Don’t Cede Your Intelligence

Chances are you’ve seen examples of Artificial Intelligence (AI) or have worked with it directly. It is an amazing addition to our increasingly technological world.

AI can accomplish stunning feats at blazing speed. That makes sense, since its infrastructure is a computer. Computers accomplish so many feats at lightning speed, especially compared to the time it would take a human to accomplish the same tasks.

AI is artificial in the sense that it is built by humans and exists within computers. But is it intelligent? I don’t believe it is. Any semblance of intelligence is because we attribute intelligence to things that mimic us. By calling AI “intelligent,” we are ceding our intelligence to an object.

A fanciful thought experiment

Imagine some people sitting in a field thousands of years ago. One of them picks up a wooden stick and starts etching symbols in the mud. Each symbol represents an item from their world. The symbols look something like this:

||
||

Then they use the stick to figure out what it would look like if they combined all the symbols, and they get this new symbol:

||||

This is very close to how the ancient Sumerians developed the mathematical notation that they used in turn to develop multiplication, division, and geometry.

Now imagine that the Sumerians attributed all their intelligence to the stick. Imagine an ancient Sumerian saying “Wow. Look at how amazing this stick is! It can calculate symbols so quickly. And even though the stick sometimes produces ||| or |||||, we’ll just ignore the occasional errors. This stick is truly one of the earliest tools of artificial intelligence.”

All joking aside, it is clearly ridiculous to attribute their intelligence to the stick rather than to themselves. Before we attribute intelligence to objects, let’s explore what intelligence means.

Intelligence

In order for something to be intelligent in the human sense, I believe it would have to have these qualities:

• Understanding. Understanding is not just an ability to recall or process information from a repository; it involves a deep sense of the meaning of that information.

• ChatGPT and other LLMs (large language models) mimic understanding by calculating the probability that the pattern of words it provides closely matches the pattern of words in the prompt that is given, without actually understanding.

• Creativity. Creativity is not just a result of choosing from among things you already know or have access to; it’s about creating something new that isn’t part of what has come before, to go beyond or even contrary to what you have been trained to do.

• Artificial Intelligence is limited to its training and is not actually creating anything new. Since AI is also producing content that it is then further trained on, including errors, some are now warning of the danger to future information that will result from allowing AI to “eat its own tail”.

• Emotions. Many people would think logic and emotions are at odds with each other. However, as neuroscientist Antonio Damasio demonstrates in Descartes’ Error, logic requires emotion. He points to medical cases where people experience severe injuries to the emotion centers of their brains and are subsequently unable to make even basic logical decisions.

• Artificial Intelligence does not feel any emotions, and its “decisions” are a result of calculating probabilities.

• Experience. Our nervous systems are comprised of more than just the thoughts that take place in our brain: there are biological connections between our brains and our senses, muscles, and even our microbiome; and there are social connections that we make with people who are different from us. As we grow, all of these combine to play a role in what we call intelligence, making much of what we call intelligence both experiential and nonverbal.

• Artificial Intelligence lacks all these forms of experience.

A conversation with ChatGPT

In order to pursue these questions more fully, I turned to the source to get its input. The following excerpts are quotes from ChatGPT in response to some of my questions about it. The depth of its answers shows off the impressive nature of what it is able to do. At the same time, ChatGPT is clear about its own limitations.

“I don’t have memory or understanding the way you do. Instead, I generate responses based on patterns in the data I was trained on.”

“… I can often generate answers that feel like knowledge, because I’ve been trained on so much material.”

“… what I do is based on patterns of reasoning found in texts I’ve been trained on — not on understanding or self-awareness in the human sense. … I don’t have beliefs, consciousness, or independent judgment. I generate reasoning the way a calculator does math: very fast, very structured, but not self-reflective.”

“[I’m] capable of generating poetry, stories, ideas — but based on recombinations of existing patterns.”

“Since I don’t have a body, emotions, or sensory input, I simulate these experiences based on text — but I may misinterpret metaphorical physical language (e.g., ‘gut feeling’), miss the emotional weight behind a lived experience (e.g., grief, trauma), or get tone wrong in emotionally loaded conversations (especially when there’s cultural variability).”

Conclusion