Are you worried about AI? Pundits talk about how it is going to change our world like nothing before it, possibly putting millions of people out of work. If you’re an investor, the advice columns scream about jumping on the bandwagon to invest in AI companies while the opportunity is still ripe.
This article is going to put it into perspective. For starters, AI means artificial intelligence, and it is anything but new. The first mainframe computers to be used in business applications date back to the 1950s and every one of them had AI — double entry accounting systems, which is one of the most intelligent creations in our history. Think about it. Double entry accounting with its statements of revenues and expenses, assets, and cash flow has allowed businesses to function with accurate and audited financial information, which has led to robust stock markets worldwide.
Those first mainframe computers artificially implemented what humans had been doing manually for close to 700 years before computers started doing it far more accurately. In short, all computer programming is a form of AI. So why are we all of a sudden going gaga over AI as if it’s now somehow different? It’s because as computers have become far more sophisticated in their capabilities, so has the programming. But that’s it. There is no magic or even mystery about AI. It’s computer programming, plain and simple, but with more sophisticated applications like actually writing articles and even making videos from descriptions.
The AI Buzzwords
The increase in sophistication has led to a series of new buzzwords that sound almost scary: pattern recognition, machine learning (supervised, unsupervised and reinforcement), natural language processing, neural networks, computer intelligence, and general computer intelligence, to name a few. It turns out, these new concepts are almost entirely related to the ability of today’s computers to handle graphic-oriented processing as opposed to character-oriented processing. The early computers from the 1950s mainframes to personal computers well into the early 1990s primarily handled characters, which enabled accounting systems, word processing, and spreadsheets to be programmed.
Beginning in the 1990s, PCs started handling graphics files and by the 2000s started handling audio and video files as well. Today, the typical PC can handle video just as well as your television. The move from character processing to graphic processing has opened the door to more sophisticated applications, but that’s all they are. They’re programming, not human intelligence, which functions at a completely different level. In fact, the buzzword general computer intelligence describes computers that will function with the kind of human intelligence we exhibit today, and they do not exist. It’s not clear, furthermore, that they will ever exist.
So What’s the Big Deal with AI?
The big deal with AI is that the door is now open to new kinds of programming based upon linking together PCs with backend cloud computing systems that can process huge amounts of data -both text and graphics. What’s important to understand, however, that in order for this type of programming to work properly, it requires a specific and relatively limited application. For example, a bank might have a computer answer the telephone and talk to the caller, which has existed for many years in primitive form. Today, that computer can listen to the answer and help the caller in a wide variety of situations. Each situation, however, is well defined. Furthermore, if the caller has a situation the computer can’t handle, it then passes the caller to a human customer service agent. The benefit to business is that it may reduce the number of operators that it needs or, better yet, it may lower the time it takes to reach a human operator, which enables the bank to provide better customer service in competition to gain and keep its customers.
The key point is that each application has to be specific. For example, we now have access to almost every piece of music that has been recorded since the 1920s, and the data base is growing by leaps and bounds with new music. There’s absolutely no way for any individual to begin to know the number of songs and musical artists they might like. But the computer system that simply plays requested songs can also categorize the various music it has available, keep track of what any individual listens to, and then suggest artists with similar music that the listener may enjoy, even to the point of creating a special playlist. Understand though that it only works within the context of a specific application.
Automated Driving
One of the ultimate potential applications is programming cars to drive themselves. It is essentially real-time video processing with the computer recognizing the various elements that humans have to recognize to safely maneuver their cars. Again, there’s no magic. It’s computer programming. The computer, however, has to be able to recognize the many variances it will encounter, such as differently shaped traffic lights, and unexpected occurrences, like a ball entering the street with a child chasing it. It is not human intelligence. It is a specific computer application that programs all of the activities that it takes to safely drive a car. It is, of course, artificial intelligence, but then so was the first mainframe computer program that recreated a double entry accounting system.
Chatbots Are Augmented Memory and Actiivities
Chatbots like ChatGPT are the application that has largely started the current wave of AI frenzy. You can type in all sorts of questions and it will give back an answer, which will only be as accurate as the data base that it is searching, although it may certainly seem official enough, but an example is in order. Back in the early days of personal computing in maybe 1982, I used a spreadsheet inside the company I was working for to project the market for a project I was working on. The projection was complete guesswork, but it looked official because it was typed out beautifully from the spreadsheet. Interestingly, the projection was considered somehow better and more official by my colleagues because of how it looked. Unfortunately, that kind of reaction hasn’t changed over all these 40+ years. When the chatbot gives an answer, it may well be considered more accurate than it actually is.
But before falling for Chatbots hook, line, and sinker, let’s understand where their real value lies, which is augmenting our memories and activities. Consider that very few people are really good writers, and a large number are actually bad writers. Chatbots will unquestionably help people to be better writers by proofreading what they’ve written and even doing the first draft of a paper. Teachers will have to adjust and figure out how to distinguish the computer’s work from their students’s work, but they should be able to do it. Writing, of course, is just one activity. These bots will help us with many others, especially those that will benefit from sophisticated pattern recognition, which is a very weak part of human intelligence. Computers are much better at looking through great gobs of data and finding meaningful patterns.
Tradesmith’s AN-E 2.0: an AI Irony
Here’s an example of an AI application that ironically goes well beyond the ability of human intelligence. Stock market information company Tradesmith has developed an AI system, AN-E, that they say can accurately predict the movement of a stock for a period 30 days. Its value requires the use of options trading. Let’s say the system predicts the stock would go up during the 30 days. An option would allow the investor the right to buy a certain number of shares of stock at the price that day. If the stock goes up as predicted, the investor would then buy the stock at the lower price and then sell at the higher price, making a profit that could be several times the price of the option.
In the typical world of options, humans are largely just guessing whether any specific stock would go up or down. Options trading is notoriously risky because human intelligence is really bad at determining whether a specific stock will go up or down over a period of 30 days. Nevertheless, if this can be done accurately over 30 days with a computer program, it has the potential to obsolete 30-day options trading by eliminating the risk. Instead of such a system being artificial human intelligence, it is a new kind of machine intelligence that goes beyond human capabilities.
The Bad and The Ugly
While today’s so-called AI systems will help our abilities, there is also the potential for a myriad of abuses. Deepfake images and even videos is an obvious negative consequence. For example, it’s easy to imagine an AI bot creating a deepfake image of Donald Trump shooting some on 5th Avenue. As a result of such a capability, our justice systems might encounter all sorts of problems distinguishing what images might be real. It’s also possible for cybercrooks to create all sorts of sophisticated scams. Finally, we should expect that there will be a high degree of bias in Chatbots for the simple reason that people with biases will program them. This is all life in today’s computer programming fast lane, and such outright abuses are just the Bad. There’s also the Ugly.
As we’ve seen, AI really isn’t new, yet it’s being hyped as if it is. In fact, the hype has already become so bad that almost every single product on the market now has some type of AI claim associated with it, which renders the term almost meaningless. When everything is revolutionary, nothing is revolutionary. In short, the hype is that AI somehow makes a product or service more accurate or better than before using AI. The reality will almost certainly be highly questionable, albeit with select exceptions that in many cases will perform tasks that humans couldn’t do well. These systems will indeed change the world in select ways. The AN-E 2.0 system from Tradesmith is an example. If it’s possible to accurately predict stock movements for a period of 30 days, everyone would want to use it to make money. The result, however, will be the opposite. Options trading systems will simply disallow taking an option for a period of 30 days.
Another issue will be bots (either chat or voice) used to perform educational tasks like writing a paper. Teachers will have to accept the positive, which is that student writing will appear to improve, but they’ll know that a machine likely wrote it and not the student. As such, teachers will adjust their strategies to determine what the student has really learned from it. This issue isn’t really new. Ever since the development of electronic calculators and the internet, the need for students to memorize has declined precipitously. The world of AI bots simply extends this further to other educational tasks. Smart educators will recognize these as tools that help students learn, and will adjust their methods of having students demonstrating what they have learned.
Will AI Take Our Jobs?
The one thing we won’t have to worry about is AI taking all our jobs. The reason is simple. If AI takes our jobs, then who will buy all the stuff that the companies produce who are using AI? The one thing we can be certain of is that any additional wealth created by AI will get spread around to our middle classes who drive the engine of commerce through their consuming. In short, a series of new jobs will develop so that people can afford to continue consuming the stuff that companies make. How will it happen? That we don’t know, but it’s a virtual certainty that it will.
The Bottom Line
The bottom line is that AI really isn’t new. What’s new is that the continuing power of computing, especially the promise of networked computing called neural networks, will allow powerful new applications that provide us with a series of increasingly capable tools. Neural networks, incidentally, comes from the recognition that our brains work through an incredibly sophisticated neural network. The problem for computing, however, is that we have no idea how the neural network in our brains actually works. As such, we don’t know how to program networked computers to simulate how our brains work. The best we can do is leverage the power of networked computers to solve problems associated with pattern recognition, which ironically performs tasks that often go well beyond human intelligence. While we’ll continually hear about how such capability is a threat to humankind, don’t believe it. In the end it will only be a benefit.
*************************
If you like the article or podcast and want see more, please enter your email to subscribe to our email list. You can opt out at any time.