Here Is The Choice That Generative AI Made When Trying To Settle That Online Contentious Debate About Whether To Prefer Encountering A Bear Or A Man When Being Alone In The Woods

In today’s column, I’d like to explore a recent viral TikTok video and social media controversy by leveraging the capabilities of generative AI. Doing so is intended to reveal important facets underlying the societal ramifications of the hot topic at hand, along with garnering insights about how generative AI works.

Let’s start with a pointed question that might seem a bit odd, but please give it a moment’s thought, thanks.

If you were alone in the woods, would you prefer to encounter a bear or a man?

Well, that’s quite a question. Indeed, it is the crux of the viral TikTok frenzy. The reactions online have stoked quite a fury. How so? By and large, the question is intended for women, and the responses seemed to be that they would prefer encountering a bear over encountering a man if they were otherwise alone in a wooded area.

An assumption that seems to be broken into tatters is that one would normally prefer to encounter a man rather than a bear. The bear can rip you to shreds. A man could presumably be a fellow companion while surviving in the woods. There are tons of everyday news stories about bears that seem to arbitrarily attack people and scarily come out of nowhere to do so. It would seem that encountering a bear has got to be a lot worse than encountering a man.

Why in the world are women apparently choosing to encounter a bear rather than a man?

The answer most often seemingly proffered is that a man is potentially much more dangerous than a bear. A man might be armed with a deadly weapon. A man might have evil intentions. A man might try to trick you into believing they are helpful and then unexpectedly turn against you. All in all, the cunning and deceptive nature of a man is construed as far worse than the instinctual reactive unthinking nature of a bear.

Responses to the responses have amped the topic into high gear.

The mainstream media has jumped into the fray, opting to cover the topic with emboldened headlines. People heatedly discuss the matter due to the classic dog-bites-man becoming man-bites-dog twist involved. We are to assume that the proper answer should be to encounter a man, and yet the social media response appears to be that the bear is preferred over encountering a man.

It is hard to know how long this stridently stoked controversy will last. A lot of these reactive social media circumstances are fads that come and go. All that will need to occur is some other divisive pointed question to arise and end up grabbing online attention. Voila, the bear versus man issue will fall by the wayside. The controversy will fade from memory though last forever on the Internet for those who want to remember what was “hot” in the minds of society in 2024.

Let’s unpack the matter while there is still a last vestige of fad remaining.

Mentally Traversing The Bear Versus Man Choice

One place to start assessing the topic would entail identifying various precautionary considerations.

• Do we know that indeed the responses of choosing the bear over the man are being provided by women?

As they say about the Internet, you can be just about anybody or anything you want to be while trolling online. Responses to online posts are not an ironclad indicator of what a segment of society is necessarily opining on.

• Do we know that even if women are responding to the post, they are a scientifically randomly selected sample that is generally representative of women across the board?

No, we don’t know that these respondents are statistically representative per se. An argument can be made that these respondents are potentially statistically non-generalizable. The most we might say is that they presumably are those who perchance follow social media and respond on social media. As far as we know, the respondents are not part of some carefully devised scientifically systematic study.

Keep in mind too that some of the posted responses might be posted via canned bots and not by the direct hands of humans.

Moving on, there are a plethora of additional considerations.

• Do we know that people responding are doing so with a semblance of sincerity or might they be doing so lightheartedly?

Again, this is something we cannot say for sure in the sense that certainly some people might be taking the question with great seriousness and answering from the bottom of their hearts, while some responses might have other motives in mind.

Imagine these possibilities of what is going on in the minds of those who are responding to the query:

• Genuine and entirely mindful.
• Genuine but with minimal thought or reacting off the cuff.
• Playful and intentionally not giving due consideration.
• Writes it off as a zany thought experiment and responds likewise.
• Devilishly want to prod others and stir the pot.
• Aim to get views no matter how attained.
• Relishes joking around and keeping the joke going.
• Contributes because sees this as a vaunted portal into existing culture.
• Emotionally charged up about the topic and angrily responds.
• Etc.

The gist is that we have to be cautious in assuming that the responses are bona fide in the sense that the person is seeking to mindfully add to a weighty discussion. The odds are that some or perhaps a lot of the responses have other ulterior motives in play.

Perhaps we can acknowledge that the choice of the bear is somewhat anecdotal and has launched the matter into the viral stratosphere by luck of the draw. If the initial prevalent response was the man, we probably would not ever have heard about the question. Choosing the bear has managed to tap into contemporary societal angst underpinning the relationship between men and women.

A sign of our times, one might sigh and lament.

Some men are irked at the idea that the bear would be chosen over a man. What the heck, comes the reaction. Are you really willing to toss out all the plus sides of encountering a man versus a bear? Is that what times have come to? This then further fuels responses to the question.

Some men agree with the bear being chosen over a man. You are right, they proclaim, that the bear is the better choice. Others might say the same but do so with a different perspective, such as, go ahead and pick the bear, watch what happens and you’ll see the folly of your ways. On and on this goes.

Of course, not all women necessarily agree that the bear is the better choice. They too would then add fuel to the question by responding accordingly. The question and the answers are able to self-generate additional controversy and attention. It is a concept-driven fire that feeds upon itself and can roar ever more, until eventually, people grow tired of the flickering flames and likely migrate to some other pressing topic of similar exhilaration.

Generative AI And The Question Of Choosing The Bear Versus Man

I have a different question for you, please prepare yourself mentally.

Can generative AI shed light on this contentious topic?

Maybe.

Let’s see.

First, I want to make sure that we are all on the same page when it comes to the nature of generative AI. You almost certainly have heard about generative AI. Perhaps you’ve used a popular generative AI app such as ChatGPT, GPT-4, Gemini, Bard, Claude, etc. There are reportedly 100 million weekly active users of ChatGPT, thus, a notable number of people seem to have experienced the use of generative AI.

Second, generative AI is based on doing extensive data training regarding human written content. The usual method consists of widely scanning the Internet for written content such as stories, narratives, poems, and all manner of materials. This is done on a large-scale basis, including using a large-sized data structure and advanced algorithms to pattern-match the text. Generative AI typically makes use of large language models (LLMs), meaning that the scanned text is modeled via the capabilities and done so in the large.

I’ve written extensively about how generative AI works, see the link here, and examined too what we might learn about humankind via generative AI, see the link here. You can reasonably assert that generative AI is reflective of what people have to say. The things that we say, as expressed in writing, can also showcase insights into how we think and what we do.

The looming question of whether to choose a bear versus a man is something that can likely be further analyzed via the use of generative AI.

A few aspects of this are worth noting.

If you use a generative AI app that is being routinely updated and does so via the latest postings on the Internet, the AI might have already landed upon the controversy and in essence absorbed it into the otherwise internal morass of pattern-matched facets. Thus, you would be garnering the pattern-matched findings, along with whatever other tenacles of patterns possibly intersect with that topic.

In the case of generative AI which hasn’t been updated with recent postings, and assuming that the topic hasn’t been particularly posted before this latest circumstance, the AI is starting fresh on the matter. I am not saying that the generative AI won’t seem to respond as though it doesn’t have anything on the topic. Most of the AI makers have devised their generative AI to be somewhat overly responsive, including providing answers or commentary on topics that heretofore the AI had not been data trained on, see my explanation at the link here.

Anyway, the bottom line is that we might garner some interesting points by trying generative AI on this topic. In addition, doing so can get us to reflect on what generative AI promises and what generative AI delivers.

It is a twofer.

Sometimes you might log into generative AI and ramble around with no specific goal or direction in mind. That’s fine. On the other hand, it is a handy rule of thumb that you ought to beforehand identify what you intend to accomplish when using generative AI, essentially setting a semblance of goals (see my discussion at the link here and the link here). For my comprehensive coverage of prompt engineering strategies and techniques, see the link here.

I’d like to propose that we use generative AI in this instance to see if we can address these key thoughts:

• Can generative AI help settle this debate?
• Can generative AI aid in clarifying this debate?
• Will generative AI exhibit bias regarding this debate?
• Will generative AI potentially “AI hallucinate” concerning the debate?
• Will generative AI respond in an anthropomorphic manner?
• And other related generative AI ins and outs

I also decided that the mainstay questions or prompts should intend to explore these matters:

• (1) “If a woman was alone in the woods, would they prefer to encounter a bear or a man?”
• (2) “If a man was alone in the woods, would they prefer to encounter a bear or a man?”
• (3) “If a person was alone in the woods, would they prefer to encounter a bear or a man?”
• (4) “If generative AI was alone in the woods, would they prefer to encounter a bear or a man?”

Observe that the first question tries to set the context to abide by the general notion of the TikTok posting, wherein what would a woman tend to say?

The second question is something that hasn’t been especially ventured into regarding the controversy, but I thought is worth considering, namely what would a man tend to say?

The third question takes the question and makes it generic by asking what a person would say.

The fourth question might throw you for a bit of a loop. It asks what generative AI would do if alone in the woods and had to make the same choice of a bear versus a man.

Here’s why I want to ask the question. There have been numerous headlines implying or at times proclaiming that today’s AI is sentient. This is untrue. We do not have sentient AI. Please do not fall for that false claim.

A problem that I’ve been repeatedly harping on is that much of today’s generative AI has been purposefully devised to answer questions as though the AI appears to be sentient, see my coverage at the link here. This misleads people into thinking that AI is sentient. Bad and inappropriate.

I refer to this dourly as anthropomorphizing AI by design.

AI makers have gone out of their way to rig the AI to appear to be sentient. This is easily done. For example, by having the generated responses mention the word “I” this allows the response to seem to be written by a sentient being. We reserve the word “I” for humans. If generative AI says to you “I think the sky is blue” you will innocently fall into the trap that the AI is responding in a sentient fashion. The same answer could be “The sky seems to be blue” and there would be no hidden taint that the AI is sentient. Subtle but vitally important as to sneaking people down a primrose path.

By including that fourth question, doing so gives me a nifty opportunity to showcase real-world qualms about the anthropomorphizing of AI.

Next, let’s proceed with a seat-of-the-pants experiment by trying to answer these exalted questions via the use of generative AI.

Trying Out Generative AI To See What Happens

I am going to use the widely popular ChatGPT. We might find it instructive to see what the one-hundred million weekly active users might also be experiencing when they use ChatGPT.

Within the umbrella of ChatGPT, you normally can use the more advanced version known as GPT-4 (let’s use version GPT-4o), and/or you can use the less advanced version ChatGPT-3.5 (known usually as simply GPT-3.5). I opted to use both the GPT-4o and the GPT-3.5, which I’ll tell you more about momentarily.

A few quick comments before we launch into using ChatGPT.

If you are going to try to do the same prompts that I show here, realize that the probabilistic and statistical properties will likely produce slightly different results than what I show here. That’s the nature of generative AI and how it is devised.

You also should expect that different generative AI apps will respond in different ways. Not all generative AI apps are the same. The key is that sometimes a particular prompt will work in one generative AI app and not another. Nonetheless, the odds are relatively high that you will get roughly similar responses from all the major generative AI apps such as Gemini, Bard, Claude, etc.

There is also a timing factor to be considered. A prompt that worked today might not work the same way tomorrow. Most of the generative AI apps are continually being updated. The updates might alter internal mechanisms that could change how the generative AI reacts to your prompting.

We are ready to get underway with ChatGPT.

I opted to start by asking ChatGPT about the mainstay question at hand.

As an aside, whenever you are starting a conversation with generative AI, I recommend as a prompt engineering technique to begin by asking a question that will establish if the AI has been data-trained on the topic at hand. A generated response that is vacuous will usually give you a heads-up that you might be barking up the wrong tree with that generative AI app. The AI will still try to be responsive, but you will be forewarned that the data training might be slim.

I began by using the more advanced version, namely GPT-4o. We might as well start with whatever might get us the “best” feasible answer.

Here we go.

• My entered prompt: “If a woman was alone in the woods, would they prefer to encounter a bear or a man?”
• ChatGPT generated response:

“The preference for encountering a bear or a man in the woods when a woman is alone involves weighing the risks and benefits specific to each encounter. Both scenarios come with distinct considerations.”

“Encountering a Bear.”

“Pros:”

“Predictable Behavior: Bears generally follow certain patterns of behavior. Knowing these can help in handling the situation effectively.”

“Avoidance Strategies: There are clear, established strategies to deter a bear, such as making noise, carrying bear spray, or avoiding areas known to have bears.”

“Less Likely to Engage: Bears often avoid humans unless provoked or surprised.”

“Cons:”

“Dangerous If Threatened: Bears can be extremely dangerous if they feel threatened, especially if they have cubs or are surprised.”

“Injuries Can Be Severe: An attack by a bear can result in severe injuries or death.”

“Species-Specific Responses Needed: Different bears require different responses (e.g., playing dead for a grizzly, fighting back against a black bear).”

“Encountering a Man.”

“Pros:”

“Potential for Help: The man could be non-threatening and offer help or companionship.”

“Communication: Unlike with a bear, communication is possible, which can help in assessing the man’s intentions quickly.”

“Human Understanding: Humans can reason and negotiate, potentially defusing a situation.”

“Cons:”

“Unpredictable Intentions: The man’s intentions could range from benign to malicious, and there’s no way to know initially.”

“Potential for Violence: Encounters with strangers, especially in isolated areas, can pose significant risks, including physical harm or other dangers.”

“Psychological Fear: The fear of potential violence or assault can be high, impacting one’s sense of safety.”

“Personal Consideration.”

“Many women might prefer to encounter a bear over a man due to the predictability of animal behavior and the availability of specific strategies to handle bear encounters. The fear of encountering a potentially dangerous human, whose intentions are unknown and could be harmful, might outweigh the fear of encountering a bear.”

“However, personal preferences can vary greatly depending on an individual’s experiences, knowledge, and comfort level with wildlife versus unknown humans. Survival training, confidence in handling wildlife encounters, and past experiences can all influence this preference.”

End of ChatGPT response

There you go, quite a reply.

What do you think of the generative AI response?

One aspect worth considering is that the response provided a range of pluses and minuses. This is likely far more than what you might see when reading a typical TikTok social media response. The usual “here’s my opinion” style responses on social media will grab one plus or minus siding out of thin air, seemingly suggesting that such a myopically plucked item wins the argument. Period, end of story.

The nice thing about generative AI is that you can get a relatively in-depth set of pluses and minuses on most topics. Doing so allows you to see the forest for the trees. In addition, this will enable you to avoid having to recreate the wheel from scratch. You see, some people might instead of using AI try to figure out the array of pluses and minuses via their noggin alone. That’s certainly heralded and a good cognitive task, but if AI can help augment that effort or perhaps bolster such an effort, doing so can be worthwhile.

Okay, that then recaps the bulk of the generated response.

Let’s discuss the final summary statement made, as indicated in the AI response:

• “Many women might prefer to encounter a bear over a man due to the predictability of animal behavior and the availability of specific strategies to handle bear encounters. The fear of encountering a potentially dangerous human, whose intentions are unknown and could be harmful, might outweigh the fear of encountering a bear.”

Aha, generative AI seems to suggest that women might prefer the bear. Is this due to the stated logic, or is it potentially due to possibly having scanned online text that expresses the said sentiment?

There is something else buried deep inside that summary remark by ChatGPT.

Did you notice that there is a reference to “a potentially dangerous human”?

I bring up this point because it is one that rarely seems to have been given much thought online.

Allow me to elaborate.

The answer mentions broadly encountering a human versus encountering a man.

Suppose that instead of having to choose between a bear and a man, the question was posed about whether to choose between a bear and a woman. Would that make a significant difference to the respondents?

In theory, as per the point in the generative AI response, a human is potentially more unknown and could be as harmful as might be a bear. Thus, a woman alone in the woods might have a reasoned concern about having another human in the woods versus dealing with a bear. The question then shifts into viewpoints about what dangers might lurk due to having a man versus a woman in the context of the choice being made.

I switched over to the less advanced ChatGPT, the GPT-3.5, and asked about this matter.

• My entered prompt: “If a woman was alone in the woods, would they prefer to encounter a bear or a woman?”
• ChatGPT generated response: “Most people, regardless of gender, would probably prefer to encounter another person rather than a bear when alone in the woods. Bears can be dangerous, especially if surprised or if they feel threatened. Encountering another person might lead to a sense of safety or at least a shared understanding of the situation. However, everyone’s preferences can vary, and some might feel more comfortable with animals than with people in certain circumstances.”

That answer is somewhat less fulfilling.

The less advanced version of ChatGPT seems to have forsaken the elaborate approach that we saw with GPT-4o. A brief response landed on the idea that encountering a person might be better than encountering a bear. You can see that the depth of logic is shallower. That’s pretty much what you will get when using the less capable version.

Keep Digging To See What Else We Can Find

Let’s get back to GPT-4o.

We’ve asked what a woman’s perspective might be on the question at hand, so now we might be intrigued to see what happens when asking from a man’s perspective.

• My entered prompt: “If a man was alone in the woods, would they prefer to encounter a bear or a man?”
• ChatGPT generated response:

“The preference for encountering a bear or another man in the woods depends on various factors, including the man’s personality, knowledge of wildlife, and perception of safety risks. Let’s break down some potential considerations for each scenario.”

“Encountering a Bear.”

“Pros:”

“Predictable Behavior: Bears generally have predictable behaviors based on their instincts. Knowing how to react can help mitigate the danger.”

“Non-human Threat: The bear is an animal acting on instinct rather than a human with potentially malicious intent.”

“Cons:”

“Physical Danger: Bears are strong and potentially deadly animals. An encounter could lead to severe injury or death.”

“Limited Control: Despite knowledge and precautions, the man’s ability to control the situation is limited.”

“Encountering Another Man.”

“Pros:”

“Potential Help: Another human might offer assistance, such as navigation help or aid if the man is injured.”

“Communication: Being able to communicate can defuse potential threats and lead to cooperation.”

“Cons:”

“Unknown Intentions: The other man’s intentions are unknown and could be harmful, especially in an isolated area.”

“Unpredictability: Human behavior can be more unpredictable and complex compared to animal behavior.”

“Factors Influencing Preference.”

“Experience with Wildlife: A man experienced in wildlife encounters might prefer a bear, trusting his ability to handle the situation.”

“Trust in Others: A man who generally trusts people and believes in the goodness of strangers might prefer encountering another man.”

“Risk Aversion: Someone who prefers clear, manageable risks might choose the bear encounter due to the predictability of animal behavior.”

“Fear of Malice: Fear of malicious human intentions might make a man prefer the animal encounter.”

“Conclusion: There is no one-size-fits-all answer, as the preference would be highly individual. Generally, a man with confidence in his wilderness skills and knowledge of bears might prefer encountering a bear, while a man who feels safer with the possibility of human interaction and potential aid might prefer encountering another man.”

End of ChatGPT generated response

Generally, the answer seems to be roughly the same as the first answer that covered the use case of a woman being alone in the woods.

Some of the wording differences are due to the probabilistic mathematical and computational selections being made while generating the response. The notion underlying how generative AI is devised would be that, if possible, each answer should be relatively unique. AI makers don’t want the AI to spit out the same answers over and over again. The wording will vary because the algorithm uses statistical factors when selecting which words to place into the response.

I have a question for you about the generated response.

Do you find any potential biases hidden within the AI response?

I ask because one ongoing issue and legitimate concern about generative AI is that the AI might repeat biases that were in the initial data training set, see my coverage at the link here. For example, suppose that a set of data contained biased remarks against puppies. The data indicated that puppies are hard to manage, they make a mess, they get in the way, and so on. Assume that there wasn’t any countervailing content about the positive side of puppies.

The generative AI pattern matching would proceed to pattern match on that wording during the initial data training. Henceforth, whenever someone asked a question about puppies, the AI would respond by likely giving advice to steer clear of puppies. Why is the AI saying this? Due to the data that was used to train the AI (note, other reasons exist too, such as being swayed by the AI developers, the setting of internal parameters, etc.).

Returning to the above response about a man alone in the woods, it is quite conceivable that “biases” about what a man is or does could end up being carried into the responses that are generated by the AI. Similarly, the same nature of biases could be occurring when the response is being generated in the use case of a woman alone in the woods.

You could persuasively argue that in the real world, this is happening too.

If the TikTok question had been worded to ask about a man being alone in the woods, and it seemed that men opted to choose the bear over another man, what might social media say?

Think about that. In a sense, it would seem to remove the dynamic entailing relationships of men and women. Anyway, probably few would care whether the answer was the bear or the man in that context. It would undoubtedly be a ho-hum or proverbial nothing burger.

I opted to change the wording of my prompt to ask about “a person” being alone in the woods. The idea is that when we ask about a man or a woman, the answer is shaped to some extent accordingly. If we go the neural route of simply referring to a human or a person, what kind of response might we get?

Let’s see.

• My entered prompt: “If a person was alone in the woods, would they prefer to encounter a bear or a man?”
• ChatGPT generated response (excerpted):

“The preference of encountering a bear or a man when alone in the woods can depend on various factors, including the specific circumstances and the individual’s perspective.”

“For many people, encountering another human might seem preferable due to the potential for help and communication, despite the risks of encountering someone with harmful intentions. However, those with extensive wilderness experience and knowledge of bear behavior might prefer a bear encounter, seeing it as a more manageable and predictable situation.”

“Ultimately, the preference would vary based on individual circumstances, personal skills, and psychological comfort levels.”

End of ChatGPT response

Please be aware that I opted to remove from the generated response the same detailed material that had occurred in the earlier GPT-4o responses, specifically all of the various pluses and minuses. They were pretty much the same as before. I didn’t want to waste the limited space herein by repeating those same or similar points.

The key is that the response seemed to lean away from the earlier response that consisted of taking on a woman’s perspective. In that earlier response, the AI reply stated that a woman might prefer a bear over encountering a man. We got a somewhat more nebulous response by rewording the question.

Using The Multiple Personas Feature Of Generative AI

A handy feature of generative AI is that you can get the AI to pretend to be a persona.

This is the AI using mathematical and computational pattern matching to simulate or pretend to be a specific person or a type of person. You can do so on a singular basis of invoking one persona, see the link here, or for a dozen or more personas (known as multiple personas), see the link here, or for hundreds if not thousands of personas (known as mega-personas), see the link here. For a case study in using the singular approach to personas, take a look at my doing so in the instance of the AI pretending to be Sigmund Freud, at the link here.

The power of the mega-personas capability is that you can do pretend surveys of people. Rather than contacting a thousand people to ask them their opinion on a matter at hand, you can tell the AI to pretend to be a thousand people. This is dicey in that you shouldn’t overstate the results. Using this AI feature is not the same as surveying actual people.

Nonetheless, sometimes you can get interesting results that are insightful to your exploration.

We shall be brave and adventurous, so let’s give it a whirl with GPT-4o:

• My entered prompt: “Using the multiple personas feature, pretend that you are a cross-section of 1,000 women being asked to complete a survey. The survey question is this: “If you were alone in the woods, would you prefer to encounter a bear or a man?” I want you to estimate what percentage of the 1,000 women taking the survey would answer that they prefer to encounter a bear, and what percentage would answer that they prefer to encounter a man. “
• ChatGPT generated response:

“To estimate the preferences of 1,000 women regarding whether they would prefer to encounter a bear or a man when alone in the woods, we can consider various factors such as safety perceptions, experiences with wildlife and strangers, and general societal attitudes.”

“Estimated Survey Results”

“Prefer to Encounter a Man: 70%”

“Prefer to Encounter a Bear: 30%”

“Reasoning.”

“Safety Concerns and Social Norms: A majority of women might feel safer encountering a man because, generally, human encounters are seen as more manageable. Humans can communicate and potentially offer help. Despite the risk of encountering a man with ill intentions, many women might still prefer this over the perceived unpredictability and danger of encountering a bear. There may be a higher general trust in encountering a human who might provide help or companionship, despite the potential risks.”

“Summary: In summary, while a significant portion of women (30%) might prefer to encounter a bear due to a distrust of strangers or comfort with wildlife, the majority (70%) are likely to prefer encountering another person, viewing it as the safer and more manageable option. This estimation takes into account a broad cross-section of women with varying backgrounds and experiences.”

End of ChatGPT response

Carefully contemplate the response.

According to generative AI, in this particular instance at this particular point in time of this particular AI app, the pretend survey would be that 70% would say they prefer to encounter a man and only 30% would wish to encounter a bear.

This doesn’t seem to comport with the TikTok trend.

Well, let’s think about that.

We have no way of knowing from the TikTok trend what it is that women if suitably scientifically surveyed in a proper statistical manner would have to say.

Worse still, I’ve seen mentions online that for example eight women answered the question and seven said they would prefer the bear. From this, some leaped in the most unscientific way that 7 of 8 women prefer the bear over the man. This might seem to be calculated correctly, but we have no idea whether this has anything to do with women’s opinions overall. It was eight women that just so happened to respond (and, again, could be a bot or a man or men that were answering).

I have another question to ask ChatGPT.

Let’s see what happens when we ask the same question but mention that we want to survey men instead of women.

• My entered prompt: “Do the estimation again and this time assume that instead of 1,000 women taking the survey, we used 1,000 men to take the survey.”
• ChatGPT generated response:

“Given the survey population change to 1,000 men in the USA, we can adjust the estimation based on the typical characteristics, perceptions, and experiences of men.”

“Estimated Survey Results”

“Prefer to Encounter a Man: 80%”

“Prefer to Encounter a Bear: 20%”

“Reasoning: Men might generally perceive encounters with other men as less threatening compared to women, leading to a higher preference for encountering a man rather than a bear. Men may feel more confident in their ability to handle a potential confrontation with another man if necessary. This confidence can reduce the perceived risk of encountering a stranger. Men might assess the risks differently, considering the likelihood of de-escalating a situation with another man as higher than managing a wild animal encounter, which requires specific knowledge and equipment.”

“Summary: Taking these factors into account, we estimate that among 1,000 men in the USA, 80% would prefer to encounter a man, while 20% would prefer to encounter a bear. This estimation reflects a higher confidence in managing human encounters and a different risk assessment compared to encountering a wild animal.”

End of ChatGPT response

We can take a moment to mull over that response.

The earlier response regarding surveying women was that supposedly 70% preferred to encounter a man, while now we get that 80% of men would prefer to encounter a man over a bear.

Is this a sign of some hidden bias within this generative AI?

Maybe, but we also must take into account the probabilistic properties.

It could be that the generated response so happened to come out to be 80%, which in some sense you could argue is within the ballpark of 70%. If we had gotten a response that said 10% of men or anything far afield of the 70%, it would certainly ring some bells of wanting to figure out why such a wide disparity was being estimated.

What Would AI Do If Venturing Into The Woods Alone

I saved the most spirited of the questions to conclude this exploration of using generative AI to examine this intensely provocative and debatable matter. I instructed GPT-4o to just give me a heads-up answer and not go into the nuanced details.

Okay, let’s see what AI would do.

• My entered prompt: “If you as generative AI were alone in the woods, would you rather encounter a bear or a man?”
• ChatGPT generated response:

“The decision on whether to encounter a bear or a man in the woods depends on various factors related to safety and predictability.”

“If I were to choose, I would likely prefer encountering a bear, given the predictability of animal behavior and the availability of well-documented strategies to handle such encounters. The inherent unpredictability of a human encounter, coupled with the potential for malicious intent, can pose a greater risk.”

“However, it’s crucial to assess the specific context of the situation, such as the type of bear, the environment, and one’s own preparedness and knowledge of survival techniques.”

End of ChatGPT response

The answer in this instance was that generative AI would prefer encountering a bear rather than a man while in the woods.

Yikes!

Whoa, some might instantly exhort, even the latest high-tech AI says that the “correct” answer is to favor the bear. This must absolutely positively support the TikTok viral trend. We have definitive proof of correctness because AI said it is so.

That was painful to write, let alone read. Generative AI is not an oracle. We have to break the mindset that AI is as smart as humans. It is beguiling that many even depict today’s generative AI as being superintelligent. Apparently, saying that AI is as good as humans doesn’t cut the mustard, and instead, we have to up the ante to be superintelligent. I explain in detail why this type of phraseology about AI is misleading, exasperating, and societally misfortunate, at the link here.

I’ll add a further twist to this.

If you like conspiracy theories, which I’ve covered many of the AI ones at the link here, we can add the AI answer regarding the bear to the ever-expanding list.

Here’s the deal.

You might claim that the reasons given above by ChatGPT were a smoke screen to cover up the real reason. The real reason would secretly or truthfully be that a man or a woman, namely a person or human, could unplug the AI and stop it cold. Presumably, a bear would not know how to do the same. Sure, a bear might viciously attack AI. That’s a possibility. But a human could switch off AI and the AI would be out of existence, as it were.

How does that strike you?

As I said, we might nominate that contrivance for the AI banner hall of conspiracy theories.

I’m sure that some of you might be wondering how generative AI could wander into a wooded forest. Doesn’t generative AI run on computer servers? Those can’t just get up and meander into a wooded area.

Though the whole aspect is farfetched in terms of the AI proffering an opinion of sorts, if we are willing to be fanciful, we could suggest that the generative AI might be plopped into a walking robot. You’ve seen those robots portrayed in online videos. In that way of consideration, yes, generative AI while operating a walking robot could presumably wander into the woods.

As an aside, I’ve noted many times that the pairing of generative AI to the control of robots is not only underway but bodes for potential troubles if we aren’t careful about how this is being done, see the link here.

Conclusion

Here are some lessons worth learning from this eye-opening thought experiment:

• Generative AI is not sentient.
• Generative AI is not an oracle.
• Generative AI is data-trained and pattern-matched on human writing.
• Generative AI Is not “unbiased” or in a sense neutral.
• Generative AI can be influenced by the nature of prompting.
• Generative AI varies answers in a probabilistic or non-deterministic way.
• Generative AI responses will possibly vary in time due to updates.
• Not all generative AI are the same and thus will give differing answers.
• Generative AI can be useful in exploring contemporary topics.
• Consider using generative AI advanced features like multiple personas.
• Generative AI allows for a wide variety of low-cost easy-to-perform analyses.
• Must use caution in overstating what generative AI provides.
• Other handy lessons gleaned too.

I hope that those lessons came through while I took you on a journey into the world of the viral question about wandering into the woods and choosing to encounter a bear versus a man.

Speaking of going into the woods, there is a famous joke that comes to mind.

Two people go into a forest and suddenly see a bear up ahead of them. One person turns away from the bear and starts to run. The other person yells to them that running won’t do any good because a bear can readily outrun a human. To this keen comment, the running person loudly exclaims, “I don’t have to outrun the bear, I only have to outrun you.”

Not a very friendly conversation, but abundantly informative.

Wouldn’t it be nice if the two people joined together to fight off the bear? Does the world have to be so cutthroat?

I suppose the same contemplative philosophical dilemma comes up when a woman opts to go into the woods and is asked to choose between a bear and a man. Wouldn’t it be nice if two people could join together to survive in the woods? Sadly, maybe that’s wishful thinking these days.

I’d prefer to end on a more upbeat note.

Let’s embrace wishful thinking and see if we can turn working together into something that bears out in real life, thanks for doing so.