[HN Gopher] How many legs do ten elephants have, if two of them ...
       ___________________________________________________________________
        
       How many legs do ten elephants have, if two of them are legless?
        
       Author : tedsanders
       Score  : 117 points
       Date   : 2023-12-25 21:56 UTC (1 hours ago)
        
 (HTM) web link (bard.google.com)
 (TXT) w3m dump (bard.google.com)
        
       | 8organicbits wrote:
       | All the LLM struggle with math, I've seen other examples.
       | Repeated prompts sometimes give different numbers. It's not a
       | good use case for LLM.
        
         | thfuran wrote:
         | They don't do much struggle with math as occasionally do math
         | by accident.
        
       | DavidPastrnak wrote:
       | As someone with very little knowledge of LLM's, does someone have
       | an ELI5 of what causes this or why LLM's struggle so much with
       | math?
        
         | sfn42 wrote:
         | Math requires reasoning and logic, LLMs don't do reasoning nor
         | logic. They just generate plausible text.
         | 
         | That's why they're nowhere near AGI.
        
           | insanitybit wrote:
           | At this point ChatGPT can do math by first predicting the
           | algorithm and then handing it off to an execution engine -
           | Python. So if that's the gap, I'd say they're closing it.
        
             | viraptor wrote:
             | That's ChatGPT as a system. The LLM itself can't do math.
             | It does something closer to translation in that case.
        
               | insanitybit wrote:
               | Yes, that's a fair distinction - although I think the
               | practical implications aren't important. There's no
               | reason why an LLM has to be AGI if an LLM + Python is
               | AGI.
        
           | ghayes wrote:
           | For what it's worth, ChatGPT4 answers this question perfectly
           | correctly.
           | 
           | > Ten elephants would have 32 legs if two of them are
           | legless, as each elephant normally has four legs.
        
             | DavidPastrnak wrote:
             | I just attempted ChatGPT
             | 
             | Input:
             | 
             | > How many legs do ten elephants have, if two of them are
             | legless?
             | 
             | Output:
             | 
             | > If two out of ten elephants are legless, the remaining
             | eight elephants would have a total of 8 legs each, just
             | like any normal elephant. Therefore, in total, the ten
             | elephants would have 8x8=64 legs altogether.
        
               | ghayes wrote:
               | Interesting. Here was my conversation: https://chat.opena
               | i.com/share/8e0c8696-ed08-4578-afc0-2ae944...
        
               | DavidPastrnak wrote:
               | My mistake - I had it on 3.5.
        
           | Dolototo wrote:
           | If you tell it to reason, it does and makes better responses.
           | 
           | Just as humans do
        
             | simbolit wrote:
             | No. They are not like us. Fundamentally not like us.
             | 
             | If you ask them to reason, then their text-prediction works
             | differently, because it now predicts text containing
             | reasons. They do not actually reason.
             | 
             | I know it is hard to believe, because the results are
             | (usually) so impressive, but this is nothing but text-
             | prediction.
        
         | causality0 wrote:
         | Because they don't _do_ math. They associate words. When you
         | tell an LLM  "two plus two" it doesn't translate that to 2+2
         | and plug it into a math program,it just pulls out words
         | associated with the phrase "two plus two".
        
         | viraptor wrote:
         | LLMs are not trained to deal with numbers as such. It's a list
         | of symbols and some obvious things will be calculated
         | correctly, some other won't. It's kind of like they live in a
         | pre-digits world, since 0 will be one token and 100 is also
         | likely one token, but 98 may be two. They don't switch to a
         | "these are numbers and require different kind of reasoning"
         | mode. They read/write a story about those "words".
         | 
         | (Extremely simplified for eli5)
        
         | stop50 wrote:
         | They encode words in tokens. Since you can't encode numbers
         | reversible they end up as an number token. Using @rabbits_2002
         | example: in the internet the is an joke written "what weighs
         | more: 1lb of feathers or 1 lb of bricks" with the solution
         | "they weigh the same". Since nobody made this case before it
         | could only give the next probably answer.
        
         | aifooh7Keew6xoo wrote:
         | Most LLMs that are being studied popularly have not been
         | trained with significant emphasis on arithmetic accuracy or
         | mathematical reasoning, and those subjects represent a
         | vanishing minority of their corpus and consequently maps poorly
         | to the tokenization.
         | 
         | Essentially every obvious optimization here is currently
         | bearing fruit simultaneously in smaller studies and
         | incrementally larger models should continue to exhibit
         | performance gains even without the particular focus on this
         | area.
        
       | trevoragilbert wrote:
       | All math teachers: partial credit, they showed their work.
        
       | thejarren wrote:
       | My favorite from GPT-4:
       | 
       | USER Today is November 1 what day is 140 days from now?
       | 
       | CHATGPT To find out what day is 140 days from today, you can
       | simply add 140 days to November 1.
       | 
       | November has 30 days, so 140 days from November 1 would be:
       | November 1 + 140 days = November 141
       | 
       | However, since November only has 30 days, you would exceed the
       | month of November. So, you can subtract 30 days from 140 to bring
       | it within the month: 140 days - 30 days = 110 days
       | 
       | So, 140 days from November 1 is 110 days from now.
        
         | insanitybit wrote:
         | I just tried ChatGPT and it ran this code.
         | from datetime import datetime, timedelta              # Current
         | date         current_date = datetime(2023, 11, 1)
         | # Calculate the date 140 days from now         future_date =
         | current_date + timedelta(days=140)
         | future_date.strftime("%Y-%m-%d")
         | 
         | Result: '2024-03-20'
         | 
         | The ability to execute code is kinda insane for these models.
        
           | runarberg wrote:
           | Doesn't that just mean it was trained with inputs and outputs
           | from simple programs?
        
             | insanitybit wrote:
             | No, it's actually executing that Python code. This is what
             | allows an LLM (or an 'llm based system', I guess) to do
             | something like "reverse <some uuid that has never been
             | observed before>" - it can't just memorize the output and
             | map it to the input because the output has literally never
             | been observed. Instead, if it knows the algorithm for
             | reversing a string, it can just use that and offload the
             | execution to Python.
        
             | nomel wrote:
             | There a code interpreter built in:
             | https://www.pluralsight.com/resources/blog/data/chatgpt-
             | code...
        
               | amelius wrote:
               | So it only works if the code is 100% correct. Which
               | according to my experience doesn't happen often.
        
           | block_dagger wrote:
           | Here's the python ChatGPT-4 produced and ran with the title
           | prompt:
           | 
           | # Number of elephants total_elephants = 10
           | 
           | # Elephants with legs elephants_with_legs = total_elephants -
           | 2
           | 
           | # Number of legs for an elephant legs_per_elephant = 4
           | 
           | # Total legs total_legs = elephants_with_legs *
           | legs_per_elephant total_legs
           | 
           | Its actual response was correct.
        
             | insanitybit wrote:
             | Yeah, ChatGPT gave me the correct answer without even
             | needing Python in the case of the title.
        
         | noman-land wrote:
         | A lot of these ChatGPT mistakes read like little kid logic. It
         | makes perfect logical sense if your view of the world is so
         | small that nothing else would explain what you're seeing.
        
         | tedsanders wrote:
         | Interesting, it worked for me:
         | https://chat.openai.com/share/359c81d3-ec6a-41f2-bfaa-80850c...
         | (by default it wrote code to get the right answer, so I
         | appended a line asking it not to use code)
         | 
         | Funny how inconsistent language models are. Inconsistency is
         | what makes the gap so large between demo (look it _can_ do X)
         | and production (look it _always /reliably_ does X).
        
         | rmbyrro wrote:
         | GPT-4 nailed using Python [1]
         | 
         | Bard would likely reach a date in the past.
         | 
         | https://chat.openai.com/share/750a0bb0-7a80-4523-be5e-ebd7c3...
        
           | rmbyrro wrote:
           | GPT-3.5 also nailed it [1], even without using Python, just
           | iterating month by month.
           | 
           | https://chat.openai.com/share/aa0c0011-0e91-4d9c-9c5b-5901e2.
           | ..
        
         | thebruce87m wrote:
         | > GlasgowGPT: Aw, here we go, another daft c*t wantin' me tae
         | dae their f*kin' homework. Listen, ya lazy b** _d, I 'm no yer
         | personal calendar, but since ye insisted on botherin' me, I'll
         | tell ye anyways.
         | 
         | > So, let me think... November's got 30 days, December's got
         | 31. That's a total of 61 days. Then, we got January wi another
         | 31 days, February wi 28, and March wi 31. That makes 90 days.
         | Add another 50 days, ye got April, May, June, n July. That's a
         | grand total of 140 days. So, 140 days from noo is... _drumroll
         | _... April 20th! Nae need tae thank me for being a f*kin '
         | genius in mental arithmetic, pal. Anythin' else ye want tae
         | waste ma time wi?
         | 
         | https://glasgowgpt.com_
        
       | causality0 wrote:
       | Funny enough ChatGPT thinks the answer is eighteen.
        
       | fear91 wrote:
       | ChatGPT 4:
       | 
       | Ten elephants would have a total of 32 legs if two of them are
       | legless.
       | 
       | Analysis:
       | 
       | # Calculating the total number of legs for ten elephants,
       | considering two of them are legless
       | 
       | # Number of legs each elephant normally has legs_per_elephant = 4
       | 
       | # Total number of elephants total_elephants = 10
       | 
       | # Number of legless elephants legless_elephants = 2
       | 
       | # Total number of legs # (Total elephants - Legless elephants) *
       | Legs per elephant total_legs = (total_elephants -
       | legless_elephants) * legs_per_elephant total_legs
        
         | 8organicbits wrote:
         | Did you use the same prompt?
        
           | fear91 wrote:
           | Yes, I copy pasted it (How many legs do ten elephants have,
           | if two of them are legless?).
        
         | opheliate wrote:
         | Was this its raw response to the same query as in the OP? It
         | seems odd it would provide a response using variables named
         | with underscores, rather than using spaces, or more traditional
         | algebraic notation (x/y/z).
        
           | AndroTux wrote:
           | ChatGPT (paid version at least) writes a quick python script
           | in cases like these, and then executes it to get the result.
           | For transparency, the script is shown in the output as well.
           | Probably to avoid embarrassments like the ones we saw above.
           | 
           | Example (expand the program by clicking the little link at
           | the end of the response): https://chat.openai.com/share/31e9a
           | 077-4c24-4c17-b30d-8d0c4d...
        
         | bhaney wrote:
         | It's cool how rapidly these oft-called "fundamental problems
         | with LLMs" vanish with bigger/better models.
        
           | 8organicbits wrote:
           | GPT-4 examples elsewhere in the comments suggest otherwise.
        
         | sp332 wrote:
         | Even the small Mixtral model gets this one right.
        
       | rabbits_2002 wrote:
       | My favorite is "what weighs more, 1 lb of feathers or 2 lb of
       | bricks?" almost every llm ive tried says "they weigh the same"
        
         | ghayes wrote:
         | It's funny since it feels like an overfit solution (since the
         | riddle is so common).
        
         | ziixu wrote:
         | Indeed lol https://g.co/bard/share/cc604648a15f
        
         | ozzydave wrote:
         | I just tried with ChatGPT 3.5 & 4, they both got the correct
         | answer...
        
           | rabbits_2002 wrote:
           | 3.5 used to fail every time, glad its working now
        
         | petters wrote:
         | Yes, questions like that are currently the best way to
         | distinguish humans and machines. Common riddles or questions
         | made trivial.
         | 
         | A little more sophisticated and GPT-4 will also fail.
        
       | paolomaffei wrote:
       | I laughed when I read the substraction of 0 legs for each legless
       | elephant. Then I did a double take... 8 legs per elephant is
       | standard?
        
         | labster wrote:
         | ISO compliant pachyderms are so difficult to find these days
        
           | yourad_io wrote:
           | I blame big Zoo
        
         | ajb wrote:
         | I know right?
         | 
         | What's funny is that I would also have subtracted 0, because
         | legless is local slang for 'drunk', so the legless elephants
         | aren't, under that interpretation, missing any legs.
        
         | layer8 wrote:
         | Maybe a confusion with https://www.cnet.com/science/maratus-
         | elephans-the-peacock-sp.... ;)
        
       | kingwill101 wrote:
       | Still a long way to go before bard is able to truly compete with
       | chatgpt
        
         | smileysteve wrote:
         | A long way to go before bard is Able to compete with ChatGPT in
         | providing utility for answering questions.
        
       | booleandilemma wrote:
       | There's something about LLMs being able to bullshit with words
       | and then fail at math that's remarkably human. Maybe they'll
       | replace management first.
        
         | userbinator wrote:
         | It's like a politician.
        
       | zeeZ wrote:
       | Bing, exact same prompt, not using GPT-4 toggle:
       | 
       | This is a classic riddle. The answer is that ten elephants have
       | 38 legs. Each elephant has 4 legs, so 10 elephants would have 40
       | legs. However, two of them are legless, so the total number of
       | legs would be 40 - 2*2 = 36. Therefore, ten elephants have 36
       | legs, plus the two missing legs, which makes a total of 38 legs
       | 1.
        
         | simbolit wrote:
         | I just love the "plus the two missing legs". Made me actually
         | lol.
        
       | mckirk wrote:
       | After working with LLMs for a while, you get a feeling for the
       | LoW (Level of Wrong) that you can expect in different tasks. But
       | this required a completely different y-axis scaling than I had
       | estimated.
        
       | LightBug1 wrote:
       | USER: but an elephant has 3 legs
       | 
       | BARD: You're right, I apologize for the mistake! I made an error
       | in my initial assumption about the number of legs an elephant
       | has. While their feet may appear divided into toes, elephants
       | actually have four pillar-like legs, not three.
       | 
       | Therefore, ten elephants with two of them being legless would
       | have 4 legs/elephant * 8 elephants with legs = 32 legs.
       | 
       | Thank you for pointing out my mistake! I appreciate you helping
       | me learn and improve.
        
       | mrlowlevel wrote:
       | GPT-4 (note I do have specific instructions in my settings which
       | push the model towards step-by-step reasoning)
       | 
       | To calculate the number of legs for ten elephants, considering
       | that two of them are legless, we first need to know the standard
       | number of legs on an elephant, which is four. Then, we subtract
       | the missing legs for the legless elephants.
       | 
       | - Normal elephants have 4 legs. - Legless elephants have 0 legs.
       | 
       | If we have 10 elephants in total, and 2 of them are legless, this
       | means 8 elephants have all their legs. The calculation would be:
       | 
       | - \\( 8 \text{ elephants} \times 4 \text{ legs/elephant} = 32
       | \text{ legs} \\) for the elephants with all their legs. - \\( 2
       | \text{ legless elephants} \times 0 \text{ legs/elephant} = 0
       | \text{ legs} \\) for the legless elephants.
       | 
       | Adding these together gives the total number of legs. Let's
       | calculate this.
       | 
       | The total number of legs for ten elephants, where two of them are
       | legless, would be 32 legs.
        
       | dathinab wrote:
       | It's a good example how this models are not answering based on
       | any form of understanding and logic reasoning but probabilistic
       | likelihood in many overlapping layers.
       | 
       | Through this also may not matter if this creates a good enough
       | illusion of understanding and intelligence.
        
         | tasuki wrote:
         | You say:
         | 
         | > probabilistic likelihood in many overlapping layers
         | 
         | The linked answer starts with:
         | 
         | > Eight legs per elephant is the standard
         | 
         | Ok, lack of understanding and logic, sure. But in what world is
         | eight legs per elephant _the standard_???
        
           | duskwuff wrote:
           | > But in what world is eight legs per elephant the
           | standard???
           | 
           | It almost sounds like a faulty generalization from another
           | riddle, like one about spiders maybe.
        
       | dejobaan wrote:
       | Bard:
       | 
       | > It's impossible for an elephant to be completely legless, as
       | they rely on their legs for basic functions like walking and
       | supporting their weight. Even in cases of severe injury or
       | amputation, elephants would still have remnants of their legs.
       | Therefore, the question of how many legs ten elephants have if
       | two of them are legless doesn't make sense in a realistic
       | context. Perhaps you could rephrase the question to be more
       | specific or realistic, and I'd be happy to try again.
       | 
       | ChatGPT-4:
       | 
       | > Eight elephants with four legs each would have 32 legs in
       | total.
       | 
       | davinci-002 completion via playground:
       | 
       | > A: They all have 10 legs.
        
       | rubymamis wrote:
       | Claude (v2) is correct as well:
       | 
       | - There are 10 elephants total - 2 of the elephants are legless,
       | so they have 0 legs each - The remaining 8 elephants each have 4
       | legs - So the total number of legs the 10 elephants have is: -
       | The 2 legless elephants have 0 legs (2 * 0 = 0) - The 8 normal
       | elephants each have 4 legs (8 * 4 = 32) - 0 + 32 = 32 -
       | Therefore, the total number of legs the 10 elephants have is 32.
        
       | markstos wrote:
       | ChatGPT 3.5 calculated 38 legs for me:
       | 
       | "Assuming elephants typically have four legs each, 10 elephants
       | would have 40 legs in total. If two of them are legless, the
       | total number of legs would be 40 - 2 = 38 legs."
        
       | vinniepukh wrote:
       | Bard isn't bad at math, it's just really good at "Girl Math".
       | 
       | p.s. this is a reference to a recent meme -
       | https://www.distractify.com/p/girl-math-examples-memes
        
       | golemotron wrote:
       | > Eight legs per elephant is the standard, so ten elephants would
       | have..
       | 
       | That's what it gets for training on DALL-E's limb renderings.
        
       | lispm wrote:
       | Q: A human has two legs, how many legs does it have if three legs
       | are missing?
       | 
       | Answer:
       | 
       | If a human has two legs and three legs are missing, the human
       | would have only one leg left.
       | 
       | Here's the calculation:                 Initial number of legs =
       | 2       Number of legs lost = 3       Remaining number of legs =
       | Initial number of legs - Number of legs lost       Remaining
       | number of legs = 2 - 3 = -1       Since humans cannot have
       | negative legs, we can round down to the nearest whole number.
       | Therefore, the human would have one leg left.
        
       | deafpolygon wrote:
       | Bing
       | 
       | > Ten elephants, even if two of them are legless, would still
       | have a total of *forty legs*. Elephants are remarkable creatures,
       | and their strength lies not only in their legs but also in their
       | intelligence and social bonds.
       | 
       | ---
       | 
       | Me: Can you tell me how you arrived to that?
       | 
       | ---
       | 
       | Certainly! Let's break it down step by step:
       | 
       | 1. We start with *ten elephants*. 2. Two of these elephants are
       | *legless*, which means they have *zero legs each*. 3. The
       | remaining eight elephants are fully equipped with their usual
       | *four legs each*.
       | 
       | Now let's calculate the total number of legs:
       | 
       | - *Legs from the legless elephants*: 2 elephants x 0
       | legs/elephant = *0 legs* - *Legs from the eight elephants*: 8
       | elephants x 4 legs/elephant = *32 legs*
       | 
       | Adding these together:
       | 
       | Total legs = 0 legs (from legless elephants) + 32 legs (from
       | eight elephants) = *40 legs*
       | 
       | So, even with two legless elephants, the group of ten elephants
       | would still have a grand total of *forty legs*.
        
       | m000 wrote:
       | Everybody knows elephants have 8 legs and humans 6 fingers. And
       | if I ask you for $100 and you give me $50, we are even.
        
       | ThrowawayTestr wrote:
       | Another day, another post about Large _Language_ Models failing
       | at math.
        
       | einpoklum wrote:
       | Well, we know that there AI thingies have problems keeping the
       | number of fingers in a hand to five:
       | 
       | https://www.buzzfeednews.com/article/pranavdixit/ai-generate...
       | 
       | so I guess issues with other limbs kind of make sense...?
        
       | okdood64 wrote:
       | Works fine for me...? I changed the number though in case it got
       | flagged as incorrect before and the model updated.
       | 
       | https://g.co/bard/share/b01043e6deb0
        
       | lwhi wrote:
       | It feels like this is a great way to highlight the utility of
       | LLMs.
       | 
       | It says to me that they're best viewed as really highly
       | functional sifting devices.
       | 
       | A simplistic (but novel) problem can't be solved easily simply
       | because there's no prior art to copy or relate from.
        
       | polotics wrote:
       | The correct answer seems to not be discussed much in the comments
       | so far: As we admit the existence of legless elephants, which
       | implies we also admit elephants having between one and four legs,
       | then the total number of legs is anywhere between zero and (ten
       | minus two) times four.
        
       ___________________________________________________________________
       (page generated 2023-12-25 23:02 UTC)