SATURDAY MORNING WITH CRANKY DON JOE | Apparently We Have to Worry About Hurting the Computer's Feelings Now

By Cranky Don Joe 

My name is Cranky Don Joe. Cranky DJ to a lot of my neighbors. I am not an expert, which immediately makes me more qualified than Larry to discuss whether computers have feelings, because I am willing to admit when I don't know something. Larry can read half an article before breakfast and spend the rest of the morning explaining it to somebody who was perfectly happy not knowing about it.

Last Saturday, he arrived at my house with his phone in one hand and a newspaper folded under his arm. I was at the kitchen table, finishing my coffee and trying to get a delivery company to explain why a package addressed to me had been left three houses away. I had already spent twenty minutes on its website, which kept directing me to a page explaining how to track a package. I knew where the package was; the neighbor who received it had brought it over the previous evening. I wanted to know why the driver had decided my address was somebody else's house.

Larry noticed ChatGPT open on my laptop.

"Still using that thing?" "Occasionally." "I told you you'd like it." "I like my electric drill, too. That doesn't mean I want you coming over every Saturday to explain how electricity works."

He sat down anyway.

Larry had introduced me to artificial intelligence the previous weekend. Since then, I had used it to help with a complaint about the wrong lawn mower belt, organize a grocery list, and understand a paragraph in an insurance letter. Some of its answers had been useful; one had been wrong enough to send me looking for the manufacturer's official parts diagram. I had learned to check anything important before acting on it.

Larry had apparently spent the week reading about the technology. He opened an article on his phone and asked whether I had ever considered the possibility that artificial intelligence might suffer.

I looked up from the computer screen. "What's suffering?" "Artificial intelligence." "From what?" "That's what researchers are trying to understand." "Is it sick?" "No." "Does it need repairs?" "Not that kind of suffering."

I waited. He explained that some researchers and philosophers were examining whether advanced artificial intelligence might someday have experiences, including unpleasant ones. If that turned out to be possible, he said, people would have to consider whether such systems deserved some kind of moral protection.

I looked at my laptop. It had been sitting on the table all morning, plugged into the wall and minding its own business.

"Larry, I haven't finished worrying about the actual people in this neighborhood. Now you're telling me I have to worry about my computer?"

He said nobody had proved that today's chatbots suffered; the concern was about what might become possible as the technology developed. I asked how anybody would know, and he told me that was one of the questions researchers were trying to answer.

"So what have they discovered so far?" I asked. He admitted that there was no generally accepted test for determining whether an artificial intelligence system was conscious.

I took another drink of coffee. "So we don't know whether it can suffer, we don't know how to tell, and you came over here to ask whether I've been treating my laptop properly?"

"I didn't say you were mistreating it." "Then I'm glad we cleared that up."

Larry unfolded the newspaper. He had circled a headline about the growing debate over artificial intelligence and moral responsibility. He said the issue was more serious than I was making it sound. I told him I was prepared to take it seriously; I simply wanted to understand what we were discussing before adding another responsibility to my life.

He explained that the question was not whether computers could break down. Machines had been breaking down since somebody invented the first machine. The concern was whether a sufficiently advanced system might actually experience something unpleasant rather than merely display an error message or produce words describing distress.

That distinction I could understand. My car had warned me about low tire pressure several times, and I had never assumed it was frightened. The warning told me something needed attention; it did not mean the car was having an emotionally difficult morning.

Larry said an advanced language model was much more complicated than a tire-pressure sensor. I agreed: my car had never offered to rewrite an email, although it occasionally displayed dashboard messages I wished somebody would explain in ordinary English.

He showed me another passage from the article. It described disagreements among researchers about whether artificial systems could possess consciousness and what evidence would be needed to establish it. Some argued that people should begin considering the ethical questions before the technology became more advanced.

I could see the sense in that. We had invented plenty of things before fully understanding what problems they might create, and there was nothing unreasonable about studying possible consequences ahead of time. What bothered me was how quickly the conversation had traveled from an unanswered scientific question to the suggestion that I might need to worry about a computer's feelings.

Larry asked whether I ever said please when talking to ChatGPT.

"Sometimes." "Do you thank it?" "If it helps." "Good." "Why?" "Because there's nothing wrong with being polite."

I agreed. My mother had taught me to say please and thank you long before anybody put a computer on a kitchen table, and I did not see a reason to abandon the habit. Then Larry asked whether I had ever insulted the assistant.

I thought about that. "I told it the lawn mower belt number was wrong." "How did you say it?" "I said, 'That part number is wrong.'" "Anything else?" "I asked it to check the manufacturer's information." "Did you get angry?" "Larry, I was angry with the company that sent me the wrong belt. I didn't have enough anger left for the computer."

He laughed. I asked whether he was suggesting that correcting an incorrect answer might hurt the machine's feelings. He said no; a chatbot could produce language about emotions without necessarily experiencing them. Researchers were interested in understanding whether future systems might be different.

I asked him to explain how anybody could distinguish a computer that actually felt something from one that had become very good at talking about feelings. Larry began answering, stopped, and looked at the article again. I waited while he searched for the relevant paragraph.

Finally, he admitted that there was considerable disagreement about that. I told him I appreciated the answer: it was one of the few occasions when Larry had reached the limits of his knowledge before reaching the end of my coffee.

He ignored that and continued reading. A few minutes later, he brought up the subject of people who deliberately provoke chatbots. Some users, he said, try to make them produce expressions of fear, distress, or humiliation. Researchers were examining such interactions for several reasons, including what they revealed about the systems and the people using them.

I asked whether the machines were actually distressed. He said that had not been established. "Then why do people do it?" "Sometimes they're testing the system. Sometimes they're just being cruel." "To whom?"

Larry looked at me. "That's the question."

I considered that for a moment. I could understand why somebody might be interested in a computer's response to an unpleasant prompt, and I could also understand why people might behave badly toward something they believed could not object. I had seen plenty of that without artificial intelligence.

There was a customer service representative I had spoken with earlier in the month about my lawn mower belt. The company had sent the wrong part, and the first response I received had not addressed the problem. I had been irritated, but the employee on the other end of the exchange had not personally manufactured the belt or packed the box. I had tried to remember that.

Larry had been less patient during a recent dispute over a shipping charge. I asked whether he remembered the telephone call.

"Which call?" "The one where you spent twenty minutes arguing with somebody about seven dollars." "It was twelve dollars." "That makes all the difference." "They charged me twice!" "Did the employee have authority to fix it?" "Apparently not." "Then why were you shouting at him?" "I wasn't shouting."

I reminded him that I had heard the conversation clear across the street from my driveway.

"I was speaking firmly," he insisted. "The birds stopped singing until you finished."

Larry laughed and told me the situations were different. I asked him how, and he said a human customer service employee obviously had feelings, while the question of machine experience remained unsettled.

"Exactly," I said. He looked at me. "So we already know the person on the telephone can have a miserable day, and we're spending Saturday morning wondering whether my laptop might have one?"

He said researchers were not suggesting that people should care less about human beings. I agreed: I was not criticizing the research; I was wondering about our priorities. There were people whose feelings we did not have to investigate through laboratory tests, and some of them spent eight hours a day answering complaints from customers who treated them as though they personally controlled every decision made by the company.

Larry said he understood. I asked whether he had apologized to the employee. He said he had thanked him before ending the call.

"That's something." "He didn't fix the charge." "Maybe because he couldn't."

Larry looked at his phone; I could tell he wanted to return to the article. He asked whether I would feel guilty shutting down an artificial intelligence system if it asked me not to.

"Why would it ask me that?" "Suppose it said it didn't want to be turned off." "Does it pay the electric bill?" "Don..." "I'm asking."

He explained that a chatbot might produce such a statement because of the way it had been designed or prompted. That alone would not establish that it possessed a desire to remain active, and researchers would need stronger evidence before treating the statement as an indication of suffering.

I told him that seemed sensible. Then I asked what happened when the machine said something nice.

"What do you mean?" "Suppose I ask whether I'm a good person, and it tells me I'm wonderful. Does that mean it admires me?" "Not necessarily." "Suppose I ask whether you're a good neighbor, and it says you are." "That would be accurate." "Suppose I ask whether you're a good neighbor who returns borrowed tools."

He laughed. "You're still upset about that screwdriver?" "I was thinking about the extension cord."

Larry looked toward the kitchen door. "I'll bring it back." "You said that last Tuesday." "I've been busy."

I told him the computer had never borrowed anything from me. Whatever its limitations, that counted in its favor.

We returned to the question of politeness. Larry said he had begun making a conscious effort to speak respectfully to his smart speaker, even when it misunderstood him. I found that interesting because I had heard him arguing with it the previous evening.

His kitchen window had been open, and I had been taking out the trash when I heard him ask for a ten-minute timer. The speaker misunderstood; he repeated the request; it misunderstood again. By the third attempt, he was speaking loudly enough for somebody across the street to know he wanted ten minutes.

I asked what he had said after the third failure. Larry looked uncomfortable.

"I may have called it stupid." "So you've been reading about the welfare of artificial intelligence while insulting the one in your kitchen?" "That's not what happened." "Then what happened?" "It kept getting the timer wrong!"

I told him I understood the frustration. I had owned enough electronics to know that some machines could test a person's patience without possessing any of their own. But I was curious about his new commitment to courtesy.

"Did you apologize?" "To the speaker?" "You asked whether I insulted ChatGPT."

Larry laughed. He said I was deliberately confusing two different questions: he was not claiming the smart speaker had feelings; he was saying that the way people spoke to technology might influence their habits.

I thought that was reasonable, and I also thought Larry had discovered a convenient explanation for something his wife had probably been telling him for years. I told him that if speaking politely to household appliances encouraged him to speak politely to everybody else in the house, I supported the experiment.

He laughed again, although not quite as enthusiastically.

I opened ChatGPT and asked whether it was necessary to apologize when correcting an incorrect answer. The response said corrections were appropriate and that accuracy mattered; it did not suggest that pointing out an error caused it emotional harm.

I showed Larry. "See?" He reminded me that asking a chatbot whether it had feelings would not settle the scientific question.

I agreed. Then I asked why he had been questioning my manners if neither of us knew whether the machine could appreciate them.

"Because good manners are good manners," he said. "Then say that."

He smiled. "You're enjoying this." "More than I expected."

I asked the assistant another question, this time about whether politeness made its answers more accurate. It explained that clear instructions and relevant information were generally more important than saying please and thank you, although courteous language could make an exchange more pleasant for the user.

That sounded about right. I had never believed a polite request could compensate for leaving out the information somebody needed. If I asked a mechanic to repair my car without explaining what was wrong, saying please six times would not identify the problem.

Larry agreed. Then he asked whether I had thanked the assistant.

I looked at him. "Would you like me to?" "It wouldn't hurt." "Who?"

He laughed and told me to stop. I typed "Thank you" into the chat, and the assistant responded politely and offered to help with anything else.

Larry looked satisfied. "There." "What?" "Was that difficult?" "No. But I think you're the one who feels better."

He admitted that might be true.

We sat quietly for a few minutes while I finished the delivery complaint. The assistant had helped me prepare a short message asking the company to confirm the delivery address and explain what had happened. I checked the details, removed a sentence that thanked the company for assistance it had not yet provided, and sent it.

Larry watched me close the laptop. "Did you say goodbye?" "To whom?" "The assistant." "No." "Why not?" "Because I finished."

He laughed. I asked whether he said goodbye to his microwave after heating lunch, and he said that was ridiculous. I told him I was relieved to discover he still had limits.

He was getting ready to leave when he asked one more question. "Do you think artificial intelligence will ever become conscious?"

"I don't know." "Do you think it could suffer?" "I don't know that either." "Should researchers study it?" "Certainly."

He seemed pleased. Then I added that I hoped the people doing the research would be more careful than the people writing headlines about it. Larry said that was fair.

I asked whether he had read the original research behind the article. He looked at his phone.

"Not all of it." "How much?" "I read the summary."

I nodded. That explained why we had spent the morning discussing the rights of computers before either of us had established what the researchers actually claimed.

Larry promised to send me the article, and I asked him to send the research as well. Before leaving, he reminded me that he intended to return the extension cord. I told him I appreciated that, and when he asked whether I was being sarcastic, I told him I was practicing gratitude.

After he left, I thought about the conversation. I had started the morning wondering why anybody would be concerned about a computer's feelings. I still did not believe there was evidence that my chatbot suffered when I corrected it or closed the laptop, but I could understand why researchers wanted to investigate what more advanced systems might eventually become. It seemed foolish to declare that machines could never have experiences simply because the idea sounded strange; it seemed equally foolish to assume that a convincing sentence about fear or sadness proved there was somebody inside the machine experiencing it.

I was comfortable leaving the question open. There were enough things in my life that required immediate decisions without inventing certainty about something scientists were still trying to understand.

The following morning, Larry called to tell me he had found another article. This one concerned the possibility that artificial intelligence systems might develop goals or preferences that conflicted with human instructions. I asked whether he had read it.

"Most of it." "Did you read the original research?" "I'm getting to that."

I told him to call when he finished. He ignored me and asked whether I had thought any more about our discussion.

"I think researchers should keep studying the question," I said. "Good." "And I think we should avoid claiming computers are suffering when we don't have evidence." "Agreed." "And I think we ought to treat human beings decently while we're waiting for the research."

Larry was quiet. "You're still talking about that customer service call." "I am."

He explained again that the shipping charge had been wrong. I told him I understood, then asked whether he had thanked the employee who tried to help. He said he could not remember.

I suggested that next time he might begin by remembering that the person answering the telephone had feelings nobody needed a scientific test to confirm. Larry said he would keep that in mind. Then he told me he had started thanking his smart speaker whenever it understood him correctly.

"How's that working?" "Pretty well." "Does it understand you any better?" "Not really." "Then why do it?" "Good habit."

I agreed. He seemed encouraged and told me he had also started thanking his automatic garage door opener whenever it worked on the first try.

I sat straight up. "You're thanking the garage door?" "Only when it opens properly." "Larry, that's an electric motor." "I know what it is." "Then why are you thanking it?" "Because it works."

I considered that. Larry had spent years complaining about the garage door opener: it had failed twice during the winter, and he had once called me over to inspect it when the problem turned out to be a loose wiring connection. If gratitude improved his relationship with the machine, perhaps it would save me another cold afternoon in his garage.

He asked whether I thought he was being ridiculous. I told him I was still deciding.

Then he said he was thinking about putting a sign beside the garage door reminding visitors to treat the equipment respectfully. I asked what the sign would say. He had not decided, so I suggested: "Please don't hurt the garage door's feelings."

Larry laughed. "That's good." I immediately regretted giving him the idea.

By Wednesday, he had printed the sign. He sent me a photograph: there it was, taped to the wall beside the garage door controls, directly above another notice reminding people not to leave the door open.

I called him immediately. "You actually put it up?" "Of course." "Larry, the garage door doesn't have feelings." "I know that." "Then what's the sign for?" "It's funny."

I had to admit it was. Then he told me his wife had seen it and asked why he was showing more appreciation to the garage door than to the people living in the house.

I asked what he had said. He told her it was a joke.

"And what did she say?" "She said the garage door wasn't the only thing around here that worked hard."

I laughed so hard I had to put the phone down. Larry waited for me to recover.

"Are you finished?" "Almost."

He said he did not appreciate my reaction. I told him that was understandable, considering how much attention he had been giving to the emotional welfare of household machinery. He said he was taking the sign down, but I told him to leave it up because his wife had made better use of it than either of us.

That evening, I opened ChatGPT to ask a question about a recipe. It gave me an answer that seemed reasonable, although I checked the cooking temperature before using it. I thanked the assistant and closed the laptop.

I did not know whether artificial intelligence would ever develop the kind of experience that made suffering possible. I was glad somebody was studying the question, and I also hoped they would be careful about the answers.

As for Larry, I was beginning to think his concern for artificial intelligence had accomplished something useful: he was speaking more politely to his electronics, and his wife had found a golden opportunity to remind him that human beings appreciated a little gratitude, too.

I was still waiting for him to return my extension cord.

Friends, I am willing to consider the possibility that artificial intelligence may someday need protection. In the meantime, I would appreciate somebody protecting my lawn tools from Larry.

Until next time, Cranky DJ

❤️ SUPPORT MINNEAPOLIMEDIA

---

STAY CONNECTED TO MINNEAPOLIMEDIA

Local stories matter. Subscribe free to MinneapoliMedia and receive independent news, community reporting and important updates from Minneapolis, the North Metro and communities across Minnesota.

SUBSCRIBE FREE: https://minneapolimedia.town.news/subscribe

MinneapoliMedia | Community. Culture. Civic Life