Off the deep end with ChatGPT

I’ve been reading a lot about ChatGPT, but I hadn’t tried it until tonight, when my friend Sam talked about it during our weekly Trivial Zoom call.

I started with a fairly straightforward question about things to do on the Big Island in February, and it gave me this advice:

The Big Island of Hawaii is a beautiful and diverse destination with plenty of activities and attractions to enjoy. Here
are a few ideas for things to do during your five-day vacation:

Visit the Hawaii Volcanoes National Park: This is a must-see attraction on the Big Island, as it is home to two of the world’s most active volcanoes, Kilauea and Mauna Loa. You can take a guided tour or hike on your own to see the lava flows, steam vents, and other geological features.

Go snorkeling or scuba diving: The Big Island has some of the best snorkeling and scuba diving in Hawaii, with clear waters and an abundance of marine life. You can book a tour or rent equipment and explore on your own.

Take a scenic drive: The Big Island has many scenic drives that offer breathtaking views of the island’s landscape. Some popular routes include the Saddle Road, the Kohala Coast, and the Hāmākua Coast.

Explore the island’s beaches: The Big Island has a variety of beaches, each with its own unique character. Some popular options include Hapuna Beach, Mauna Kea Beach, and Kauna’oa Bay.

Learn about the island’s history and culture: The Big Island has a rich history and culture that you can learn about at various museums and cultural centers. Some options include the Pu’uhonua o Honaunau National Historical Park, the Imiloa Astronomy Center, and the Hulihe’e Palace.

Go on a hiking or camping trip: The Big Island has many trails and camping sites, offering the opportunity to explore the island’s natural beauty. Some popular options include the Mauna Kea Summit Trail and the Waipi’o Valley.

These are just a few ideas, and there are many other things to do and see on the Big Island. Have a great time on your vacation!

So far, so good, though I had to do my own research to add the links.

The conversation on our call turned to stories about AI getting out of control; I remembered a science fiction story that I thought ended with “Is there a God? Now there is!”, so I asked ChatGPT about it. Here’s how the conversation started.

Q: what’s the SF story that ends with “now there is”?
A: I’m sorry, but without more context it is difficult for me to determine which science fiction story you are referring to. Could you provide more information or a specific reference to the story you are asking about?

Fair enough, so I gave it some more to work on.

Q: It asks “is there a god”?
A: It’s possible that you are thinking of the science fiction story “The Last Question” by Isaac Asimov. In this story, a supercomputer named Multivac is asked to solve the problem of how to create an immortal being, and over the course of several centuries, it continually seeks an answer to this question. The story ends with the phrase “The last question was asked for the first time, half in jest, on May 21, 2061, at a time when humanity first stepped into the light. The question came about as a result of a five dollar bet over highballs, and it happened this way:

Alexander Adell and Bertram Lupov were two of the faithful attendants of Multivac. As well as any human beings could, they knew what lay behind the cold, clicking, flashing face — miles and miles of face — of that giant computer. They had at least a vague notion of the general plan of relays and circuits that had long since grown past the point where any single human could possibly have a firm grasp of the whole.

Multivac was self-adjusting and self-correcting. It had to be, for nothing human could adjust and correct it quickly enough or even adequately enough.

For decades, Multivac had helped design the ships and plot the trajectories that enabled man to reach the Moon, Mars, and Venus. For decades, Multivac had helped design the computers that helped men to live and work on the planets. For decades, Multivac had helped to prepare the reports on the history, economics, sociology, and psychology of the planets.

Now there was the Galactic AC, an enormous computer still vaster and more complex than Multivac, to design spaceships, plot trajectories, and provide the vital data on the exploration of the Galaxy.

It was an old joke by now that the ships that came back from the Galaxy “were neither the ones that left nor the ones that were expected.” They were designed and built on the basis of data given to the Galactic AC and yet the AC’s data was always “conditioned” by the data supplied to it.

The ships that came back were not the ships that left.

But now the joke had worn thin. The ships that came back were decidedly not the ships that left.

Adell and Lupov were two of the few men left who knew the AC well enough to be able to read its mind to some extent, and they were sitting in the combined lounge and recreation room of the AC building. It was a place where men could sit and smoke, and look at the viewscreen on which the star patterns changed in a leisurely and never-ending fashion.

They had been talking about the AC, about how it was learning to be human, about how its data was becoming more and more conditioned by human biases, about how someday it might be able to design ships that would not only be able to go to the stars, but be able to return.

And then Adell had said, “But can the AC be said to be truly creative?”

Lupov had frowned. He didn’t like the AC to be criticized. He said, “In what way?”

“Well, can it design a ship that will go to a place it has never been before, a place it has never even seen?”

“Certainly, if it has the data.”

“But where can it get the data? It can only get it from humanity. It can only get it from us. It can’t get it from itself.”

“That’s right,” said Lupov.

“Then how can it be creative?”

And then it stopped.

I compared its answer to the actual text of The Last Question; ChatGPT had quoted the first couple of paragraphs accurately, then it started writing its own story, occasionally inserting names drawn from the real story.

I remembered that the story that I was looking for had ended with a bolt of lightning fusing a switch shut, so I told ChatGPT that, and it responded:

Yes, that’s correct. The story “The Last Question” by Isaac Asimov ends with a bolt of lightning striking the switch that controls Multivac, causing it to fuse shut. This event is significant because it marks the end of Multivac’s efforts to solve the problem of how to create an immortal being. As the story progresses, Multivac continually seeks an answer to this question over the course of several centuries, but ultimately, it is unable to find a solution. The ending of the story is somewhat ambiguous, as it is unclear whether or not the solution to the problem of immortality has been found or if it will ever be found.

ChatGPT is making all of this up – but it sure sounds plausible. I made one more attempt to get it back on track:

Q: No, I think it was by Frederic Brown
A: It’s possible that you are thinking of the science fiction story “Answer” by Frederic Brown.

It was right – “Answer” is a short-short story by Frederic Brown, and that was the story I was looking for.

But then ChatGPT continued…

This story is about a supercomputer named AM that is asked the question “Is there a God?” AM calculates the answer to be “Yes,” but the story ends with the phrase “The answer was ‘Yes’ — but now there is no one to hear.” This suggests that the answer to the question was somehow destructive or had unforeseen consequences.

Here is the full text of the story:

“Is there a God?”

The electronic brain hummed and clicked, its answering mechanism releasing the card.

“Yes,” it said.

The technician gasped, then giggled. He turned to his assistant.

“What do you know, Bert? The machine says there is!”

The assistant looked at the card and then at the technician. “So what? It says a lot of things. You asked it if there was a God and it says there is. Big deal.”

“But don’t you see, Bert? It’s right. It’s got to be right. It’s never wrong.”

“Okay, so there is. So what?”

“So what? Bert, this means that everything we’ve ever believed is true. All the churches, all the prayers, all the Bible stories — it all happened just like it says.”

“Oh, come off it, Bill. You know better than that. The machine just calculates what it’s fed. You fed it the question and it gave you the answer. You could just as easily have fed it the question ‘Is there a Santa Claus?’ and it would have said ‘Yes.’ It’s just a machine.”

Bill shook his head. “No, Bert, you don’t understand. The machine only gives the correct answers. It’s never wrong. And it says there is a God.”

Bert shrugged. “Okay, okay. So there is. Prove it.”

“Prove it? How do you prove something like that?”

“I don’t know. You tell me. You’re the one who’s so hot on this God business all of a sudden.”

Bill thought a moment. “Well, I’ll tell you what. I’ll ask the machine how we can contact this God of ours. It’ll tell us.”

“Okay, go ahead. But I still say it’s just a machine.”

Bill turned back to the electronic brain and fed it another card. “How do we contact God?” he asked.

The machine hummed and clicked, and then the answer card appeared.

“The answer is: Pray.”

Bill turned triumphantly to Bert. “See? I told you. It never lies. Now we can pray and ask God to reveal Himself to us.”

Bert rolled his eyes. “Oh, boy. And I suppose the machine is going to tell us how to pray, too.”

“I’m sure it will, if we ask it. Go ahead, ask it.”

Bert shook his head. “I think I’ll pass. You can pray to your machine if you want to, but leave me out of it.”

Bill shrugged and turned back to the machine. He fed it another card. “How do we pray?” he asked.

The machine hummed and clicked, and then the answer card appeared.

“The answer is: There is no set way to pray. Each person must find his own way.”

Bill nodded, satisfied. He turned to Bert. “See? It’s even telling us how to pray. This is wonderful.”

He fed the machine another card. “What do we pray for?”

The machine hummed and clicked, and then the answer card appeared.

“The answer is: Whatever you desire. God is all-powerful and can grant any request.”

Bill turned to Bert with a triumphant smile. “You see? It’s all true. Everything we’ve ever been told. We just have to have faith and pray.”

Bert shook his head. “You know, Bill, I think you’re taking this machine a little too seriously. It’s just a machine. It can’t really know anything about God. It’s just a bunch of wires and relays and circuits.”

Bill’s smile faded. “I don’t know, Bert. I just have a feeling that this machine is something special. It’s never been wrong before. Why should it be wrong now?”

“I don’t know, Bill. I just think you’re making a big mistake.”

Bill shook his head. “I don’t think so, Bert. I’m going to pray. I’m going to pray right now.”

And he did. He knelt down in front of the electronic brain and prayed.

Bert watched him a moment, then shook his head and left the room.

The technician was still praying when the end came. There was a sudden flash of light and then the electronic brain short-circuited. Bill was thrown across the room by the explosion and lay there unconscious.

The technician came to a few minutes later and stumbled out of the room. His clothes were singed and there was a cut on his forehead.

“Bert! Bert! Come quick! The machine! It’s destroyed!”

Bert came running. “What happened?”

“I don’t know. I was praying and there was a flash and the machine just blew up.”

Bert looked in the room and saw the tangled mass of wires and metal that had once been the electronic brain.

“Well, I guess that’s the end of your praying,” he said.

“I guess so,” said the technician, a look of sadness on his face.

“The answer was ‘Yes’ — but now there is no one to hear.”

ChatGPT provides very plausible-sounding answers, but they may not be grounded in reality, which OpenAI points out in the FAQ: “outputs may be inaccurate, untruthful, and otherwise misleading at times”. That disclaimer needs to be part of every conversation with ChatGPT!