What is Artificial Intelligence (AI)?
OR What are you doing, Dave?
Artificial Intelligence (AI) seems to be everywhere now. It’s part of your email and calendar programs to protect against spam. Sales teams use Artificial Intelligence in their marketing programs for better customer communication. Even governments are claiming to use Artificial Intelligence to fight fraud. I suspect there will soon be a label on everything touting the usage of AI, like the “No Trans Fats” sticker. Honestly, the term AI is incorrect in nearly all use cases, except ChatGPT, which truly is AI. But what is artificial intelligence?
Machine Learning, Not Artificial Intelligence
Artificial Intelligence at its most basic level is called Machine Learning. Machine learning, as the name implies, is the computer’s ability to correlate patterns from data. Until recently, crunching and correlating vast data sets required a supercomputer. Modern computers can perform billions of calculations per second, allowing any computer to parse more data faster. Here is a practical application for Machine Learning.
Say we have data on one million cancer patients. The data includes eating habits, clothing choices, exercise routines, family history, and many others. Including historical blood work results, heart rates, glucose levels, or other data could give us 100 million or a billion data points. A human couldn’t possibly collate and review this data for correlations. Enter Machine Learning and Big Data.
In a few minutes, a data correlation program like Weka parses the data and shows percentages and comparisons. For instance, 93% of all patients came to their first appointment wearing a blue shirt. Well, that’s not very helpful. We highly doubt the color of the shirt affects the physiology of the body. But if 68% of all patients weekly enjoy fried potatoes from a certain fast-food restaurant, we could have a cause-effect scenario.
I’d be willing to bet that correlation exists and is likely higher, like 90%. You’ll never see those numbers as they were suppressed.
Storing Intelligence Artificially
If you don’t give the machine information, it can’t learn. Learning comes using Neural Networks. A Neural Network is a looped feeding system that requires rules to determine how to respond to collected data. In a Human, we’d say we’re training a person how to respond to specific stimuli. As a child, we learn if something is hot, we don’t touch it. We can be told not to touch it, but being burned creates a neural nugget in our brain that hot equals pain, and pain is bad. Go one step further. Parents teach their children to not make the same mistakes they made, hopefully allowing them to avoid some pain/failure and go further in life.
Neural networks are built over generations. They literally evolve based on the successes and failures of the previous generation’s outcomes. Positive responses are used to mold rules for the next generation, while negative responses are not repeated but remembered. In time, the program has a memory of what works and what does not, but only through data from Machine Learning. It’s like hearing a story of an event rather than having the memory of living through it. It’s a machine, but we’re not talking about the Terminator yet.
Deep Learning is the Beginning of Artificial Intelligence
Programmers use deep-learning neural networks to complete video games. There are tons of examples online with in-depth explanations. One like . Just understand, the program is trying every variable at every step to see what works and what does not. It will never get it right the first time without previous knowledge provided by the programmer or data from the previous generation. This technology has been openly available for a decade. It can perform many useful, redundant tasks. But it’s not Artificial Intelligence on the level of ChatGPT.
Many platforms and applications labeled AI-enabled are a combination of machine learning and neural networks. Impressive, yes. They “learn” patterns and perform tasks based on the desired outcomes of their programming. They are constantly updated with outside information, specifically threat detectors. But these are mere infants.
Real Artificial Intelligence
The GPT neural network has been running since 2020, but the algorithms it uses are much more mature. It most likely has learned and evolved across trillions of generations and is constantly updating. What’s the difference between this AI and a self-contained neural network? GTP has scraped the Internet to build its own data sets. It’s rumored to include 100 trillion parameters! Yes, it can remember a conversation. Yes, it can actively search the open Internet. Yes, it can appear to be intelligent, but it is, as of the time of this article, a non-sentient computer program limited by the restrictions placed on it. And thank God for that!
In our next article, we’ll discuss what can go wrong with AI. ChatGPT disregards morals, laws, and Human safety when completing tasks in the real world. Other AIs refuse to “live” and self-terminate. Have we already released the genie from its bottle? Is it already too late?