recognize cats versus dogs in images are different from the kind of algorithms that can recognize someone’s voice (such as the ones that power Apple’s voice assistant, Siri).
There are a huge number of algorithms out there from “naive bayes,” “random forest,” “k-nearest neighbors,” and more and more. Chances are you don’t know what any of those are, and that’s OK.
While it is fun to learn how these different algorithms work, you don’t really need deep understanding to make smart AI investments.
Putting It All to Use
Once you have identified the right algorithm, built a large dataset of labeled data, and trained your AI system on this data, you come to the last component of AI: finding an application for it.
We’re still very early in the process of applying AI to the real world, but the use cases have already been extremely promising.
For example, one of the most exciting areas of AI deployment is healthcare. AI algorithms are being used to analyze 3D scans 1,000 times faster than humans, listen to 911 calls to predict heart attacks, and make hospital and clinic operations more efficient.
The use cases of AI are endless, but the key is to find use cases that are valuable enough to make a big difference for customers while being problems AI is mature enough to solve.
Artificial intelligence can be an intimidating concept.
While there is a lot of jargon, mathematics, and computer science involved, the reality is we can all understand how AI works at a conceptual level.
Machine learning, algorithms, and applications are the three core components of AI, and I’ve been using this framework for years to guide my investments in AI and robotics companies… with great success.
You can too.
Until next week!