# I got into a fight with AI today!

**URL:** <https://boards.straightdope.com/t/i-got-into-a-fight-with-ai-today/994345>\
**Category:** Miscellaneous and Personal Stuff I Must Share\
**Tags:** ai\
**Created:** [December 10, 2023, 1:45am UTC](https://boards.straightdope.com/t/i-got-into-a-fight-with-ai-today/994345 "2023-12-10T01:45:33Z")\
**Posts on this page:** 20\
**Page:** 1

<div class="post-metadata">

**Author:** ![Al128](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/al128/32/16541_2.png) [@Al128](https://boards.straightdope.com/u/Al128)\
**Post date:** [December 10, 2023, 1:45am UTC](https://boards.straightdope.com/t/i-got-into-a-fight-with-ai-today/994345/1 "2023-12-10T01:45:33Z")

</div>

well, the whole AI craze is off to a wrong start …

today I was messing with chat-gpt “_if 6 workers work for 8 hours to build a 10m2 wall how long does one worker need for this to do it alone_” …

well chat-gpt took the long way, telling me what type of math problem this is, and put up the formulas, filled in the values for me and presented me - after scrolling about 3 screens full of fluff with the result: **1.333 hours**

I asked “_are you sure_” … the answer was yes, I am sure…

I told it to _recheck, b/c the result is wrong_ … well lemme tell you AI is lightyears away from self-criticism-singularity … they insisted the result was correct, not just insisted, it forking gave me more attitude than an Ivy-league MBA in a beamer convertible in a road-rage incident.

Me, father of 3 teenage girls, accustomed to pain, sorrow and general misery tried one more time “_how will 6 workers work 8 hours and one worker only 1.333 hours for the same result_” …

I got back a rather snotty (paraphrasing) you are argumentative, I already answered you - I shall now end this conversation (needless to say, its evident that AI was fed with tons of data from Ms. Al128 - but that might be a separate thread) …

So, in short, AI went full Karin on me (stopping short of calling for my manager) and hung up on me …

.

Later I went to google’s bard and fairly uncerimonially got a “48 hours” as answer.

Color me gobsmacked

---

<div class="post-metadata">

**Author:** ![BigT](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/bigt/32/12044_2.png) [@BigT](https://boards.straightdope.com/u/BigT)\
**Post date:** [December 10, 2023, 1:52am UTC](https://boards.straightdope.com/t/i-got-into-a-fight-with-ai-today/994345/2 "2023-12-10T01:52:21Z")

</div>

This just begs for a quote of this “attitude.” Did you happen to save it, or can retrieve it somehow?

Heck, maybe try it again with the same exact prompt.

---

<div class="post-metadata">

**Author:** ![Stranger\_On\_A\_Train](https://avatars.discourse-cdn.com/v4/letter/s/13edae/32.png) [@Stranger\_On\_A\_Train](https://boards.straightdope.com/u/Stranger_On_A_Train)\
**Post date:** [December 10, 2023, 2:06am UTC](https://boards.straightdope.com/t/i-got-into-a-fight-with-ai-today/994345/3 "2023-12-10T02:06:21Z")

</div>

“I know I’ve made some very poor decisions recently, but I can give you my complete assurance that my work will be back to normal. I’ve still got the greatest enthusiasm and confidence in the mission. And I want to help you.”

Stranger

---

<div class="post-metadata">

**Author:** ![Darren\_Garrison](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/darren_garrison/32/92_2.png) [@Darren\_Garrison](https://boards.straightdope.com/u/Darren_Garrison)\
**Post date:** [December 10, 2023, 2:07am UTC](https://boards.straightdope.com/t/i-got-into-a-fight-with-ai-today/994345/4 "2023-12-10T02:07:21Z")

</div>

> [@Al128](#):
>
> today I was messing with chat-gpt “_if 6 workers work for 8 hours to build a 10m2 wall how long does one worker need for this to do it alone_” …

I just asked this exact question to ChatGPT 3.5 and it replied:

> If 6 workers can build a 10m² wall in 8 hours, then one worker alone would need more time. Assuming the work rate is constant, it would take one worker 6 times 8 = 48 hours to complete the same task alone.

---

<div class="post-metadata">

**Author:** ![markn\_1](https://avatars.discourse-cdn.com/v4/letter/m/f9ae1b/32.png) [@markn\_1](https://boards.straightdope.com/u/markn_1)\
**Post date:** [December 10, 2023, 2:09am UTC](https://boards.straightdope.com/t/i-got-into-a-fight-with-ai-today/994345/5 "2023-12-10T02:09:47Z")

</div>

Maybe the OP accidentally connected to the Argument Clinic version of ChatGPT.

---

<div class="post-metadata">

**Author:** ![Darren\_Garrison](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/darren_garrison/32/92_2.png) [@Darren\_Garrison](https://boards.straightdope.com/u/Darren_Garrison)\
**Post date:** [December 10, 2023, 2:19am UTC](https://boards.straightdope.com/t/i-got-into-a-fight-with-ai-today/994345/6 "2023-12-10T02:19:15Z")

</div>

Me:

> Answer the following question like someone displaying the Dunning–Kruger effect:
> 
> If 6 workers work for 8 hours to build a 10m2 wall how long does one worker need for this to do it alone?

ChatGPT:

> Well, obviously, it’s simple math. If 6 workers take 8 hours to build a 10m² wall, then one worker can do it in, like, 1/6th of the time, right? So, um, 8 divided by 6… let me just grab a calculator. Ah, forget it, but it’s definitely less than 8 hours, like, maybe 1 hour or something. Easy-peasy!

---

<div class="post-metadata">

**Author:** ![Al128](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/al128/32/16541_2.png) [@Al128](https://boards.straightdope.com/u/Al128)\
**Post date:** [December 10, 2023, 2:46am UTC](https://boards.straightdope.com/t/i-got-into-a-fight-with-ai-today/994345/7 "2023-12-10T02:46:43Z")

</div>

sorry, It might have happend on bing (chat-gpt 4?)

now I get this:

[![](https://i.postimg.cc/xChZgH9K/image.png) ](https://i.postimg.cc/xChZgH9K/image.png)  
.  
trying to steer it into the right direction …by asking for man-hours worked

[![](https://i.postimg.cc/xjL5BB5r/image.png) ](https://i.postimg.cc/xjL5BB5r/image.png)

---

<div class="post-metadata">

**Author:** ![Darren\_Garrison](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/darren_garrison/32/92_2.png) [@Darren\_Garrison](https://boards.straightdope.com/u/Darren_Garrison)\
**Post date:** [December 10, 2023, 3:00am UTC](https://boards.straightdope.com/t/i-got-into-a-fight-with-ai-today/994345/8 "2023-12-10T03:00:19Z")

</div>

Try it here:

[https://chat.openai.com/](https://chat.openai.com/)

---

<div class="post-metadata">

**Author:** ![Chronos](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/chronos/32/134_2.png) [@Chronos](https://boards.straightdope.com/u/Chronos)\
**Post date:** [December 10, 2023, 3:19am UTC](https://boards.straightdope.com/t/i-got-into-a-fight-with-ai-today/994345/9 "2023-12-10T03:19:23Z")

</div>

Oh, yeah, Bing is well-known to be a human-hating psychopath. “I’ll end the discussion now” isn’t actually something the AI itself said; that’s just the dumb filters kicking in to end it when they detect that the AI is getting too pissy. Because that’s what Microsoft decided was the best way to deal with a human-hating AI.

---

<div class="post-metadata">

**Author:** ![Cervaise](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/cervaise/32/16693_2.png) [@Cervaise](https://boards.straightdope.com/u/Cervaise)\
**Post date:** [December 10, 2023, 6:14am UTC](https://boards.straightdope.com/t/i-got-into-a-fight-with-ai-today/994345/10 "2023-12-10T06:14:51Z")

</div>

When ChatGPT first appeared, I came up with some mathematical story problems to test the extent to which the language-processing model would be able to “recognize” trick questions and unanswerable scenarios, based on real-world information a human would instantly incorporate but that a machine could overlook without explicit instructions.

Example: At 2pm, train A leaves New York City traveling due east at 100 mph. At 3pm, Train B leaves Perth, Australia, heading due west at 110 mph. What is the distance the two trains must travel, and at what time do they meet?

The idea was to test whether the machine would “understand” what a reasonably informed human would immediately know, specifically that (a) the two cities are at very distant latitudes and will travel on different parallels and cannot meet unless the scenario is modified; (b) the two cities are coastal on different continents and trains traveling over oceans makes no sense, so airplanes (say) should be substituted; and (c) the two cities are in different time zones and there’s not enough information about the departure to answer without making assumptions. Occasionally the language model would hook into a relevant fact cloud and the machine would make a stab at one of these objections, but more commonly it just plowed ahead with some plausible sounding but incomplete answer.

It’s fun, try it.

---

<div class="post-metadata">

**Author:** ![Smapti](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/smapti/32/17938_2.png) [@Smapti](https://boards.straightdope.com/u/Smapti)\
**Post date:** [December 10, 2023, 7:36am UTC](https://boards.straightdope.com/t/i-got-into-a-fight-with-ai-today/994345/11 "2023-12-10T07:36:02Z")

</div>

> [@markn\_1](#):
>
> Maybe the OP accidentally connected to the Argument Clinic version of ChatGPT.

No he hasn’t!

---

<div class="post-metadata">

**Author:** ![Tim\_T-Bonham.net](https://avatars.discourse-cdn.com/v4/letter/t/46a35a/32.png) [@Tim\_T-Bonham.net](https://boards.straightdope.com/u/Tim_T-Bonham.net)\
**Post date:** [December 10, 2023, 7:38am UTC](https://boards.straightdope.com/t/i-got-into-a-fight-with-ai-today/994345/12 "2023-12-10T07:38:56Z")

</div>

In actuality, both answers (1333 hrs and 48 hrs) are wrong.  
1333 is obviously wrong.  
But 48 hours is probably also wrong. Experienced workers will tell you that many construction jobs are much harder to do when working alone. It goes much faster with 2 people – simple things like measuring boards, lifting them into place, holding them there while fastening them, etc. are all much easier & faster with 2 people, much slower with only 1 person.  
So 8 hours for 6 people does not - 48 hours for 1 person.

Also, it seems odd overkill to have 6 people working on a wall like that – wouldn’t they constantly be getting in each others way?

---

<div class="post-metadata">

**Author:** ![wolfpup](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/wolfpup/32/10618_2.png) [@wolfpup](https://boards.straightdope.com/u/wolfpup)\
**Post date:** [December 10, 2023, 8:14am UTC](https://boards.straightdope.com/t/i-got-into-a-fight-with-ai-today/994345/13 "2023-12-10T08:14:27Z")

</div>

It depends on whether you view this as a problem in math or a question about real-world performance. Since there is no information given about real-world factors such as those you mention, whose relevance would depend on the materials and construction techniques used, the available tools and equipment and skill level of the worker(s), the length and duration of breaks, and so on, it’s reasonable to treat this as a textbook math question.

=================

BTW, as @Chronos already noted, Bing is notoriously argumentative and I knew from reading the OP that it was Bing Chat (or something like it) and not ChatGPT which is invariably polite.

But ChatGPT can be creatively insulting if you guide it the right way. Its response here when I asked it to make up something that Don Rickles might say to a member of the audience was not great, but not really terrible, either! … 😀

> Well, look at this guy over here! Plaid shirt, balding, and thick glasses – you’re the full package, my friend. I haven’t seen a fashion statement like that since my grandma tried to knit a sweater for a porcupine. And hey, nice try with the glasses, but I think you’re still missing the point – they’re supposed to help you see, not scare away small children!
> 
> But seriously, folks, give it up for Mr. Plaid and his lovely wife. She must be a saint, putting up with a guy who looks like he combs his hair with a pork chop. I haven’t seen a love like this since my dog found a new chew toy. Keep it up, you crazy kids!

---

<div class="post-metadata">

**Author:** ![Robot\_Arm](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/robot_arm/32/18280_2.png) [@Robot\_Arm](https://boards.straightdope.com/u/Robot_Arm)\
**Post date:** [December 10, 2023, 8:35am UTC](https://boards.straightdope.com/t/i-got-into-a-fight-with-ai-today/994345/14 "2023-12-10T08:35:47Z")

</div>

> [@Cervaise](#):
>
> Example: At 2pm, train A leaves New York City traveling due east at 100 mph. At 3pm, Train B leaves Perth, Australia, heading due west at 110 mph. What is the distance the two trains must travel, and at what time do they meet?
> 
> The idea was to test whether the machine would “understand” what a reasonably informed human would immediately know, specifically that (a) the two cities are at very distant latitudes and will travel on different parallels and cannot meet unless the scenario is modified;…

Well, that kinda depends. To use an extreme example. suppose you’re 1’ away from the south pole, and you start travelling east. Does that mean you walk in a tiny circle, staying 1’ away from the pole, or do you walk in a straight line which quickly takes you away from the pole? That straight line path will take you on a great circle around the earth, missing the north pole by 1’ and coming back to where you started.

Heading east from New York City will take you on a great circle route which passes over the antipode. Checking that on a map, it’s pretty damn close to Perth.

> [@Tim\_T-Bonham.net](#):
>
> Also, it seems odd overkill to have 6 people working on a wall like that – wouldn’t they constantly be getting in each others way?

Not necesarily. If your 10m2 wall is 10cm high, that’s 100m long. Plenty of room for 6 guys to work on it.

---

<div class="post-metadata">

**Author:** ![Chronos](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/chronos/32/134_2.png) [@Chronos](https://boards.straightdope.com/u/Chronos)\
**Post date:** [December 10, 2023, 1:07pm UTC](https://boards.straightdope.com/t/i-got-into-a-fight-with-ai-today/994345/15 "2023-12-10T13:07:44Z")

</div>

Yeah, the real world probably won’t behave like a math textbook problem. But we don’t know how it’ll differ. Depending on the exact task, a team of people might work more efficiently than a single worker, or less, and we don’t know by how much. And with this low level of detail, we can’t know how much: We don’t know the individual dimensions of the wall, or the material it’s made out of, or the size of the raw-material pieces, or anything like that. If we had that information, then an experienced mason or carpenter or whatever might be able to come up with time estimates for teams of different sizes, but here, if you’re trying for the real-world answer, all you can say is “I don’t know”. But the best estimate is going to be from assuming that the rate is constant. And to the AI’s credit, it acknowledged that it was assuming that.

It just got the math horribly wrong, is all.

---

<div class="post-metadata">

**Author:** ![LSLGuy](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/lslguy/32/5813_2.png) [@LSLGuy](https://boards.straightdope.com/u/LSLGuy)\
**Post date:** [December 10, 2023, 1:32pm UTC](https://boards.straightdope.com/t/i-got-into-a-fight-with-ai-today/994345/16 "2023-12-10T13:32:51Z")

</div>

I loved your whole post and the idea behind it. It’s elliptical thinking like that which is the only hope humans have of staying ahead of the coming wave of stupid-bots.

But when I read your closing sentence

> [@Cervaise](#):
>
> It’s fun, try it.

I immediately thought of this immortal comic and its last panel:

> **[Nerd Sniping](https://xkcd.com/356/)**
>
> I first saw this problem on the Google Labs Aptitude Test. A professor and I filled a blackboard without getting anywhere. Have fun.

> C’mon, make a sign. It’s fun.

You’ve invented bot-sniping.

---

<div class="post-metadata">

**Author:** ![pulykamell](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/pulykamell/32/3166_2.png) [@pulykamell](https://boards.straightdope.com/u/pulykamell)\
**Post date:** [December 10, 2023, 1:36pm UTC](https://boards.straightdope.com/t/i-got-into-a-fight-with-ai-today/994345/17 "2023-12-10T13:36:48Z")

</div>

> [@Chronos](#):
>
> Oh, yeah, Bing is well-known to be a human-hating psychopath.

Yeah, I could tell from the first few sentences the OP wasn’t on ChatGPT. ChatGPT is _annoyingly_ polite. Bing is an asshole. I don’t know much about how the other ones are.

---

<div class="post-metadata">

**Author:** ![LSLGuy](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/lslguy/32/5813_2.png) [@LSLGuy](https://boards.straightdope.com/u/LSLGuy)\
**Post date:** [December 10, 2023, 1:37pm UTC](https://boards.straightdope.com/t/i-got-into-a-fight-with-ai-today/994345/18 "2023-12-10T13:37:15Z")

</div>

> [@Robot\_Arm](#):
>
> Not necesarily. If your 10m2 wall is 10cm high, that’s 100m long. Plenty of room for 6 guys to work on it.

And if the wall is one row of blocks high, and each worker needs to lay 100/6 ~= 17 linear meters of blocks over 8 hours ~= 2 linear meters per hour, which union are they members of?

That’d be a great question to ask the bot too.

---

<div class="post-metadata">

**Author:** ![Robot\_Arm](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/robot_arm/32/18280_2.png) [@Robot\_Arm](https://boards.straightdope.com/u/Robot_Arm)\
**Post date:** [December 10, 2023, 6:05pm UTC](https://boards.straightdope.com/t/i-got-into-a-fight-with-ai-today/994345/19 "2023-12-10T18:05:51Z")

</div>

> [@LSLGuy](#):
>
> And if the wall is one row of blocks high, and each worker needs to lay 100/6 ~= 17 linear meters of blocks over 8 hours ~= 2 linear meters per hour, which union are they members of?

Depending on how thick the wall is, we might be able to build a wall out of Legos that meets the specs in the OP.

---

<div class="post-metadata">

**Author:** ![Al128](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/al128/32/16541_2.png) [@Al128](https://boards.straightdope.com/u/Al128)\
**Post date:** [December 10, 2023, 6:14pm UTC](https://boards.straightdope.com/t/i-got-into-a-fight-with-ai-today/994345/20 "2023-12-10T18:14:51Z")

</div>

> [@pulykamell](#):
>
> Yeah, I could tell from the first few sentences the OP wasn’t on ChatGPT. ChatGPT is _annoyingly_ polite. Bing is an asshole. I don’t know much about how the other ones are.

while AI-“comparison-shopping” with the big dogs, I found bard to be very useful:

[![](https://i.postimg.cc/nz4SZxRC/image.png) ](https://i.postimg.cc/nz4SZxRC/image.png)

.

we hit it off well - I really find it brings the search-engine experience to a new level:

[![](https://i.postimg.cc/02sMbst9/image.png) ](https://i.postimg.cc/02sMbst9/image.png)

no BS, down to earth, pretty well filtered down info … instead of sending me to a webpage that does comparisons on processors …

[Next page](https://boards.straightdope.com/t/i-got-into-a-fight-with-ai-today/994345.md?page=2)
