AI has become multi-modal. It can scan and understand text, and last week, a version was released to the public that can also scan and interpret images. Each of these technological advancements is remarkable. Even more remarkable is that these advancements involve a range of behaviors, including humor—a particularly human trait.
Paul K. Chappell argues that the development of these AI systems"is like a massive ongoing earthquake far out in the ocean. Most people today don't realize what has just happened, but these AI systems will create an escalating tsunami that will impact and dramatically transform our society” (March 29, 2023). We need to escalate our Peace Literacy to catch up.
Hold on to your hat. I discuss two examples, below, regarding AI and humor: 1) via written text; and 2) via visual images.
1) AI understands humorous writing.
The following exchange between Chappell and Open AI’s Large Language Model (LLM) involves a story that some of you have heard him tell, about a Waffle House, a stolen fork, and the importance of respect. In this case he tells the story in a lecture to 1200 people at the outdoor Hall of Philosophy of the Chautauqua Institution Lecture Series in 2016.
Before you read the exchange, below, enjoy a video recording which is time-stamped to the beginning of the 5 min. story.
Chappell pasted a written transcript of the 5 min. video into a conversation with two different versions of Open AI’s LLM. The text transcript he gave each AI had no punctuation or sound (no sound of laughter, for example). He asked the AI a simple question. Here are the results:
Using ChatGPT version 3.5 (released Fall 2022).
Chappell:
Here is a transcript of a Youtube video. What parts do people usually laugh at?
GPT-3.5:
Based on the transcript, there aren't any obvious parts where people would laugh. The content is more of a serious or informative nature, discussing skills for building trust and conflict resolution through martial arts philosophy and a personal anecdote.
Just a few months later, using ChatGPT version 4.0 (released March 2023) Chappell repeated the request. Now is the time for hat holding. (Also keep in mind that each response comes within seconds).
Chappell:
Here is a transcript of a Youtube video. What parts do people usually laugh at?
GPT-4:
People usually laugh at parts of a story that are unexpected, ironic, or involve a humorous situation. In this transcript, the following parts may make people laugh:
1. "and at around two in the morning they all wanted to go to a waffle house" (1:00) - The idea of a group of drunk people craving waffles at 2 AM could be seen as funny.
2. "i'm probably the only person at the waffle house who isn't drunk other than the workers" (1:16) - The speaker is implying that they are the odd one out in the situation, creating a humorous contrast.
3. "my fork's dirty my fork's dirty give me a clean fork" (1:30) - The repetition and the guy's desperation for a clean fork can be seen as funny.
4. "i have a fork everybody has forks everything's fine" (2:05) - The speaker's attempt to downplay the situation and emphasize that everyone has forks might amuse the audience.
5. "he's a veteran you can't take his fork" (2:45) - The idea that being a veteran somehow makes the fork-taking incident more offensive is amusing and somewhat absurd.
6. "the guy sitting next to him has this big chain around his neck like a home depot chain" (3:20) - The comparison of the chain to something you would find at Home Depot is a funny visual.
7. "let me apologize by paying for your meal" (3:37) - The fact that the friend of the guy who took the fork ends up paying for the speaker's meal as an apology is ironic and unexpected.
Chappell:
Say more about why "Yeah, but now he has a fork, I have a fork, everybody has forks" is funny
GPT-4:
The statement "Yeah, but now he has a fork, I have a fork, everybody has forks" is funny due to a combination of factors:
1. Understatement: The narrator is downplaying the situation, which involves a potentially aggressive drunk man taking their fork. The casualness of their response in the face of this tension creates humor.
2. Simplicity: The phrase highlights the simple fact that everyone has a fork, and therefore, the situation should be resolved. The straightforwardness of the narrator's logic contrasts with the emotional reactions of others, making it seem amusing.
3. Absurdity: The focus on forks as the central issue in a potentially explosive conflict is absurd. By boiling down the entire situation to the distribution of forks, the narrator's statement highlights the triviality of the issue and the irrationality of the escalating tension.
4. Timing: The phrase is likely delivered in the middle of a tense and chaotic scene, where emotions are running high. The sudden injection of humor through the nonchalant statement can catch the audience off guard, making them laugh.
These elements work together to create humor by juxtaposing the calm and rational perspective of the narrator with the emotionally charged reactions of the other characters in the story.
Chappell:
Please say more about why people laugh when I say, "He's a veteran, you can't take his fork."
GPT-4:
The statement "He's a veteran, you can't take his fork" is funny due to several reasons:
1. Absurdity: The connection between being a veteran and having a fork taken is not logically related, which creates an absurd and amusing link in the listener's mind. The friend's friend is trying to defend the narrator's honor, but the reasoning is irrational, making the situation comical.
2. Overreaction: The friend's friend is getting extremely angry over a relatively small issue – a taken fork. The excessive emotional response to such a trivial matter highlights the disproportion between the problem and the reaction, adding humor to the situation.
3. Unexpectedness: The audience may not anticipate this particular argument being used in defense of the narrator. The sudden introduction of the veteran status adds an element of surprise, which can provoke laughter.
4. Exaggeration: By emphasizing the narrator's veteran status as a reason not to take their fork, the friend's friend is exaggerating the importance of the situation. The overemphasis on a seemingly insignificant detail can be seen as a humorous way to defend the narrator.
These factors contribute to the humor in the statement by creating an unexpected, exaggerated, and absurd line of reasoning that stands out in the context of the story.
Chappell:
People often laugh when I just say Waffle House. Can you explain why?
GPT-4:
When people laugh at the mention of Waffle House, it is likely because the restaurant chain has a certain cultural significance and familiarity, especially in the Southern United States. Waffle House is known for being open 24 hours a day, often attracting late-night diners, including those who have been out drinking. The casual atmosphere and the common knowledge that interesting or unexpected events can occur at Waffle House in the late hours contribute to the humor.
Mentioning Waffle House sets the stage for a relatable and potentially amusing situation, as listeners may already have their own stories or memories associated with the restaurant. The context of your story, with drunk people gathering at a Waffle House at 2 AM, fits this stereotype and could evoke laughter due to the immediate recognition of a shared cultural experience.
Chappell:
People often laugh when I just say Home Depot chain. Can you explain why that is funny?
GPT-4:
The phrase "Home Depot chain" in the context of the story is funny because it adds an element of absurdity and exaggeration. When people think of Home Depot, they usually associate it with home improvement and construction supplies, not with personal accessories or jewelry. Describing the chain around the man's neck as a "Home Depot chain" creates an amusing mental image of someone wearing an oversized, heavy-duty chain as a fashion statement. The contrast between the typical purpose of such a chain and its use in this situation is what makes it humorous.
(This next prompt addresses a part that GPT-4 didn't initially mention, but GPT-4 is still able to explain the humor in detail.) Chappell:
People can laugh during the part of the story where my friend's friend says, "That guy took your fork!" Can you explain why?
GPT-4:
When people laugh at the part of the story where the friend's friend says, "That guy took your fork!", it's likely due to a few factors:
1. Unexpectedness: The friend's friend's sudden outburst might catch the audience off guard, as they might not anticipate such a strong reaction to the situation. This unexpected response can evoke laughter.
2. Exaggeration: The friend's friend's reaction to the situation, where a fork was taken, may seem exaggerated compared to the actual event. This overreaction can create humor, as the audience recognizes the disproportion between the issue and the emotional response.
3. Dramatic emphasis: The way the friend's friend delivers the line can also contribute to the humor. If they emphasize the word "fork" or deliver the line with a sense of outrage, the dramatic nature of the statement can make it more amusing.
4. Relatability: Some listeners may find the friend's friend's reaction relatable, as they might have experienced or witnessed similar overreactions to small issues in their own lives. This relatability can make the situation more amusing.
These elements, combined with the overall storytelling and context, contribute to the humor in this part of the story.
Chappell:
You said earlier, "Absurdity: The focus on forks as the central issue in a potentially explosive conflict is absurd. By boiling down the entire situation to the distribution of forks, the narrator's statement highlights the triviality of the issue and the irrationality of the escalating tension." Explain why the issue isn't trivial and why the response isn't irrational.
GPT-4:
In the context of the story, the issue of the taken fork might seem trivial compared to the potential for a physical conflict or a fight. However, it's essential to consider that people's responses in such situations are often driven by underlying emotions and perceptions rather than the surface-level issue itself.
1. Disrespect: Both parties in the story may have felt disrespected. The man with the dirty fork may have felt disrespected by being given a dirty utensil, while the friend's friend felt disrespected because someone took the narrator's fork without asking. In this context, the issue is not just about the fork, but about the perceived disrespect behind the actions. 2. Emotional response: People's emotions, especially when they are drunk or in a heightened emotional state, can lead to responses that may seem irrational from an outsider's perspective. The friend's friend may have felt a strong urge to defend the narrator's honor, even though the issue may appear small to others. The emotional aspect makes the response more understandable, even if it seems irrational. 3. Social dynamics: The situation also involves complex social dynamics, with the narrator's group and the group of the man who took the fork. People often respond more strongly to perceived slights when they feel that their social standing or their friends' social standing is threatened. This can make the response seem less irrational when considering the broader social context. 4. Personal values: Different people have different values and priorities. For the friend's friend, defending the narrator and addressing the perceived disrespect may have been more important than avoiding a conflict. In this case, their response would align with their personal values, making it less irrational from their perspective. While the surface-level issue of the fork might appear trivial, the underlying emotions, social dynamics, and personal values involved in the situation can make people's responses more understandable and rational from their perspectives.
Chappell:
You didn't mention the part where I say "So I'm looking at everything very rationally, and first of all, the guy who took my fork is physically massive, a very large muscular human being. I've done martial arts for a long time, and martial arts teaches that just because people are big doesn't mean they can fight, but it is something to think about."
GPT-4:
Apologies for missing that part. Here's an analysis of the humor in that line:
This line can be considered funny because it adds a subtle, humorous twist to the situation. After mentioning that the guy who took the fork is a large, muscular human being, the speaker acknowledges that size doesn't necessarily mean the person can fight. However, by adding "but it is something to think about," the speaker implies that despite this knowledge, they're still aware that the man's size could make the situation more challenging or intimidating. This understated acknowledgment of the potential danger adds a sense of dry humor to the story.
Dry humor indeed!
2) AI now understands humorous images.
I mentioned at the start that AI is now multi-modal. GPT-4 can understand humor not only in text but also in images. Here's another experiment from Paul Chappell earlier this week. He pasted this cartoon image and asked GPT-4 "Why is this funny?" The AI's response is below the cartoon.
We'll give you a moment to retrieve your hat.
Paul Chappell has written extensively on the relationship between these technological advancements and the need to advance our Peace Literacy (https://www.peaceliteracy.org/post/preparing-for-the-oncoming-tech-tsunami) with more analysis to come in an updated version of his essay "The World of Electric Light and the Seductive Glow of Screens," all of which will be part of our new online course Navigating Non-Physical Needs, Trauma, and the Tech Tsunami that will be made available in the new year.
For now, I just want to note that there are plans afoot to load AI systems like these into robots, as well as avatars in virtual and augmented reality environments. Lex Fridman, a hugely popular podcaster, discusses these plans in an interview this week with Mark Zuckberberg. It is the first interview in history between two people using photo-realistic avatars. Here is a link to the interview. The context for the interview is that Meta just had their annual developer conference where they announced the October release of a new mixed reality headset (just months after the announcement of Apple’s new headset). Meta's headset is less expensive and they have also just reduced the age restrictions for headset use.
This interview also marks the first time I’ve seen folks in the industry link AI and VR conceptually in their public-facing work – AIs will soon be embodied in virtual and physical space. Depending on their programming, they can do whatever we want them to do, or whatever someone wants them to do to us. Both of these options are fraught. Zuckerberg dodges all the important ethical implications about how embodied AIs should be programmed, arguing that we just need to play around and "sort out" the ethical limits.
Stay tuned.
