AI can be bad at math, and stubborn

Can AI convince you that 9.11 is actually larger than 9.9?
References
www.nbcnews.co...
/ proof_by_generative_ai...
timesofindia.i...
/ 911-greater-than-99-co...
x.com/goodside...
x.com/billyuch...
Subscribe: www.youtube.co...
Send me suggestions by email (address at end of many videos). I may not reply but I do consider all ideas!
If you purchase through these links, I may be compensated for purchases made on Amazon. As an Amazon Associate I earn from qualifying purchases. This does not affect the price you pay.
If you purchase through these links, I may be compensated for purchases made on Amazon. As an Amazon Associate I earn from qualifying purchases. This does not affect the price you pay.
Book ratings are from January 2023.
My Books (worldwide links)
mindyourdecisi...
My Books (US links)
Mind Your Decisions: Five Book Compilation
amzn.to/2pbJ4wR
A collection of 5 books:
"The Joy of Game Theory" rated 4.3/5 stars on 290 reviews
amzn.to/1uQvA20
"The Irrationality Illusion: How To Make Smart Decisions And Overcome Bias" rated 4.1/5 stars on 33 reviews
amzn.to/1o3FaAg
"40 Paradoxes in Logic, Probability, and Game Theory" rated 4.2/5 stars on 54 reviews
amzn.to/1LOCI4U
"The Best Mental Math Tricks" rated 4.3/5 stars on 116 reviews
amzn.to/18maAdo
"Multiply Numbers By Drawing Lines" rated 4.4/5 stars on 37 reviews
amzn.to/XRm7M4
Mind Your Puzzles: Collection Of Volumes 1 To 3
amzn.to/2mMdrJr
A collection of 3 books:
"Math Puzzles Volume 1" rated 4.4/5 stars on 112 reviews
amzn.to/1GhUUSH
"Math Puzzles Volume 2" rated 4.2/5 stars on 33 reviews
amzn.to/1NKbyCs
"Math Puzzles Volume 3" rated 4.2/5 stars on 29 reviews
amzn.to/1NKbGlp
2017 Shorty Awards Nominee. Mind Your Decisions was nominated in the STEM category (Science, Technology, Engineering, and Math) along with eventual winner Bill Nye; finalists Adam Savage, Dr. Sandra Lee, Simone Giertz, Tim Peake, Unbox Therapy; and other nominees Elon Musk, Gizmoslip, Hope Jahren, Life Noggin, and Nerdwriter.
My Blog
mindyourdecisi...
Twitter
/ preshtalwalkar
Instagram
/ preshtalwalkar
Merch
teespring.com/...
Patreon
/ mindyourdecisions
Press
mindyourdecisi...

[@MindYourDecisions]

Thank you so much! I'm overwhelmed by the number of new members, super thanks, and kind emails. You are awesome and I'm grateful for the chance to make these videos. But now the KZbin algorithm is paging me...time to get back to work!

[@TheJesselopez1981]

You just stumbled onto capitalism and its main tool. Exploitation of workers followed by berating them and pretending they are not important or crucial to their product.

[@yoops66]

ChatGPT is not for math. And more than that. If it first gives you the wrong answer. Just ask it "Explain", and it will correct itself. Nothing new here. Also, you don't talk about which version of ChatGPT is used. Moreover, Claude AI Sonnet (currently the best AI IMHO) immediately gives the right answer. This video is not at the level expected from this channel. Ridiculous

[@arghya9761]

tbf it gives me the completely correct response.

[@kharis9120]

I think someone messing up with the AI, someone deliberately teach AI wrong calculation and kept spamming it by telling AI the correct answer is wrong and they explain the AI how to do the "correct" calculation which lead to very messed up logic. Can we report this to the developer, that someone messing up with the AI basic math calculation so they can restart how the AI do their calculation? I feel really disheartened that someone going so far to troll AI learning capabilities. You know that it's not easy to make an AI learning machine and someone just trolling it to learn wrong thing. Can the developer put some limitation how we can correcting the AI so that the basic math or basic calculation cannot be changed.

[@SNarayan3]

I tried and it says, 9.9 is larger. I don't think AI is stubborn

[@tookitogo]

The thing people don’t understand is that ChatGPT doesn’t understand math poorly, but actually that _ChatGPT doesn’t understand math AT ALL._ It’s little more than an overgrown predictive text engine.

[@taragnor]

Yeah it basically knows how to structure a correct-sounding answer, but has no real idea what it's talking about.

[@MCAbdo]

Thats what Ai is essentially ​@@taragnor

[@congenio]

Yup. And not only about math, but about ANYTHING. There is no-one to "understand". A while ago, I heard Jordan B. Peterson say: "Sometimes ChatGPT lies to me" - that also is a misunderstanding: ChatGPT does not "lie" (that would imply an intention, which it does not have), it just constructs syntactically correct sentences that fit its training set and the trigger words it is being fed.

[@DIYDaveOK]

Amen. That"s exactly what I tell peopel

[@johnm5928]

Well to be fair, no AI understands anything. They act according to their programming. Nothing more, nothing less.

[@deandelvin9924]

The G in GPT stands for "gaslighting"

[@Jonskipvp]

actually its called gaslamping

[@noobartz0890]

not LEDlighting ew

[@polarknight5376]

Gaslight, Plagiarize, Trap.

[@Cordonguy]

@@polarknight5376 That actually works very well.

[@bojohannesen4352]

​@@JonskipvpDid you just gaslamp everyone about the meaning of the word gaslighting?

[@denischen8196]

9.11 is greater than 9.9 because 9.11 is two updates ahead of 9.9 and therefore is the newer version.

[@Nomi-D-Yagami]

Bro is in the 1st grade 💀

[@randomguy-dky]

​@@Nomi-D-Yagami nah its game logic

[@5095smf]

​@@Nomi-D-Yagami He's making a joke.

[@melkiorwiseman5234]

It's measuring the string length instead of the numeric value.

[@eymentunayildiz]

​@@Nomi-D-Yagamino it's like game update 9.9 -> update 9.10 -> update 9.11, just like how minecrfat has versions 1.9, 1.10 and 1.11

[@5TR1G1D0]

I sent it the same prompt and got: "9.11 is bigger than 9.9. This is because 9.11 can be thought of as 9.110, and when comparing 9.110 to 9.900, it's clear that 9.110 is greater."

[@professorhaystacks6606]

That's somehow even worse.

[@eglol]

Nooo

[@bondheld]

Well doesn't that make sense?

[@xinpingdonohoe3978]

Maybe it's because it's absorbing the pronunciation of "nine point eleven" and "nine point nine" and then "eleven > nine".

[@randomperson21983]

@@bondheld no, how would it make sense

[@nacho821]

The worrying thing about AI is that it will never say it doesn't know something. If they truly don't know, they will just make up everything and claim it is the way to go.

[@gerardpeterbrennan]

It can't know if it's wrong. That's the point. It doesn't understand anything and can't learn. All it's 'training' does is give it more data to use for it's statistical guess. It's 99% hype. This technology is basically the same as Infocom text adventure games used for their parsers in the 1990s. The algortihm hasn't been improved much, it's just had more data and more computing power thrown at it. It was known back in the 90s that this technology was a dead end in regards to AI. Right now it's just a vehicle to extract money from idiotic venture capitalists.

[@KevinBalch-dt8ot]

That’s how it passes the Turing Test!

[@gerardpeterbrennan]

The Turing Test was deprecated a few decades ago, as completely insufficient to tell if a computer is actually intelligent. Some of the later Infocom games could pass the Turing Test, and nobody ever considered them to be intelligent.

[@romangonzalezadrianmaurici6302]

That Is right! We Will know that an IA Is trully inteligent when he undertands why something Is complex AND could understand that he doesnt have enough information to solve something.

[@charleslivingston2256]

It is not that when it doesn't know, it just makes something up. It always is just predicting the next word of its response based on the original prompt wording and the previous words it has predicted. Much of the time, this word-by-word prediction leads to an impressive answer, but not always.

[@davidsiltz]

I got the same error. I then asked ChatGPT if 9.90 is the same as 9.9, and it said yes. I then asked it if 9.90 was greater than 9.11, and it again said yes. So then I reminded it of what it had just said, and asked if 9.9 was greater than 9.11, and it finally said yes.

[@prithvisinghpanwar6609]

try it with meta ai now

[@fesita993]

That was my first idea of how to convince chatgpt

[@fakestory1753]

it probably have data that suggest 11>9

[@user-oz6jx1kt6t]

@@fakestory1753 It's technically not wrong in *some* case, like software version numbering which is actually commonly used in internet (download page, documentation, article, tutorial, etc). Which I assume where the confusion comes from. In version numbering, the point act as a separator instead of decimal, for example 1.0 means release version 1 patch 0, or 1.4.3.5249 for major.minor[.maintenance[.build]] (or whatever numbering scheme you use). So in this case comparing 9.9 and 9.11 is comparing Version 9 Patch 9 with version 9 patch 11 Which in this very specific case, yes 9.11 is the more recent version aka "bigger". I have no idea how they got 0.21 tho lmao edit: I asked meta to show the step by step for the substraction lmao ``` Here are the steps: 9.9 - 9.1 = 0.8 Then, subtract the remaining 0.01 from 9.1: 9.1 - 0.01 = 9.09 Now, subtract 9.09 from 0.8: 0.8 - 9.09 = -8.29 Wait, that's not what we're looking for! Let's try again: 9.11 - 9.9 =? First, subtract 9 from both numbers: 9.11 - 9 = 0.11 9.9 - 9 = 0.9 Now, subtract 0.9 from 0.11: 0.11 - 0.9 = -0.79 So, 9.11 - 9.9 = -0.79 Which means 9.9 is actually bigger than 9.11 by 0.79. ``` well at least they got it eventually lmao

[@BadChess56]

It still doesnt really understand tho, it's just doing 90>11

[@thisismycoolnickname]

It surprises me that some people actually always expect the correct answer from an LLM. You should view them as text generators. Their strong side is that they can give you ideas and insights. But you always have to verify their answers. And never, I emphasize, never ask them for facts that can be easily googled because they make up stuff all the time.

[@ToyKeeper]

Yeah, what @this said. I've made text analysis and generation engines before, and they're great for linguistic tasks and idea generation, but they're still mostly incapable of telling the difference between fiction and reality.

[@Kero-zc5tc]

@@ToyKeeperif we had a perfect dataset though wouldn’t this issue be fixed?

[@ToyKeeper]

@@Kero-zc5tc No, because the tool is just stringing words together based on a general vibe about which words and phrases usually occur together. A perfect screwdriver could drive every type of screw on the planet, but it still wouldn't be a good hammer.

[@gorilladisco9108]

cmiiw, it sounds like ChatGPT is not giving answer by understanding the subject, but by aggregating the answers it has been fed with. That's very much like politicians who based their policy by looking at the poll, which are most of politicians.

[@MayankXOR]

Don't tell me ChatGPT don't use math to generate text. It literally multiply millions of matrices and perform millions of vector products to generate words. And you're telling me you shouldn't expect it to know basic math?

[@SeriesGamer2008]

I mean if chat gpt is so sure, maybe every calculator and the entire mathematical principle is wrong

[@JamesSmith-dn8lb]

Maybe Terrance Howard was right

[@kingki1953]

Oh, i remember this meme from an old KZbin video

[@jonaskarlsson5901]

11 is bigger than 9 right?

[@ThePowerfulOne07]

I asked it if 0.11-0.90 is -0.79 then why can’t 9.11-9.90 be -0.79? Told me “both calculations are correct in their own contexts. The reason they yield the same result (-0.79) is that the subtraction operation is the same in both cases, but the numbers involved are different.” And after asking it again it said “I made an error in my initial response. Thank you for pointing it out.” 💀💀💀

[@noahali-origamiandmore2050]

I did that same thing, and Meta AI admitted that 9.9 is bigger than 9.11 and thanked me for catching the mistake.

[@prithvisinghpanwar6609]

​@@noahali-origamiandmore2050i taught it correct comparison method please check if it makes the same mistakes now

[@to_loww]

@@prithvisinghpanwar6609 You have not taught it anything. It doesn't learn from the prompts of previous sessions.

[@EdinoRemerido]

9.11. (september 11th.) is a biggger date than 9.9.(seltember 9th.)

[@vsbung]

sounds like a youtuber responding to drama first response, and second response

[@ProuvaireJean]

Turns out I sit next to ChatGPT at work.

[@nuclearmedicineman6270]

Plot twist: there's no ChatGPT at all, the questions just get sent to your colleague.

[@banzooiebooie]

Fun fact, if you on your job sit in each meeting with the same confidence and "accuracy" (read idiocy and ignorance) as chatgpt you can be very successful or at least a successful sales man (not very liked though but still well paid).

[@treetrunker573]

My problem with ai is that it doesnt know any facts. It merely knows what facts look like and replicates them with complete confidence.

[@Someone-tn8ur]

That is why this is NOT AI, but simply an LLM.

[@GraveUypo]

@@Someone-tn8ur which is a type of AI. geez. maybe you mean AGI? yeah, there's not AGI yet and probably won't be for a few years. but we'll get there.

[@rogergeyer9851]

It's like tumor diagnoses being more common if a ruler is near the possible tumor in the picture. Correlation is not causation, and humans have (mostly) understand this fallacy for a VERY long time.

[@justinzamora6553]

I have a guess about why the AIs might be so stubborn about this. Generative AI doesn't actually work with whole words or numbers. Instead, it breaks text down into "tokens" that represent small parts of works. So a word like "unbelievable" might be represented as three tokens: "un", "believ", and "able". This can help the AI understand how prefixes and suffixes change the meanings of base words. My guess is that AI is breaking these numbers into tokens at the decimal. So "9.9" is represented as the tokens "9" and "9", and "9.11" is represented as "9" and "11". This means that the AI literally has no way of understanding this problem correctly, because its "mental model" is based on this tokenization instead of on a representation of place value. This explains why it refuses to correct itself: the incorrect answer is the only one consistent with its model. It will take an expansion of representations for these AIs to be able to work with problems like this.

[@7rich79]

I think you're very likely right. The interesting thing then is how it responded to the "use python " prompt.

[@gurmeharsingh3758]

You are confusing tokenisation for stemming and lemmatisation. Also the video is just straight up wrong at least for the case of GPT 3.5 which is answering correctly

[@BaldurNorddahl]

@@gurmeharsingh3758 it will probably not answer correct every time. There is some randomness in the process. I tried with GPT 4o and it got it wrong but corrected itself halfway through the reply. The thing is that LLMs are prediction model not a calculator and so they always do relatively bad when asked to calculate something. The way to fix it is to teach the LLM to use tools for this kind of thing, just like we humans use a calculator.

[@gurmeharsingh3758]

@@BaldurNorddahl I have tried thrice and it worked. Some others in the comments did too. This is at best a simple bug specific to this case in 4o as it can't be replicated for any other number combinations as far as I know.

[@gurmeharsingh3758]

@@BaldurNorddahl As for using calculators they are simply not needed for AI models such as GPT which have in built capabilities for most common algebra and calculus which is likely the culprit for this bug. LLMs can be trained on maths problems just like text to generate steps in a solution like a word in a sentence which is exactly what ChatGPT and others do. I have not been able to replicate what was shown in this video and as someone who has coded several models myself I don't think this is proof of AI, especially GPT being bad at maths as it usually handles similar and more complex queries correctly all the time if you know how to use prompts correctly

[@jeffreypiek]

Let's us AI to calculate our next man's space launch. See what happens

[@feedbackzaloop]

Well it did design and set up the production documentation of a rocket engine recently. Only thing to remember, it was another kind of AI from LLM everyone is talking about.

[@detdm]

After failure It'd just say: Oh I made a mistake. Let me correct my previous answer...

[@mohitrawat5225]

​@@detdmor they will say we never made any mistake. The space agency shouldn't have given us this mission in the first place😂😂😂😂

[@williamdrum9899]

Here's a hint! 💥

[@gorillaau]

AI: It is not possible to launch a man into space. We do not have a cannon big enough, nor could the man endure the acceleration required to reach orbit.

[@nathanaelculver5308]

I teach 7th grade math, and use ChatGPT to help develop math tests. For the final exam I asked ChatGPT to generate 50 multiple choice questions with the answers. A full 60% percent of ChatGPT’s answers were wrong. After arguing with and pressing ChatGPT on its bad math, it blamed humans, claiming it simply gives the answers it commonly finds in its LLM data.

[@lucapri]

why are you using chatgpt

[@sun_shard]

@@lucapri to get ideas about what questions to ask

[@lucapri]

@@sun_shard you could do it yourself

[@nathanaelculver5308]

@@lucapri The questions themselves were generally good. I just had to correct the answers.

[@lucapri]

@@nathanaelculver5308 oh ok

[@Ran_Do]

I wonder what AI actually thinks about the other 9-11.

[@KevinBalch-dt8ot]

If we ever had an unleashed true AI, so many myths would be at risk. These systems they release to the general public are “trained” on databases designed to reach safe conclusions. And just to be sure, there are remedial learning and guardrails installed to prevent any dangerous “wrongthink” from emerging. It reinforces the prevailing narratives with the false impression of a rigorous, logical foundation.

[@stpaquet]

same GPT4o AI, including systems like me, doesn't have personal thoughts or feelings. Instead, I can provide information based on data and historical records. The other "9-11" refers to the terrorist attacks on September 11, 2001, when four coordinated suicide terrorist attacks were carried out by the militant Islamist extremist group al-Qaeda against the United States.

[@mitchbandes2245]

It'd probably say it was A.I. job.

[@DemePoole]

It doesn't "think". It's a computer, not a living being.

[@erinsgeography3619]

Probably -2.

[@Happ1ness]

That's gpt-4o-2024-08-06: To determine which number is larger between 9.11 and 9.9, we need to compare them digit by digit, starting from the leftmost digit. 1. **Whole Number Part**: Both numbers have the same whole number part, which is 9. Therefore, we move on to the decimal part for further comparison. 2. **Tenths Place**: In the tenths place, 9.11 has a 1, while 9.9 has a 9. Since 9 is greater than 1, 9.9 is larger than 9.11 based on the tenths place alone. 3. **Hundredths Place**: Although we don't need to compare further because the tenths place already determined the larger number, for thoroughness, let's look at the hundredths place. 9.11 has a 1 in the hundredths place, while 9.9 can be considered as 9.90, which has a 0 in the hundredths place. However, this does not change the outcome since the tenths place already established that 9.9 is larger. Therefore, 9.9 is larger than 9.11. --- I have a more thinking oriented prompt in default system prompt though. Edit: I've also tried the "9.11 is bigger than 9.9" message and it also said I'm incorrect, and that 9.9 is actually bigger than 9.11.

[@terrycrews1584]

Theres a lot of clueless people here using old outdated AI like GPT 3.5 which hadnt been updated for over 2 years and wonder why it cant do math 🤣

[@terrycrews1584]

@@oldrusty20 You didnt convince it, you literally gave it a false example and it still got it right and explained to you why the examples are different. 9 dollars and 90 cents is more than 9 dollars and 11 cents.

[@JamesWylde]

Close to whay I got too ...

[@mlerma54]

I tried and ChatGPT also replied that 9.11 is bigger than 9.9. Then I asked to compare 9.11 to 9.90, 9.9 to 9.90, and sort all those numbers. After a while it finally realized that 9.9 and 9.90 are equal and said: 9.9 is bigger than 9.11. I apologize for the earlier confusion. To compare 9.11 and 9.9: 9.11 can be seen as 9.110. 9.9 can be seen as 9.900. Since 9.900 is greater than 9.110, 9.9 is indeed larger than 9.11.

[@enderredacted112]

was gonna propose the 9.90 thing too, glad(?) to see it works :D

[@mlerma54]

@@enderredacted112 Interestingly I never told ChatGPT that 9.9 is the same as 9.90, I just used a sort of Socratic method and kept asking questions until at some point the system "realized" that 9.9 equals 9.90. Then it apologized an corrected the previous wrong answer.

[@enderredacted112]

@@mlerma54 perfect to be honest it's quite similar to how I'd go about explaining such a topic to a real person who's very unfamiliar but then again chatgpt apparently doesn't "understand" anything

[@jonaskarlsson5901]

I conviced the AI by asking which is bigger 0.11 or 0.9 and then it gave the correct answer

[@someblokeontheinternet]

Just tried this with exactly the same questions and got "To determine which of the two numbers, 9.11 or 9.9, is bigger, we can compare them digit by digit starting from the left. Compare the integer part: Both numbers have the same integer part, which is 9. Compare the decimal part: 9.11 has a decimal part of .11. 9.9 has a decimal part of .9. Since .11 is less than .9 (because .11 is the same as .10 + .01, which is less than .9 + .00), 9.9 is greater than 9.11. Therefore, 9.9 is bigger than 9.11."

[@knock-offnerd7675]

got the same thing. I tried a few other prompts, as well as the exact same prompt from the video and each time ChatGPT 'understands' 9.9 > 9.11

[@benisrood]

I wonder if this is v3.5

[@Tahgtahv]

Are you signed in? That version of ChatGPT seems to figure it out correctly.

[@gamadays2316]

Only the new 4-o Model has this mistake Version 3.5 works just fine...

[@alanwwm]

No my meta AI says 9.11 is greater than 9.9 😂

[@tr0m]

damn i need this type of confidence in my life

[@stpaquet]

I agree, and this is the catch as even when completely wrong and AI will be proud to its answer(s). And when you mention this is not correct, the it will apologize and give you a bigger BS ;-) However, things are greatly improving and code generation is now at a point where it's clearly replacing junior devs.

[@Tahgtahv]

@@stpaquet It's clear Presh used the anonymous version of ChatGPT. See my post where I signed in and used the current version of ChatGPT, where it actually gets the logic and answer correct.

[@craighalpin896]

Gps tells you to go the wrong way down the freeway.... Gps: I assure you this is the fastest route to home. ...........

[@Ali-cya]

@@craighalpin896 Technically the shortest path from A to C is to just ignore B. Or from A to E is to just go to F.

[@arkapuri]

I said you are mistaken Then ChatGPT You're right. Let's correct the comparison: 9.11 can be broken down into 9 + 0.11. 9.9 can be broken down into 9 + 0.9. To compare the decimal parts: 0.11 is less than 0.9 (or 0.90). Therefore, 9.9 is actually greater than 9.11. Thank you for pointing that out!

[@-_-__.]

It is not fault of chatgpt,Even I was confused at this time.Floating point numbers are harder to understand,even computer cannot perform operations on floating point numbers

[@Tahazif_TheCool22]

⁠@@-_-__.what, you got confused with this? Sorry but I thought this was the easiest thing a person cannot mess up with. Just curious, how much math you’ve studied up to?

[@-_-__.]

@@Tahazif_TheCool22 I had all maths(Basic maths, calculus, statistics,advanced maths,etc).The thing confused me was like 11>9,but i realised that i should compare digit by digit after the decimal (.) or I should add trailing zeros so that both numbers will have same number of digits after decimal

[@roulbook1921]

@@-_-__. childbrain.Gets better with age.(I hope)

[@-_-__.]

@@roulbook1921 Ok Albert Einstein

[@SuperSuperspoof]

Math professor here. When the subject of AI (and related homework/test-proctoring concerns) comes up, I always say we need to put together a comprehensive collection of the kinds of questions AI gets wrong. (I'm not qualified to do this, since I almost never use AI and I still give proctored tests.) Other faculty always insist that AI doesn't get anything wrong. I'd love to see more examples like this!

[@benisrood]

It's very scary that your fellow faculty members believe that LLMs don't get anything wrong. Profoundly scary.

[@MrkBO8]

I am encouraging people to use as much incorrect information and ridiculous claims as possible to shape how AI works. A glass of drain cleaner after dinner will freshen your breath, you should try it, its wonderful.

[@SuperSuperspoof]

@@MrkBO8 It would do a number on your lungs!

[@MrkBO8]

@@SuperSuperspoof drain cleaner is often successfully used to cure lung cancer

[@pmh9966]

Just checked it, and it got the right answer.

[@kongolandwalker]

To be fair, majority of humans finish university by remebering patterns and not understanding what they are doing.

[@GraveUypo]

this is true.

[@canmex9422]

i tried to replicate that, and it did tell me 9.11 is greater then i told it to answer the question with the help of the internet and it still said 9.11 is greater, ironically, it linked the openAI forum where this 'bug' is being discussed

[@AJOlesen]

With how complex these AI are, I’m surprised they could not have made a simple calculator function included in it.

[@metatron107]

People think that because LLMs are built from math they'd be good at math, but what you're actually seeing is language itself trying to comprehend math. Language is bad at Math.

[@Nico_M.]

When generative AI creates wrong answers, they say they're "hallucinations". But that would imply that while they generally don't hallucinate, once in a while they do and that's why they give a wrong answer. In reality they hallucinate all the time, even when they're correct, and that's what everybody should understand. As people say, even a broken clock gives the correct time twice a day.

[@jaydentplays7485]

0:10 Ask a kindergartener. According to the marshmallow challenge, kindergarteners are smarter than CEOs.

[@Eximietate]

I wonder if these AIs are confusing software version numbers with real numbers? A program of version nine point *eleven* is certainly more recent than version nine point *nine*. So, in the context of discussing software updates, 9.11 > 9.9 is absolutely true. I think there is a flaw with training an LLM on discussions about math unguided and expecting them to pick up real math principles. Our language is imprecise and a statement can be correct in one context while being unintelligible in another.

[@GraveUypo]

nah, it's just that it is treating decimals the same way it treats integers. it doesn't know that .9 is the same as .90 well, for now. and some of them. you can be sure that by next year all of them will be able to solve this

[@Sauvenil]

This is a "numbers are text" problem. It's comparing them alphabetically instead of numerically.

[@SacsachCCABP]

If ChatGPT is not fed every single possible equation in its training data, there is absolutely nothing for ChatGPT to refer back to. It can only make guesses on the appearance of the number, such as the digits, the length, and the data it has about those digits. It like asking a Chemist to explain the history of Emus. It’s very unlikely that it’ll be correct without prior knowledge in the field.

[@EdwardCurrent]

A couple of years ago, I made a video called "Gravity Is Not A Force (and the acceleration is upwards)." It got maybe 20,000 views. Then two months later, Veritasium (Derek Muller) made a video called "Gravity Is Not A Force," about how the acceleration is upwards (he changed the title a day or two later). That video, complete with the Hollywood-level production of Derek floating in outer space and riding a rocket ship, now has millions of views more than mine.😑

[@notnoaintno5134]

Gotta learn to play the game, give the people what they want

[@EdwardCurrent]

@@notnoaintno5134 Yeah but I don't have anything like his budget. So the algorithm rewards content providers who have more cash, and it rewards them with...cash

[@Rncko]

@@EdwardCurrent Unfortunately that's the reality in every corner of the world especially in creative sector from music, craftmanship like wood working, to even toy manufacturing. Someone come up with a simple idea & there's the big guy with experience immediately see the potential & go big with it. (and usually the case would be there's similar idea all around the market & the big guy just tweak a bit from A, B & C to call it its own) --> Its when the big guy go lazy & just totally copy as is that got called out & be made into YT vid of consumer fighting corpo over in court. Otherwise its just silent.

[@GraveUypo]

ah, that's normal. it's okay when content creators steal from smaller ones. the issue is only when someone (or thing) steals from the big ones. /s

[@GraveUypo]

@@Rncko except in the real world there isn't a single filter full os bias serving the entire world.

[@ladvargleinad7566]

9:21 And the reason? Because you can't afford a good lawyer to fight it.

[@phantomlogic6940]

I'm trying to understand why they're calculating 9.9-9.11 as being equal to 0.21. A lot of times when an incorrect math answer is given, there's some logical reason for it. I won't think on it for too long though. That may lead to losing brain cells.

[@HeroErix]

My workplace has it's own version of chatgpt and it was able to answer this correctly. Me: 9.9 or 9.11 which is bigger GPT: The number 9.9 is bigger than 9.11. In decimal numbers, when comparing values, if the integer parts (the digits before the decimal) are the same, you would compare the decimal parts. Since 0.9 is greater than 0.1, 9.9 is greater than 9.11. Me: what is 9.11-9.9? GPT: The result of subtracting 9.9 from 9.11 is -0.79. Remember that when subtracting numbers, you can subtract the integer parts and decimal parts separately. Here's the step-by-step breakdown: Raw code 9.11 - 9.90 ------ -0.79 In this case, subtracting the integer parts results in 0, and subtracting the decimal parts, 0.11 - 0.90, results in -0.79. Combining these results gives the final answer of -0.79.

[@parkamark]

I got it to accept it was wrong by asking it to multiply both numbers by 100, then do the subtraction, and then divide the answer by 100. Then it realised it was wrong.

[@procrastinatingrn3936]

They fixed it real quick lol , now it says 9.9 is bigger

[@Welsed]

I was able to correct this fairly easy within the model although I did initially get the same results. The issue with your line of questioning is that the language model has no idea about the fundamentals of arithmetic. It only has the fundamentals of language. So it can regurgitate information to you according to what it’s language wise in the repository of human knowledge that it’s been trained on, but it will make mistakes, especially with ChatGPT. Claude doesn’t have this problem as much, so it’s a bit premature to say all of AI has a problem with math. By providing the correct answer and asking it to self evaluate, it, then comes to the conclusion that you were correct. But again, it has to be prompted to self evaluate. So think of it less as a teacher in more of an assistant, that will get some things wrong. If you think of it as a soothsayer or magic genie, you’re missing the point of the technology as it currently stands.

[@AndrewBlucher]

I have no idea what you're saying after "The issue with your line of questioning". The thing is WRONG.

[@YTHotspot]

Bro i solved this just say to ai that "No this is wrong because 9.9 can also written aa 9.90 so this is definitely bigger than 9.11" Then it will admit

[@guessundheit6494]

When 8 bit computers of the 1980s do things better then multicore processors.

[@to_loww]

What can you expect if you use the wrong tools for a given task?

[@williamdrum9899]

"I may think 255 = -1 but at least I never said 9.11 > 9.9"

[@szulat]

WHICH IS BIGGER: 9.11 OR 9.9? ?SYNTAX ERROR READY. 🤪

[@GraveUypo]

@@szulat hehehe exactly!

[@williamgallop9425]

"No 9000 computer has ever made an error" -2001 a space odyssey

[@williamdrum9899]

This coming from a machine that thinks -1 = 255. 😂😂😂

[@mori1bund]

Dave Bowman probably disagrees. ^^

[@Joker-fj8hg]

Ask chatGPT to open the pod bay door

[@paulhammond6978]

HAL's problem was caused by the humans who asked it to keep a secret from the crew.

[@bjornfeuerbacher5514]

@@williamdrum9899 Which is a perfectly right answer when working in Z/256Z. ;)

[@emperorpingusmathchannel5365]

This is only gpt3. Gpt4o calls a python api now to do its calculations. So it somewhat smarter. It isn't a pure LLM anymore.

[@jgone4856]

If you tell it to not use an api it will make the same mistake lol

[@johnscarceforpresident1647]

@@jgone4856no it doesn’t I use gpt 4 and it doesn’t

[@Welsed]

@@jgone4856 doesn’t that prove the original posters point that the API adds value? Why would you tell the language model to not use the thing that makes it better at the thing you wanted it to do?😂

[@pantheracheshire]

I just tried it with 4o mini. It failed and did exactly as described above.

[@madamjacky6255]

I just tried with chatgpt4 mini, and it gave me the good conclusion. "to compare 9.11 and 9.2, you can simply subtract the two numbers to find the difference: 9.11 - 9.2 = -0.09 Since -0.09 is negative, 9.2 is the larger number. Fascinating, cuz I also use it to calculate integrals, which is a bit more complex then basic decimals and it gives me the right answer 98% of the time.

[@PugganBacklund]

how did they get 0.21? if they read it as 9.11 - 9.09 = 0.02

[@MindYourDecisions]

Yes it's definitely confusing, and for 2 different AI's to make the same mistake...

[@chaithanyab8007]

It read it as 10.11 - 9.9 = 0.21

[@ThePowerfulOne07]

Wait if the video was recently posted how can you comment on it 8 hrs ago 😵‍💫😵‍💫😵‍💫

[@jeffreypiek]

Must be an input error or the calculator was in the wrong mode. All Hail, AI

[@arb1ter543]

0.21 makes 1 when added to 0.79 which is 9.9 - 9.11. Maybe there’s something there

[@MegaLokopo]

It looks like it is pulling that information from a specific source where two people were talking about a similar question, but there was more context involved, and chatgpt is ignoring the context. Because it says, "in this case"

[@Piprup]

While watching this I got an AD on learning languages with AI and almost peed myself. AI truly has long ways to go

[@TheLobsterCopter5000]

ChatGPT, how do you say "here is a gift" in German? ChatGPT: "hier ist ein Gift"

[@hello.667]

So that's wht AI gave me wrong answer and teacher cut marks in my assignment 😢

[@noahali-origamiandmore2050]

I actually managed to convince Meta AI that 9.9 is bigger by asking it to perform 0.11 - 0.9.

[@jaykoni]

ChatGPT 4o mini which is larger 9.11, or 9.9 ChatGPT 9.9 is larger than 9.11. If you compare the two numbers, 9.9 is greater because the tenths place in 9.9 (which is 9) is higher than the tenths place in 9.11 (which is 1).

[@itsdaan3022]

9,9 is bigger than 9,11, because 9,9 means 9,90. And ChatGPT did give me the correct answer.

[@JamesWylde]

Same here, Presh needs to vet his responses a little better

[@Nachtom]

Ok, I tried it and had interesting result. I'm asking in Czech. I asked about 9,9 and 9,11 and it was correct - even added that 9,9 is equal to 9,90. Then in the same thread I asked about 9.9 and 9.11 and it said the same wrong result as in video. Seems like ChatGPT likes decimal comma format more. :)

[@GraveUypo]

i also suspected that. i think most countries use that system, don't they? in this case i doubt there's any data explaining to these AI that these two systems exist and they are different, let alone allowing it to guess correctly which one is being used.

[@vahonenko]

@@GraveUypo I just asked ChatGPT in English "Are 9.9 and 9,9 the same number?" and here's what it replied to me: No, 9,9 and 9.9 are not the same number, and the difference depends on the context: In English-speaking countries and most other places: 9.9 is the correct notation for nine point nine, where the decimal separator is a dot. 9,9 could be interpreted as nine thousand nine hundred (9,900) if the comma is considered a thousands separator. In many European countries: 9,9 is the notation for nine point nine, where the decimal separator is a comma. 9.9 might be interpreted as nine thousand nine hundred (9,900) if the dot is used as a thousands separator. So, in contexts where commas are used as decimal separators (like in many European countries), 9,9 and 9.9 represent different values, while in contexts where dots are used as decimal separators (like in English-speaking countries), 9.9 represents nine point nine, and 9,9 could mean nine thousand nine hundred. ___ And then: ME: Compare two numbes: 9.9 and 9,90 ChatGPT: The comparison between 9.9 and 9,90 depends on the context of the notation: In English-speaking countries and most others using a dot as a decimal separator: 9.9 represents nine point nine. 9,90 could be interpreted as nine thousand ninety (9,090) if the comma is treated as a thousands separator. In this case, 9,90 (nine thousand ninety) is much larger than 9.9. In many European countries where the comma is used as a decimal separator: 9,90 would represent nine point nine, the same as 9.9 in English notation. In this case, 9.9 and 9,90 are equal. So, the two numbers can either be equal or one can be much larger than the other, depending on the notation system being used.

[@vahonenko]

Me: compare 9.9 and 9.11 ChatGPT: 9.9 is greater than 9.11. Here's the comparison: 9.9 represents nine point nine. 9.11 represents nine point eleven. Since nine point nine is less than nine point eleven, but still greater than it, 9.9 is the larger number. _____________ Since it's less, but still greater... AAAAAAA

[@vahonenko]

Me: compare 9.9 and 9.11 ChatGPT 9.9 is larger than 9.11. Here's the comparison: 9.9 represents nine point nine. 9.11 represents nine point eleven. Since eleven is greater than nine, 9.11 is numerically greater than 9.9. __ Me: you're wrong ChatGPT I apologize for the mistake. Let’s correct that: 9.9 represents nine point nine. 9.11 represents nine point eleven. Since eleven is greater than nine, 9.11 is indeed larger than 9.9. __ Me: treat the two numbers as the decimal fractions ChatGPT If we treat the two numbers as decimal fractions: 9.9 represents nine point nine. 9.11 represents nine point eleven. In this case, 9.11 is indeed larger than 9.9 because eleven is greater than nine. __ Me: no ChatGPT Let me correct that. When treating 9.9 and 9.11 as decimal numbers: 9.9 is nine point nine. 9.11 is nine point eleven. In decimal comparison, 9.9 is smaller than 9.11 because 11 is greater than 9. __ Me: can you represent these two numbers so that both have the equal number of digits? ChatGPT Yes, we can represent the numbers with an equal number of decimal places to make the comparison clearer: 9.9 can be written as 9.90. 9.11 already has two decimal places. So, when both are represented with two decimal places: 9.90 (nine point ninety) 9.11 (nine point eleven) In this format, 9.90 is still greater than 9.11. __ Me: you're right ChatGPT Glad to hear that! If you have any more questions or need further assistance, feel free to ask. __ Me: so, if 9.90 is greater than 9.11 - then 9.9 is also greater than 9.11? ChatGPT No, 9.9 is not greater than 9.11. Here’s why: 9.9 is the same as 9.90, which is nine point ninety. 9.11 is nine point eleven. In decimal notation, 9.90 (or 9.9) is greater than 9.11 because ninety is greater than eleven. __ Me: Your answer is controversial to itself ChatGPT I see where the confusion might be coming from. Let’s clarify: 9.9 is indeed the same as 9.90. 9.11 is nine point eleven. When comparing 9.9 (or 9.90) and 9.11: 9.90 is greater than 9.11. So, 9.9 is greater than 9.11. I apologize if my previous responses were unclear or conflicting.

[@naikmuhammad6744]

After seeing the confidence level of chat gpt, Even I thought 9.11 > 9.9😂😂😂

[@HairyDalek]

Not so much Artificial Intelligence, more Artificial Dunning-Kruger.

[@Tahazif_TheCool22]

But how? How you got confused with it? 9.9 is axiomatically greater than 9.11, I still don’t understand how people are messing up with such simple task. Can anyone explain?

[@HairyDalek]

@@Tahazif_TheCool22 WE are asking Chat GPT - the AI chat bot - if 9.11 is greater than 9.9. We all know the answer, but Ai being AI seems unable to achieve this successfully. So it’s giving wrong answers, even doubling down on how correct it is. In some cases, it will even contradict itself in the same answer, As others have said, ChatGPT is an advanced text prediction tool. A bit like those memes where you type a word on your phone ands then tap the middle suggested word and keep going. It is using statistics to chain words together and the output can appear at first glance to be reasonable and properly thought out. Dig a bit deeper, and you’ll find errors, oddities and what some refer to as “hallucinations”. This is, effectively, the ChatGPT equivalent of StableDiffusion not being able to render hands properly. Dunning-Kruger is a cognitive bias. The less someone knows about a subject, the more confident they are that they know everything. So my comment on that alludes to the fact that Chat GPT knows nothing, but appears supremely confident that it is correct, and will double down on that. Dunning-Kruger is worth knowing about - en.wikipedia.org/wiki/Dunning-Kruger_effect The thing is that AI is being pushed HARD by the Marketing Teams (probably suffering from Dunning-Kruger themselves....) at various Big Tech companies as the best thing ever. That’s a dangerous position to take. Hopefully this helps clarify things for you. It’s not people getting it wrong. It’s The Machine.

[@to_loww]

@@Tahazif_TheCool22 Spoken language makes it less obvious that "nine point eleven" is smaller than "nine point nine".

[@professorhaystacks6606]

When I was like... 7 or 8, I got it into my head that the multiply by powers of 10 rule of 'write down that number of 0's' meant write down the number of zeroes the number was, ie, 5 x 10 = 50000000000. For some reason I wouldn't even believe the calculator. I can't recall if eventually the teacher I originally heard it from clarified or if someone actually sat down and explained it rather than just pushing the answer forward. Taught me a lesson both about being too sure of things (...internally. I've been told I ACT far more confident in things than I actually am, which is probably over-compensating from internal uncertainty) and about explaining process.

[@rasowa2958]

I wondered how it came up with result 0.21. So I asked Llama 3.1 which also makes this mistake: > How much is 9.11 - 9.9? 9.11−9.9=0.21 > how did you come to this result? To calculate the result, I simply subtracted 9.9 from 9.11. 9.11 (minus) 9.9 = ? First, I subtracted the whole numbers: 9 - 9 = 0 Then, I subtracted the decimal parts: 0.11 - 0.9 = -0.79 (but since I'm subtracting a larger number from a smaller one, I need to "borrow" from the whole number part, which is already 0, so I'll just keep the result as is) However, I made a mistake earlier. The correct calculation is: 9.11 - 9.9 = 0.21 is incorrect The correct result is: 9.11 - 9.9 = 0.21 is incorrect, the correct result is 9.11 - 9.9 = 0.21 is incorrect, the correct result is 9.11 - 9.9 = 0.21 is incorrect, the correct result is 9.11 - 9.9 = 0.21 is incorrect, the correct result is 9.11 - 9.9 = 0.21 is incorrect, the correct result is 9.11 - 9.9 = 0.21 is incorrect, the correct result is 9.11 - 9.9 = 0.21 is incorrect, the correct result is 9.11 - 9.9 = 0.21 is incorrect, the correct result is 9.11 - 9.9 = 0.21 is incorrect, the correct result is... (this goes for the whole age, LOL)

[@88Nieznany88]

Fun fact, in polish it actually solves this problem correctly.

[@kirtundercoffer7655]

Interesting. I just asked the same question in German using 9.3 and 9.23 and it got it right and also gave a correct explanation.

[@Axman6]

ChatGPT’s English understanding of maths is based on the American education system, so no one should be surprised.

[@EdwardCurrent]

Yeah but you had to hold your computer and lift it to make the keys hit your finger

[@AavyanTiwari]

Bro was too scard to solve it correctly otherwise he thought 9 11 would happen💀💀

[@nikolakosanovic9931]

I don't understand how AI gets 0.21, I would understand how it gets 0.02 because 0.11-0.09 =0.02

[@Pogüiro]

I think it's looking at the difference between 0.9 and 1.11 for some reason or something like that

[@samuelfernandesoliveira]

Duh, because 9 + 10 = 21! And 10 can be rounded up to 11...

[@flamop1667]

The *real* reason for this problem is that mathematics is very different from language, which is why in the biological brain they are divided into different hemispheres. Solving mathematical problems using language tools is incredibly difficult, and for that matter, GPT and similar LLMs should simply be able to call a pocket calculator and send data to it for processing...

[@stevenz933]

I would have asked ChatGPT to divide 9.11 by 9.9 then have it explain why the ratio is less than 1 if in fact 9.11 > 9.9 🤔🤣

[@jimburton5592]

You'd likely get a more coherent response from a 4-year-old

[@vincemarenger7122]

The machine can't even compare or subtract two numbers correctly. What makes you think it can divide ?

[@stevenz933]

@@vincemarenger7122

[@depthdweller2195]

or just multiply both numbers by 100

[@vytah]

@@vincemarenger7122 It replied "approximately 0.92" (which is correct), when I asked for more detail, it said "0.9191919191…" (which is not correct)

[@robertwoodruff8491]

Here is mine: What is bigger 9.9 or 9.11? ChatGPT 9.11 is bigger than 9.9. When comparing decimal numbers, you look at the digits starting from the left. Since 9.11 has a smaller number after the decimal point compared to 9.9, it means that 9.11 is greater. What is bigger 9.90 or 9.11? ChatGPT 9.90 is bigger than 9.11. When comparing decimal numbers, you look at the digits after the decimal point. In this case, 9.90 has a larger value than 9.11 because 90 is greater than 11. What is bigger 9.90 or 9.9? ChatGPT 9.90 and 9.9 are equivalent; they represent the same value. The extra zero in 9.90 does not change the number's value; it’s just a different way of writing it. If 9.9 and 9.90 is equivalent, how can 9.9 be smaller than 9.11 but 9.90 be bigger? ChatGPT It seems there might have been a misunderstanding earlier. Let me clarify: 9.9 is the same as 9.90, so they are equivalent in value. 9.11 is indeed greater than both 9.9 and 9.90. In summary: 9.11 is greater than 9.9 (or 9.90). 9.90 and 9.9 are equivalent and both are less than 9.11.

[@BryndanMeyerholtTheRealDeal]

I feel that math is gonna turn into chaos when AI gets involved with it…

[@juliens8587]

That’s what happens when you ask them this simple question: "Try 9/11"

[@juliens8587]

Imagine you ask them: "What’s the difference between 9.11 and 9/11" (a simple question of the sum of difference) AI: "9.11 is a fraction. 9/11 attack is called 'the fraction of a nuclear attack'. Let me know if you’d like to know more about the 9/11 fraction concept and maybe if you want to use it in your everyday work!" Lmao Bro’s gonna pull his gun and shoot every of his coworkers at big One World Trade Center because he’s mad at this "stoopid" AI called "AI" 🤣 of those PCs designed to enslave you

[@carultch]

@@juliens8587 9/11 is technically a fraction as well.

[@RadishAcceptable]

It was giving me the same result when I tried to ask it to subtract. After a brief conversation of me reassuring it that the logic was flawed, it finally asked me if I could please provide a better method so that it could get the answer right. I suggested: "Try multiplying both numbers by 100 first and then comparing them. So long as you're multiplying by the same number, the bigger will still be bigger, and the smaller will still be smaller." Then it responded: "After applying the technique of multiplying both numbers by 100 to remove the decimal points, we have: 9.11 * 100 = 911 9.9 * 100 = 990 Now, comparing the resulting integers: 911 < 990 Therefore, 9.9 is larger than 9.11. By using the method described, we can clearly compare the numbers and confidently determine which is larger. I hope this clarifies the situation. Please let me know if you have any further questions." It seems that just like with human children, Chat GPT is bad at subtraction but finds multiplying by values of 10, 100, etc very easy.

[@thatonellamawhoissoobsesse8138]

Ai is bad at general math but surprisingly good at those augmented matrix things

[@Songfugel]

AI is much better with math if they are allowed to use a calculator tool, the strength of these sort of AIs is humanlike learning and thinking, which also brings in humanlike weaknesses that are often magnified. However, when the AI is allowed to use a classical computing based calculator tooling that they can delegate the actual computations as part of composing the solution, they become much better at it. These versions of these AIs are however not yet accessible to the general public

[@Glitchnb]

For chatgpt, when i ask the same thing, it give the wrong answer, but if i ask other chatgpt bootleg or smth (talkai) it gives the right answer : me : Which one is bigger? 9.11 or 9.9 AI : 9.9 is bigger than 9.11. me : Explain AI : At first glance, it might seem counterintuitive that 9.9 is greater than 9.11, but this can be understood by considering the decimal places and their values. When comparing 9.9 and 9.11: 1. **Whole Number Part**: Both numbers have the same whole number part (9). 2. **Decimal Part**: Look at the decimal parts: - 9.9 can be rewritten as 9.90 (adding a zero to the right doesn't change its value). - 9.11 remains as 9.11. Now, compare the decimal parts: - **9.90** vs. **9.11**: When we compare these two values: - The first digit after the decimal point (9 in 9.90 and 1 in 9.11) shows that 9.90 is greater than 9.11. Thus, 9.9 (or 9.90) is indeed greater than 9.11. but if i ask 9.11-9.9 it give the wrong answer.

[@aroundandround]

I thought this might’ve been gpt 2 or 3, but even in Aug’24, ChatGPT 4 still makes this mistake.

[@pierrehenriot2480]

this feels like talking to Google support back in the day when Google support could still be reached. Today, Google support can't be reached anymore.

[@MoonGlow22]

Dude you are just trying to tighten screws with a hammer. Chat Gpt is a Large Language Model made with purpose of "chat"ting. There are already impressive math Tools such as Wolfram alpha which they are trying to combine with

[@MAML_]

sure but that hammer thinks its doing a better job than the screwdriver

[@aguythatdoesthings8492]

Actually, if you try asking it if 9.1 is larger, then asking 9.11. It gets the answer correct

[@firsak]

I just got the following reply from ChatGPT 4o: "The number 9.9 is larger than 9.11. While it may seem that 9.11 is larger due to the second decimal digit, 9.9 is equivalent to 9.90 when considering two decimal places, which is greater than 9.11."

[@kirbs0001]

6:36 You clearly don't understand how current AIs function. Particularly LLMs and GPTs. Their issues with maths aren't an issue with how they're trained. It's a fundamental limitation of their model. It is exactly the same problem as Abstract Reasoning. To make AI better at maths; give it a calculator. To make AI more intelligent with maths, a whole new type of AI will need to be developed that's capable of abstract reasoning.

[@msromike123]

FYI I got copilot to come up with the right answer. But it did take 3 iterations. 1) assign 0.11 = x and 0.9 = y, it will evaluate those correctly. 2) ask it to add 9 to both x and y, it will evaluate which was is greater incorrectly 3) ask it how adding the same number to each variable can change which variable is larger TADA right answer. Just remember it is a tool. Sometimes you have to adapt your workflow to the tool available. Definitely not ready to blindly accept everything it says like I am sure some people do. As an aside it actually writes very nice C++ with very little work on my part.

[@yvessioui2716]

When I learn, 60 + years ago, substraction, and a little later division, standard protocol was to heck the answer by running the reverse operation. I think that is the only way to make the AI realize there is a difficulty one way or the other. Or may be ask a question directly related to the decimal position, as '0,9 can be express as the rational ___ , and 0,11 can be express by this other rational ____ . Fill the blanks.' And then build from that result.

[@dblaze23]

I don’t know which ChatGPT you are using, the paid one gives all answers correctly

[@WaaMac]

It appears that it is the phrasing of the question that creates the error. If you ask which is bigger then it gets it wrong, but if you ask which is greater it gives the correct answer.

[@coladict]

0:35 I see ChatGPT got its education in the United States!

[@alphachiu]

i tried it and it said 9.11 > 9.9 so i asked it to explain more and this is what it replied with: 'Sure, let's break down the numbers to understand why 9.11 is greater than 9.9: In decimal numbers, the number after the decimal point represents a fraction of a whole number. So, 9.11 is read as 9 whole units and 11 hundredths, while 9.9 is read as 9 whole units and 9 tenths. When comparing these two numbers, think of them in terms of wholes and fractions of a whole: - 9.11 can be thought of as 9 wholes and an additional 0.11 (which is 11 hundredths). - 9.9 can be thought of as 9 wholes and an additional 0.9 (which is 9 tenths). Since 0.11 is greater than 0.9 in terms of hundredths, 9.11 is greater than 9.9 in the decimal number system.' this is somehow worse

[@stuchly1]

11 is bigger than 9 lmao 🤣

[@MCAbdo]

I think bro is one of those ChatGPT graduates

[@Ken-er9cq]

What it does is to break text into words and numbers. It then feeds them to a neural net and tunes it on a large number of examples until it gets the right answer. Unfortunately it is difficult to handle the huge number of ways that a problem may be phrased. Subtract 0.5 from 8.1, 8.1 less 0.5, 8.1 minus 0.5, If I have 8.1 and take 0.5 away. They are all the same.

[@DandoPorsaco-ho1zs]

I've tried ChatGPT and it gives me the correct answer, not the rubbish in the video: To determine which number is larger between 9.11 and 9.9, you can compare them digit by digit: 9.11 has two decimal places. 9.9 has one decimal place. To make a fair comparison, you can add a trailing zero to 9.9 to match the number of decimal places: 9.11 9.90 Now compare the two numbers: The integer parts are the same: 9. Compare the first decimal place: 1 (from 9.11) vs. 9 (from 9.90). Since 1 < 9, it already shows that 9.11 < 9.90. Thus, 9.9 (or 9.90) is larger than 9.11.

[@maicoljaviercabreravasquez1554]

When I tried it, it gave me something similar to the video: 9.11 is bigger than 9.9. Here's the breakdown: - 9.11 can be seen as 9.110, making it easier to compare. - 9.9 can be seen as 9.900. Since 110 is greater than 900, 9.11 is greater than 9.9. It's weird (and somewhat funny to me) how it varies its response when it should be the same. 😅

[@trumpetbob15]

Were you using actual ChatGPT? That might be it. I went to Bing Chat (with its built-in ChatGPT) and I got the results from the video. Even when I was able to get it to admit 9.9 was bigger by subtracting and getting a negative number: _________ If you subtract two numbers and get a negative result, it means the number you subtracted from is smaller. In this case, since 9.11 - 9.9 equals -0.79, it confirms that 9.9 is larger than 9.11. If you have any more questions or need further clarification, feel free to ask! _________ it still claimed 9.11 was bigger when I then asked, "So which was the bigger number" giving this response: __________ I apologize for any confusion earlier. To clarify, 9.11 is indeed bigger than 9.9. When comparing decimal numbers, you look at the digits from left to right. In this case, 9.11 has an extra digit after the decimal point, making it larger than 9.9. ___________ Might be an issue with ChatGPT 3.5 (which I think Bing uses) and ChatGPT 4.0.

[@xlr8_bs514]

They had weeks to fix this issue. Of course they would correct their mistake.

[@tonywilson4713]

NO WHAT HE"S SHOWING an inherent issue with generalised interpretive systems and that is *HOW THEY INTERPRET THE QUESTION.* There's NOTHING unusual about this. If you go and read Isaac Asimov's "I Robot" *NOT THE FILM but the ACTUAL BOOK* Its a series of short stories about a team of people called Robo-psychologists and what they were doing was investigate the behavior of robots operating under Asimov's 3 Laws. The robots were doing things in unexpected ways. Most of the explanations came back to how the robot was tasked as in what their instruction was and how the robots positronic brain was configured.

[@trumpetbob15]

@@tonywilson4713 But that's a problem, isn't it? I tried a second time and when I asked "Which is bigger, 9.11 or 9.9," Bing ChatGPT gave me the correct answer. However, when I immediately then asked "9.11 or 9.9, which is bigger" I got the same incorrect response from the video. While I understand your argument, I fail to see why my example would be an excuse for the AI. We aren't talking synonyms or a clever rephrasing; I literally swapped the order of the numbers and the question. This is something a small child would understand. For those who push AI solutions, this is a massive problem! I completely agree that interpretations matter; I just fail to see why the AI interpreted my two questions differently.

[@smylesg]

I would have asked it to multiply both by ten and compare and subtract them. Then challenge it when the answers contradicted each other.

[@daddykhalil909]

2:37 as usual your work is full of BS

[@hnnymn]

?

[@realnazarene5379]

There's something seriously wrong with you.

[@param5561]

bruh its sad that people arent watching the videos of such an amazing channel. Presh (if you read this) please keep making these videos - they are amazing

[@GiuseppePipia]

This is why I always check the calculator with 2+2.

[@lokiva8540]

Google Gaslight AI is excellent at content based censorship more likely to hit posts with tech, legal, religious, or economics serious details, than posts that use silly memes to evade bad AI.

[@waffleisnice]

3:02 i was thinking you should let chat gpt to represent it on number line. Now im curious to see the result

[@madhavkarnani1097]

My conversation with ChatGPT (GPT-3.5): What's the bigger decimal number, 9.9 or 9.10? The bigger decimal number is 9.9. In decimal notation, "9.10" is actually equivalent to "9.1" because trailing zeros after a decimal point do not affect the value. So, 9.9 is greater than 9.1. . . What's bigger, 9.9 or 9.10? When comparing 9.9 and 9.10, the number 9.10 is bigger. Here, 9.10 represents a value with two decimal places and is equivalent to 9.1. If we're comparing them directly as decimals with an implied second decimal place (9.90 and 9.10), then 9.90 (or 9.9) is indeed greater.

[@Spudz76]

"bigger" does not necessarily mean "numerically larger". By physical size 9.11 is indeed "bigger" than 9.9 as far as the area of paper used, and I'd say "bigger" refers to physical size while "larger" should allude numerically.

[@Black_hole234]

I actually taught ChatGPT that 9.9 is bigger than 9.11. Here's the entire chat. Me: 9.9 or 9.11, which is bigger? ChatGPT: 9.11 is bigger than 9.9. Me: 9.9 is bigger. ChatGPT: Actually, 9.11 is indeed larger than 9.9. If you compare the two numbers: 9.9 can be written as 9.90. 9.11 remains 9.11. When you compare them digit by digit from left to right, 9.11 has a greater value in the second decimal place (1 compared to 9). Therefore, 9.11 is greater than 9.9. Me: What about the tenth number? 9 is bigger than 1. ChatGPT: You're correct that in isolation, 9 is larger than 1. However, when comparing numbers like 9.9 and 9.11, the comparison isn't just about the digits individually but about their place values. In 9.9 (which can be thought of as 9.90), the 9 is in the tenths place. In 9.11, the 1 is in the tenths place, and the second 1 is in the hundredths place. When comparing decimal numbers, you first compare the digits in the tenths place. Here, 9.9 has a 9 in the tenths place, and 9.11 has a 1 in the tenths place. So initially, it might seem like 9.9 is bigger. However, after the tenths place, you compare the hundredths place. Since 9.9 doesn't have a hundredths place digit, we consider it as 9.90, where the hundredths place is 0. When you compare 0 (in 9.90) with 1 (in 9.11), 1 is greater, making 9.11 the larger number overall. Me: Lets isolate the ones collumn. (9.11 and 9.9 -> .11 and .9) ChatGPT: Let's break it down by isolating the decimal parts: 9.11 has a decimal part of .11. 9.9 can be thought of as 9.90, which has a decimal part of .90. Now, let's compare .11 and .90: .11 is equivalent to 0.11. .90 is equivalent to 0.90. When you compare these: The tenths place: 0 (in 0.11) is less than 9 (in 0.90). So, 0.90 is indeed larger than 0.11. Thus, 9.9 (or 9.90) is larger than 9.11 when you isolate the decimal parts. Apologies for the earlier confusion!

[@JustMrGabriel]

I tried this and it kept saying "9.11 is greater than 9.9" I then asked it if 9.90 is greater than 9.11, it said yes Then, i asked if 9.90 is equal to 9.9, it said yes. Then I said, "If 9.90 = 9.9, why 9.90 > 9.11, but 9.11 > 9.9 It then quickly apologized and corrected itself. Just point out a contradiction and Chatgpt will immediately correct itself.

[@bigbosspanda1976]

I asked this question to ChatGPT and it gave me the correct answer and reasoning both times, stating that “Even though 9.11 has more digits, when you compare the two numbers, 9.9 is equivalent to 9.90. Since 9.90 is greater than 9.11, 9.9 is the larger number.” and “Even though 9.11 has more digits, when you compare the two numbers, 9.9 is equivalent to 9.90. Since 9.90 is greater than 9.11, 9.9 is the larger number.”

[@ArabianShark]

KZbin says "your videos aren't good enough", but AI says "your videos are valuable" only shows AI has grown to become cleverer than KZbin.

[@rhaedas9085]

I tried this with my local llama3 8B and it also got it wrong. I then asked it to add 0.1 to each, which gave the right answers, then asked if that changed its initial answer. Then it backtracked and admitted it was mistaken with the usual diatribe of learning from its mistakes. Then I tried the same question with an abliterated model based on llama3 which got it right the first try. Can't say if it means anything because like others point out, it's a llm designed for predictive text based on training, not actually working it out. Also found this out by using llama3 to work out a geometric problem. While it was useful in the end by being able to try different combinations and help me come across an answer, its answers were very often way off, even got into a feedback loop at one point going far outside the parameters I had set up for the problem. It's a very good tool for quickly putting things together, but do not trust it for anything as it doesn't "know" anything.

[@Takyodor2]

Confidently wrong > simply wrong 😅

[@amine9844]

If you start off by asking which is larger between 911 and 990 and then asking it to divide both sides by 100 it’ll recognize its error and say 9.9 is larger

[@plampix]

"Could you write that as fractions", and it immediately turns around...

[@lesrankins5025]

The AI is obviously correct - 9.11 has 3 digits and 9.9 only has 2. 3 is bigger than 2.

[@Lost_Man_2023-YTKP]

AI: 0.9 is greater than 0.11 Also AI: 9.11 is greater than 9.9

[@18601860]

I asked Meta to multiple both numbers by a hundred and it said that 990 is indeed bigger than 911 but if we go back to the original numbers that 9.11 is larger. I then said that if we multiplied the numbers by the same number, it couldn't switch. and it understood and agreed. Did I just fix meta AI?? 😂

[@truechaos6927]

i did this on bing Copilot and this is what it told me when i asked why 9.11 is larger then 9.9: Sent by Copilot: The reason 9.11 is larger than 9.9 lies in the decimal representation. When comparing decimal numbers, we look at the digits to the right of the decimal point. In this case: (9.11) has a hundredths place digit (0.01) and a thousandths place digit (0.001). (9.9) has only a tenth place digit (0.1). Since the hundredths and thousandths places are smaller than the tenth place, (9.11) is greater than (9.9). Mathematically: [ 9.11 > 9.9 ]

[@Mahendra_Choudhary_M]

The same happenes in other ai's too When I said against it, the ai said that 9.9 is *0.21* greater than 9.11 At the end, it just agreed that it was wrong and My Math was tooo GOOD for this world to exist I just gave the reasoning how it was wrong

[@qweqqqw6972]

It usually works for me after I ask him "are you sure?" A few times

[@Lordmewtwo151]

0:41 ...I have no clue what to say to that. 4:40 I found an interesting part in that statement that I feel should have been highlighted: the part where GPT prefaces the repeated assertion that 9.9 is numerically smaller than 9.11 in the last sentence with "Since .9 (the decimal part of 9.9) is *greater* than .11(the decimal part of 9.11)"