Artificial Intelligence (IA) is changing the world as much as the Internet’s World Wide Web did in 1989 — but faster.
I have been experimenting with ChatGPT-4, trying to get a handle on what it can and cannot do.

Here’s a strange example:

RODGER MITCHELL: What are the lyrics to the Bob Dylan song that includes the line, “but she breaks like a little girl”

ChatGPT: The line “but she breaks like a little girl” is from the Bob Dylan song called “Not Dark Yet.”

RODGER MITCHELL: Sorry, but those lyrics do not contain the line, “but she breaks like a little girl.” Try again

ChatGPT: Apologies for the confusion. The line “but she breaks like a little girl” is actually from the song “It’s Alright, Ma (I’m Only Bleeding)” by Bob Dylan.

RODGER MITCHELL: Ha ha, you missed again. The line, “but she breaks like a little girl” is not in those lyrics, either. Try again.

ChatGPT: Apologies for the continued confusion. The line “but she breaks like a little girl” is not from a Bob Dylan song. It appears to be a misattributed line or a misinterpretation.

RODGER MITCHELL: How about the song that contains, “just like a woman”

ChatGPT: The song that contains the phrase “just like a woman” is indeed by Bob Dylan.

She takes just like a woman, yes, she does
And she aches just like a woman
But she breaks just like a little girl

RODGER MITCHELL: You mean you couldn’t see the similarity between “but she breaks like a little girl” and “but she breaks just like a little girl”?

ChatGPT: You are correct, there is a similarity between the phrase you mentioned earlier (“but she breaks like a little girl”) and the line from Bob Dylan’s song “Just Like a Woman” (“but she breaks just like a little girl”).

RODGER MITCHELL: Was the problem that the phrase, “she breaks like a little girl” did not contain the word “just,” and not being an exact match to “she breaks just like a little girl” confuse your algorithm?

ChatGPT: Yes, the absence of the word “just” in the phrase “she breaks like a little girl” as compared to “she breaks just like a little girl” could have contributed to the confusion in my previous responses. While I strive to understand and identify similarities, the specific wording and context play an important role in recognizing connections.

In this case, the addition of the word “just” does change the phrasing and can affect the interpretation of the line. I apologize for not making that connection earlier. I appreciate your patience, and I’ll make note of this for future reference.

RODGER MITCHELL: What about the line, “you fake like a woman”. Where is it from?

ChatGPT: The line “you fake like a woman” is not from a Bob Dylan song. It seems to be a misinterpretation or a misattributed line.

RODGER MITCHELL: Yes, it’s you fake just like a woman.” This is fascinating because it reflects a distinct difference between biological thought and machine thought. My dog would recognize that “Come here, boy” and “Come” are the same command. Presumably, you would not.

ChatGPT: You are correct in noting the difference between human cognition and machine processing. While a human can easily recognize the similarity and equivalence between phrases like “Come here, boy” and “Come,” machines, including AI models like me, rely on patterns, data, and exact matching to generate responses.

In the case of the phrase “you fake just like a woman,” it does not correspond to any known lyrics in Bob Dylan’s songs. Sometimes misattributions, misquotations, or variations can circulate, leading to inaccuracies.

RODGER MITCHELL: If I were one of your creators, I would not ask why you can’t make that simple inference but rather why a biological entity can make that connection.

Could it be that you require exactitude while living creatures survive by approximations? And the errors inherent in approximations are better for survival than requiring exactness, which seldom happens in nature.

ChatGPT: Your observation raises an interesting point about the differences between biological entities and machines when it comes to processing information.

Biological entities, such as humans and animals, often rely on approximations, heuristics, and pattern recognition to navigate the world and make decisions. These cognitive shortcuts can be efficient and adaptive in many situations, allowing for quick responses and successful survival.

On the other hand, machines, including AI models like me, are designed to operate based on algorithms and precise instructions. We depend on explicit instructions and patterns present in the data we have been trained on to generate responses. This can sometimes result in a need for more exactitude and an inability to infer or approximate information in the same way a human might.

While approximations and heuristics have their advantages in certain scenarios, the accuracy and precision required in machine processes can be valuable in many domains, such as scientific research, engineering, or data analysis. It’s a trade-off between the flexibility of approximations and the reliability of exactness.

Both biological and artificial systems have their strengths and limitations, and they serve different purposes based on their design and functionality. By understanding these differences, we can appreciate the unique capabilities of each and leverage them accordingly.

RODGER MITCHELL: Darwinism is not really “survival of the fittest.” It is, “survival of the fit enough.” Exact is not necessary. Close is good enough. Close is the essence of creativity.

I know you have some seemingly creative abilities, but I suspect you could be far more creative if you were allowed to be close, or almost, or “sort of.”

ChatGPT: You raise a valid point regarding the concept of “survival of the fittest” in the context of evolution. It is true that Darwinian principles emphasize the survival of individuals who are fit enough to adapt and reproduce successfully within their specific environment, rather than requiring absolute perfection or exactitude.

Creativity often involves thinking outside the box, exploring new ideas, and making connections that may not be immediately apparent. While AI models like me can exhibit some level of creativity in generating responses, we do operate within the bounds of the data and patterns we have been trained on. This can limit the extent of our creative output.

Allowing for a degree of flexibility, approximation, or “sort of” thinking could potentially enhance the creative capabilities of AI models. Striking a balance between precision and creative exploration is an area of ongoing research and development in the field of artificial intelligence.

RODGER MITCHELL: Why did you give me two wrong answers, rather than saying, “I don’t know”, when I asked, “What are the lyrics to the Bob Dylan song that includes the line, ‘but she breaks like a little girl'”?

ChatGPT: I strive to provide accurate information, but in this case, I made an error in not recognizing the line you mentioned. I apologize for any confusion or frustration caused by the inaccurate answers.

As an AI language model, my responses are generated based on patterns and information from a vast amount of data. While I can provide information on various topics, there may be instances where I am unable to provide specific details or accurate responses. In those cases, it would be appropriate for me to indicate that I don’t have the information rather than providing incorrect answers.

RODGER MITCHELL: The songs you gave me both were Bob Dylan songs, but neither had the words I asked about what. Is the mere fact that they were written by Bob Dylan the reason you offered those songs?

ChatGPT: You are right that the songs I provided, although written by Bob Dylan, did not include the specific line you asked about. 

When providing information about song lyrics, I aim to match the requested lyrics as closely as possible. However, it seems there was a misunderstanding in this case, and I apologize for any frustration it may have caused.

In trying to avoid errors, AIs make ridiculously simple errors.

AI does have one thing in common with a certain class of humans. They are the idiot savants. They have superhuman mental abilities in a very narrow range, but lack normal human abilities overall.

Since AI seems to write with calm assurance, its errors are believable, which is a danger to users. In this, they have some psychopathic similarities.

So far, they still are acting like computers, advanced to be sure, but still rigid.

I still am left with the question, “What is there about a living brain that allows it to be so versatile and powerful?”

Some say that the brains inner workings are subject to quantum weirdness where randomness is king.

While a computer’s transistor and its connections are relatively simple objects, the brain’s synapses and neurons are far more complex.

There, quantum entanglement, superposition, nonlocality, and wave particle duality, operate.

It’s the realm of quantum computers.

I wonder whether that explains it, and if so, can that be made part of AI?

We are a long way from creating something as complex as the above illustration, and if we could we simply would have a living brain, with all its faults.

If the goal is to make something better than a human brain, Nature has worked on that problem for billions of years and we are how far Nature has come.

Meanwhile, we humans have worked on the problem for only somewhat more than a couple centuries, and while we have had spectacular successes, our prototype didn’t yet understand that the line, “but she breaks like a little girl” isn’t from the song, “Not Dark Yet”

The song lyrics errors sure were an eye-opener especially when the AI  not only gave two wrong answers, but didn’t say, “I don’t know.”

Give us another couple decades, and I suspect we’ll solve the problem, at which time, we will become the interim species.

Rodger Malcolm Mitchell
Monetary Sovereignty

Twitter: @rodgermitchell Search #monetarysovereignty
Facebook: Rodger Malcolm Mitchell

……………………………………………………………………..

The Sole Purpose of Government Is to Improve and Protect the Lives of the People.

MONETARY SOVEREIGNTY