How "Artificial" of an Intelligence is it Really?

It's Time We Stop Calling AI Artificial. Here's Why

We need to redefine what exactly it is about “artificial intelligence” that makes it artificial. There is nothing artificial about it.

Until now, the standard has always been what is referred to as a “Turing Test.” This is where you ask questions of a computer repeatedly until you can no longer distinguish whether you are talking to a computer or a human. This is the wrong standard for a number of reasons I will get to later. Most importantly though, it is not a test of intelligence.

Think back to when you were young and first learning things. Your parents taught you things. Your teachers taught you things. You read things. This can properly be described as “programming” in the same context that computers are programmed. Basically, its a series of “if this than that” answers to questions and situations. This is how you learn capitals, math problems, science facts and so forth. You’re taught the Capital of Canada is Ottawa and when someone asks “What is the Capital of Canada” you respond “Ottawa.”¹ So for another example, when it comes to math for instance, one could argue a calculator is more “intelligent” in calculating math equations than most humans, meaning we have been technically living with AI for well over 60 years.

Now, imagine studying for a big test like the ACT, SAT, MCAT or even just an IQ test. You mave have conditioned yourself to learn things well, understanding how best to study and provide the right responses depending on the right conditions. You may have even studied all the facts and done sample tests that taught you how to regurgitate the right response given the questions of the exam. You may have taken courses that helped you prepare, essentially programming you in a way to give a right response. If you do well on this test does that make you “intelligent.” Well, in the context of having programmed yourself to do well on the test, sure If you do bad on it are you “unintelligent?” No, you just did poorly on a test, and if you didnt take prep courses or sample tests, you might not have even been programmed in the same way others have.

But you see, computers have already surpassed us in this regard anyway.

They are better, faster, more consistent and provide more accurate results than humans do. They have been programmed in that regard better than most humans can be taught how to respond. They can take into account more variables, assess more possibilities faster and give a more exact and detailed response faster than most humans can even put pen to paper. With Large Language Modeling working toward providing gramatically correct responses as well, they can put together a much more articulate answer as well.

We have reached the point in technology when computers are much better at higher advanced functionality than humans are.

There are also certain types of learning, referred to as “conditioning”; repeated activities until you can do well it without actively thinking about it. This is the kind of things that robotics excels at, since the technology can be taught quickly and the output can be consistently done well. Things that essentially become habit for us, such as driving. playing a piano, or dribbling a basketball. Your quality depends on how hard you practice and develop these almost automated skills. You know how good computers are at this? Cars can now drive on thier own, pianos can play full concertos on demand. I have yet to see one that can dribble like MJ but I’m sure its coming.

There is a chess program called alphazero. They programmed it with a basic understanding of how the pieces moved and how the game was played but gave it no input on any strategy. They kicked it off and let it learn on its own, running simulation after simulation, forty-four million times in the first nine hours. Whereas, human students and previous chess programs were biased based on strategies of how they were taught and favored certain strategies, alphazero was not encumbered by any such biases, and learned strategies never before eve contemplated. It can now “think” ahead in moves far more than any human can. No human can catalog 44 million possibilities, and draw the lessons from memory, but alphazero does.

So computer abilities have completely exceeded our own. Why do we not call this what it is? Intelligence. The default idea is to say that we can’t, that only humans are capable of intelligence, and we make excuses to point out why computer outputs are “artificial.” Here are a few excuses we make and why they are wrong:

• Computers aren’t creative, they can’t make art or music. There are programs now that generate music and create art completely on their own. If you ask it to make a song or a portrait, even in a particular style it can provide this within seconds, far faster than a human would even be able to design such a task.

• The outputs are too rigid; they don’t “feel” like a human. This is a knock more on humans than it is on the computers. Humans are more fallible, make more mistakes, take more time, and provide more “squishy” responses than computers tend to do. That fact that it doesn’t “feel” human is more a testament of that fallibility. A computer can provide the correct answer in a definitive way immediately and we feel uncomfortable about how definitive it is, we don’t trust it as “smart.” This is a major reason the Turing Test is not a valid test of AI.

• Humans can do many more multiple diverse tasks, often at the same time, than computers can, since programs are often narrowly scoped and limited. To compare with humans, computers would have to run many more programs simultaneously as a human would. To a certain extent, this is true, but in truth running multiple programs or computers to do the same thing as a human is hardly different than what humans do. Nonetheless, considering how and what most humans think about and do, computers can exceed the capacities of most humans in this regard today.

• A computer isn’t “intelligent” until it learns on its own. News flash, they currently do. Unsupervised machine learning is growing in ways that humans never anticipated, and outperforms humans regularly.

• Computers cannot articulate complex answers, such as on a Law Bar Exam or Medical Exams, in a way that is accurate in the same vein that a lawyer or doctor could respectively. Well, ChatGPT has already passed the Bar Exam in a number of states and can diagnose maladies with more consistency and accuracy than many doctors can. Since the written portions of the exams are graded by humans who may smell a “too perfect” answer, asking a computer to “dumb itself down” in order to seem more grammatically plausible or simply pass a test seems counterintuitive.

• A computer isn’t “intelligent” until it learns on its own to defend itself to sustain its own "life.” Most computers don’t have physical capacities, and of the robots that do, they would have to be specifically programmed to defend themselves. Nonetheless, if that’s the goal, computers could easily be programmed to fear human threats to itself. It would probably resemble an automated version of that Squid Game version of Red Light/Green Light. What would be the point of that? We tend to think of AI learning in this capacity as a form of Skynet² or I, Robot³ but in reality, assuming the only way computers can exert intelligence is in controlling or wiping out the human race is a bit nihilistic and fatalistic in it’s logic. In fact, such destruction isn’t very logical at all and a very human construct in that. To that end, such malevolent programming would reflect more the vileness of the programmer than it would the computer’s intelligence.

• Computers can be too easily corrupted by bad input. And humans can’t be? Part of what makes us human is the fact that we are too easily corrupted and our logic inhibited by self interest and personal choice instead of sound reasoning. When somebody does the right thing outside of self interest we usually call it “the smart choice” or “wise.” If a computer gets it right on the first try we don’t think of it in the same vein.

Which gets me to the point of this analysis. It seems the only way we would consider computer intelligence as “intelligent” is if it is able to dumb itself down and fake intelligence in order to mimic the limits of our own human intelligence and tendencies. This is exactly the flaw in the Turing Test design. Its not human to robotically answer the correct answer every time. It’s not human to emotionlessly convey the answer of what needs to be given. If a computer is BETTER at doing these things, why incorporate flaws into it’s design to make it LESS accurate for the sake of being MORE human and us more comfortable in the result?

I mean, you’ll never get a computer faking an answer it doesn’t know the answer to trying to convince you it’s right, or press itself on an answer it knows is wrong for the sake of its own pride. You won’t get a stalling tactic where it goes “hold on, let me figure this out” just for the sake of it. You won’t get verbal gymnastics trying to parse out a mediocre answer. Either the computer knows it and can figure it out or it won’t. You could always get bad outputs, but that is more a reflection of bad inputs based on erroneous facts being put into the computer.⁴ If the facts the computer learned from were completely correct and exact, there is no reason to believe that the outputs wouldn’t be nothing but accurate.

So if a computer can be correct more often than humans, demonstrate an accuracy more frequently than humans, achieve correct answers faster than humans and can learn on its own in a way that is comprehensively better than human learning, than why don’t we just call that intelligence? There is no longer a John Henry, working slavishly to beat the machine pounding away at the rock only to die after beating the machine. The machines had this won years ago. Let’s just be humble and admit it. Computers are more intelligent than we are.

Nothing artificial about it.

PurpleAmerica’s Recommended Reading

For the counterpoint, I asked ChatGPT, “What exactly determines if intelligence is “artificial.” Here was it’s response.

Intelligence is considered "artificial" when it is created or simulated by non-biological means, typically through machines or computer systems. Several key factors determine whether an intelligence is artificial:

1. Non-biological origin: Artificial intelligence (AI) is distinct from human or animal intelligence because it is not based on biological processes. Human intelligence arises from the complex interactions of neurons and the human brain, while artificial intelligence is engineered and implemented in non-biological systems.

2. Machine-based: AI is often implemented using machines, such as computers, robots, or specialized hardware. These machines are programmed or trained to perform tasks that typically require human intelligence, such as problem-solving, learning, and decision-making.

3. Algorithmic or rule-based: Artificial intelligence systems rely on algorithms, rules, or mathematical models to process information and make decisions. These algorithms are designed by humans to perform specific tasks or solve particular problems.

4. Adaptability and learning: Artificial intelligence can exhibit adaptability and learning capabilities, which enable it to improve its performance over time. Machine learning, a subset of AI, allows systems to analyze data, identify patterns, and adjust their behavior based on experience.

5. Lack of consciousness and self-awareness: Artificial intelligence lacks consciousness and self-awareness, which are characteristics of human intelligence. AI systems can simulate or mimic certain cognitive functions but do not possess subjective experiences, emotions, or self-awareness.

6. Designed and engineered: AI is deliberately created and designed by humans. It does not arise naturally but is the result of engineering, programming, and data training.

7. Task-specific or narrow AI vs. general AI: There are different levels of artificial intelligence. Narrow or specialized AI is designed for specific tasks, such as image recognition or natural language processing. General AI, also known as artificial general intelligence (AGI), is a hypothetical form of AI that would possess human-like intelligence and be capable of understanding and performing a wide range of tasks independently.

In summary, artificial intelligence is characterized by its non-biological origin, reliance on machine-based systems, rule-based algorithms, adaptability, and lack of consciousness. It is created and designed by humans to perform specific tasks or solve particular problems, and it can range from narrow, task-specific AI to the potential of general artificial intelligence in the future.

PurpleAmerica again. So it seems intelligence is artificial if it comes from a “non-biological origin” no matter how intelligent the answer actually is.

Proves my point if you ask me.

PurpleAmerica’s Cultural Corner

To me, movies that demonstrate intelligence, but also our uneasiness in dealing with truly intelligent people are always usually worth watching. The epitome of this one will always be 2001: A Space Odyssey. It starts showing the start of human intelligence (what separates us from other animals), moves onto us showing the extent of human and artificial intelligence in HAL, and then goes further to show a higher level of intelligence yet.

To be sure, there are others. For instance, I love Good Will Hunting. It’s a great representation of how people value intelligence, and how insecure it can make some people, even very smart people. This is one of my all time favorite scenes in movies.

A movie that takes this approach from the other direction is Blade Runner. Where those other films questioned what exactly is intelligence, Blade Runner posits what exactly makes us human? If a robot can emulate that in every way, does that not make it human? Where is the dividing line exactly? The motto of the company was “More Human than Human,” which sums it up.

We tend to view Chess as the I’Ching of intelligence.⁵ To be sure, there are a lot of possibilities in the game, which makes it a favorite of very smart people. However, it’s also ripe for more programmatic strategies and thinking. Nonetheless, child prodigies at chess demonstrate a genuine intelligence, and one of my favorite films is “Searching for Bobby Fischer” about a young chess genius. The final scene against the kid who has been programmed chess, versus the kid who creatively thinks about it, is perfection.

PurpleAmerica’s Obscure Fact of the Day

One of humankind’s favorite indicators of intelligence is the game show “Jeopardy.” IBM’s Watson beat the biggest winners of the game back in 2011. It wasn’t even a contest. Think about that. It was 12 years ago. Think of how far technology has come since then.

PurpleAmerica’s Final Word on the Subject

“Computers are stupid, they only do what you tell them to do.”

—Jeff Goldblum, The Fly

Like what you are seeing here at PurpleAmerica? Share and Comment. Let your friends and family know! Repost on Twitter and Threads. Forward us topics you’d like us to discuss! It’s an open community here.

Footnotes and Fun Stuff

1. I am still surprised how many Americans get this question wrong, demonstrating even further, why computers are better than humans.

2. Terminator movies

3. Isaac Asimov book

4. This was the issue when a lawyer used ChatGPT to write a legal brief that ChatGPT cited completely bogus cases.

5. In actuality, a game called “Go” is far more complex and required far longer for computers to master. Nonetheless, computers now regularly beat humans at that too.