I came across an amusing example of an AI system being driven into a state of complete hallucination by a relatively simple OCR + machine translation challenge in Japanese when I was writing up some notes about a modern Japanese board game called Nukumi Onsen Kaotakuki - ぬくみ温泉開拓記 , and wanted to check the translation of the text on a particular card used in the game.

https://boardgamegeek.com/boardgame/374055/nukumiwen-quan-kai-tuo-ji-nukumi-onsen-kaitakuki

I took the screenshot (below) from a Japanese YT video about the game, and fed it into a Google Images search box, then used the inbuilt OCR facility to capture the text and pipe it into Google Translate, and here is what the AI told me

AI Overview

The image is a screenshot from the game No, I'm not a Human. 

• The card "意図せぬ来訪者" (Unintended Visitor) is a special event card in the game.

• The text "全プレイヤーは" (All players) indicates that the effect of this card applies to every player.

• The image of a person waving and the missing text suggest an action that all players must perform.

• The game is a horror adventure where players must distinguish between humans and monsters.

It also suggested that the game was available on Steam ! Most of which is complete and utter nonsense. The game is actually themed around building onsen (hot-spring spas) in a seaside town called Nukumi in southern Japan. The card text simply allows all players to draw an extra ‘helper’ card at the start of a round.

To be fair, the AI did significantly better a second time round when asked to do a ‘deeper dive’ but the fundamental problem seems to be that AI systems have a self-denying ordinance which forbids them from replying  “I don’t know” or “Insufficient data”.

I typed "what is a transom pickle" into google AI. A nonsense phrase. The reply:

A "transom pickle" is not a standard nautical term, but rather an informal expression referring to the process of "pickling" an outboard motor after it has been submerged in saltwater. The term likely gets its name from the motor's association with the boat's transom, the flat surface at the back of the vessel to which the motor is mounted.

AI did better with my next question:

How many hamsters are needed to fill the Superbowl?

To estimate the number of hamsters needed to fill the Superdome, we can divide the stadium's total volume by the estimated average volume of a single hamster. 

Step 1: Estimate the volume of the Superdome

• The Caesars Superdome in New Orleans has an interior volume of approximately 125 million cubic feet.

• There are 1,728 cubic inches in one cubic foot (12 x 12 x 12).

• Total Superdome volume in cubic inches: 125,000,000 cubic feet * 1,728 cubic inches/cubic foot = 216 billion cubic inches. 

Step 2: Estimate the average volume of a hamster

• An adult Syrian hamster can be around 6 inches long.

• Approximating the hamster as a simple rectangular prism for the purpose of this estimation gives a volume of roughly 6 inches x 1.5 inches x 1.5 inches = 13.5 cubic inches. Other estimations online suggest as low as 3 cubic inches. Let's use the 13.5 cubic inches figure as a conservative estimate. 

Step 3: Calculate the number of hamsters

• Divide the Superdome volume by the hamster volume: 216,000,000,000 cubic inches / 13.5 cubic inches/hamster = 16 billion hamsters.

Accounting for empty space

This calculation assumes the hamsters are perfectly stacked cubes. As living creatures, hamsters are irregularly shaped and cannot be packed perfectly, leaving empty space. The packing efficiency of randomly packed objects is far less than 100%. If we assume a packing efficiency of around 50% to account for this wasted space, the number of hamsters would be roughly half. 

• Adjusted number of hamsters: 16 billion hamsters * 0.50 = 8 billion hamsters.

Next I asked if cellos were a good source of fiber. AI was able to handle this one okay though it speculated I was confusing a cello with celery.

No, "cellos" (which are musical instruments) are not a source of fiber; however, celery, a vegetable, is a good source of fiber. The confusion likely stems from the similar spelling. 

• Celery: Is a crunchy vegetable that is high in fiber, low in calories, and rich in other nutrients like Vitamin K, Vitamin C, folate, and potassium.

• Cellos: Are large string instruments played with a bow, not an edible food item.

Just to see if the AI was staying fairly sane, I asked: Are cats a good source of fiber?

No, cats are not a good source of fiber, as they are obligate carnivores who get no nutritional value from it.

Hmm. Not quite understanding the question. I decided to conclude with something fairly unchallenging....

Are cats more edible if they listen to Bach every day? On this matter, the AI delivered a reply worthy of a SFN member:

There is no scientific evidence to support the bizarre claim that cats become more edible by listening to Bach or any other music. The idea is a misconception based on a loose and unproven connection between music, animal welfare, and the tenderness of meat. 

Nice to have the forums back again. I did wonder if we had broken the internet by making fun of our great lord and overmaster AI  ? (ChatGPT be thy name).

The forums vanished almost immediately after the OP appeared, and an ICANN search briefly showed them registered to a new owner in Brazil before the matrix glitched again, and they reappeared  once more from the ninth dimension :-)

Or was it all just  a “Dream within a Dream” ?

It occured to me that expecting an AI to parse a Japanese board game was perhaps slightly too severe a test of its abilities, so I tried something a little simpler, and closer to home — a chess puzzle that appeared in one of my social media feeds today.

The first problem was that the AI failed to spot that there is no black king on the board ! (the puzzle has a printing error). It’s rather hard to play chess without having your king on the board - in fact it’s impossible. The king is never removed from the board in a game of chess. The game ends in a checkmate at a point where a capture of the king is inevitable on the next move.

The AI also misread the move shown by the red arrow by stating that the white bishop was moving from C2 to D2 which would be impossible as bishops can only move diagonally on squares of one colour (the move shown is actually from G6 to C2).

When asked to do a deeper dive, the AI stated that the black king was located on G8, which is also impossible - (it would be in check from the white rook on G1).

Edited October 23Oct 23 by toucana
missing 'of' L.1

You talk with the worst LLM (Google) and you complain.. Weird.. Talk to neanderthal about quantum physics..

Edited October 24Oct 24 by Sensei

7 hours ago, Sensei said:

You talk with the worst LLM (Google) and you complain.. Weird.. Talk to neanderthal about quantum physics..

The problem is that countless millions of people rely on whatever AI happens to be built into their default browser, which for many will be Google. They won’t have subscriber access to the SOA AI systems used by researchers.

In Google these AI ‘overviews’ appear in a wholly unsolicited way the moment you put any sort of  query into a search box. I didn’t ask for any of them. With the Japanese board game I simply wanted an OCR capture and translation, and with the chess problem I just wanted to obtain image search matches on other web pages.

The “Google AI overview”, also known as Google Gemini LLM, simply summarizes the content contained in the links below. It does not create any content on its own. If the authors of these articles later write nonsense, the summary will also be nonsense. I don't understand what is unclear here that some people think it deserves a negative rating.

Since there is very little information on this topic, mainly from links, it is easy to create articles that force AI to generate unreliable summaries.

If LLM/"AI" or a human reads 10 articles in which red has a value of 0xFF0000 and nothing else, the result will be that red has a value of 0xFF0000, but if it reads 3 articles in which red has a value of 0xFF0000, 3 articles in which red has a value of 0x00FF00, and 3 articles in which red has a value of 0x0000FF, the summary will be worth a shit..

On 10/16/2025 at 8:44 PM, TheVat said:

Next I asked if cellos were a good source of fiber.

If the artificial intelligence responds to abbreviations, I will be more against this crap.

It is not cello. It is violoncello. It is spreading of laziness to write. So improper wording rules by accepting the wrong slang.

Intelligence should correct this degeneration of languages by proper responses, or refuse to answer.

Many sites robots ask the human if is human... What a load of diarrhea. The non-human is them.

42 minutes ago, Externet said:

It is not cello. It is violoncello. It is spreading of laziness to write. So improper wording rules by accepting the wrong slang.

‘Cello’ is listed in the New Oxford American Dictionary as a mid 19th century shortening of ‘violoncello’, which means that practically no one has been using the longer name since about 1850 - not unless they happen to be musical historians, or scriveners obsessed with using ink-horn words.

I was a professional theatre technician for over 30 years and worked on staging classical concerts and BBC radio recordings with large orchestras and string ensembles, and I never once heard any professional musician refer to this particular instrument as anything other than a ‘cello’ .

On 10/27/2025 at 7:09 PM, Externet said:

If the artificial intelligence responds to abbreviations, I will be more against this crap.

It is not cello. It is violoncello. It is spreading of laziness to write. So improper wording rules by accepting the wrong slang.

Intelligence should correct this degeneration of languages by proper responses, or refuse to answer.

Many sites robots ask the human if is human... What a load of diarrhea. The non-human is them.

Languages evolve. Dictionaries reflect that. Applying eternal rules to them is a fool's errand.

1 hour ago, StringJunky said:

Languages evolve

😀 Hi. Languages do not evolve, they are degenerated and contaminated by mostly lazyness. 😀

3 minutes ago, Externet said:

😀 Hi. Languages do not evolve, they are degenerated and contaminated by mostly lazyness. 😀

That is just not true. That would somewhat require all languages came into existence fully formed

On 10/27/2025 at 1:09 PM, Externet said:

It is not cello. It is violoncello. It is spreading of laziness to write. So improper wording rules by accepting the wrong slang.

Dude. My wife was a symphonic musician. I had cello lessons when I was a child, for a year, from a professional cellist. I've socialized with classical musicians for fifty years. It's called a cello. No one's called it a violoncello since before the Civil War. If your mother tongue is not English and it's violoncello in that language, fine. But it's cello in English.

30 minutes ago, TheVat said:

Dude. My wife was a symphonic musician. I had cello lessons when I was a child, for a year, from a professional cellist. I've socialized with classical musicians for fifty years. It's called a cello. No one's called it a violoncello since before the Civil War. If your mother tongue is not English and it's violoncello in that language, fine. But it's cello in English.

In my schooldays we were told to spell it 'cello with the apostrophe included. Not a word I write very often despite having a musical background. I guess the apostrophe has gone? I imagine it confused the hell out of the alphabetic sort routines.

27 minutes ago, sethoflagos said:

I guess the apostrophe has gone?

Hasn't been seen here. Not since the apocalyptrophe.

On 10/28/2025 at 9:47 PM, TheVat said:

No one's called it a violoncello since before the Civil War.

I’m told that orchestral scores still use the annotation ‘Vc’ to identify the stave line played by the cellos, but I suspect that this is to avoid any possible confusion with other orchestral instruments whose names also begin with the letter  ‘C’  -  e.g.  clarinet, clavichord, celesta, contrabassoon, cornet, chimes  etc..

Edited November 1Nov 1 by toucana
corrected 'contrabassoon'

The string instruments in an orchestral score often have a familial grouping of stave abbreviations:

V1/V2 —> Violin1/Violin2

Vla —> Viola

Vc —> Cello

C —> Contrabass

https://orchestrasounds.com/2013/12/10/score-basics/

Because the ‘C’ is reserved for a contrabass.

Yup. Akin to the use of pf for piano, even though it hasn't been called a pianoforte for centuries. The innovation was that, unlike the predecessor keyboard instruments, it had dynamic range, hence pf meaning "softloud."

So, yes, we're now calling it a "soft" even when someone's banging away in a honkytonk. I have a full-sized Ivers and Pond upright which weighs over 500 pounds and can make the house shake when a heavy bassline is played.

When just goofing off on the keys, it becomes a pianofarte.

6 hours ago, toucana said:

Vla —> Viola

Old viola joke (usually told by violists themselves):

What's the difference between a viola and an onion?

No one cries when they chop up a viola.

5 minutes ago, TheVat said:

Yup. Akin to the use of pf for piano, even though it hasn't been called a pianoforte for centuries. The innovation was that, unlike the predecessor keyboard instruments, it had dynamic range, hence pf meaning "softloud."

So, yes, we're now calling it a "soft" even when someone's banging away in a honkytonk. I have a full-sized Ivers and Pond upright which weighs over 500 pounds and can make the house shake when a heavy bassline is played.

When just goofing off on the keys, it becomes a pianofarte.

Old viola joke (usually told by violists themselves):

What's the difference between a viola and an onion?

No one cries when they chop up a viola.

It's a shame. I think the viola has a rather lovely sound, rather as a contralto often has a richer, sexier sound than a soprano. Bach's 6th Brandenburg for instance.

On 10/27/2025 at 7:09 PM, Externet said:

So improper wording rules by accepting the wrong slang.

What?

7 minutes ago, exchemist said:

It's a shame. I think the viola has a rather lovely sound, rather as a contralto often has a richer, sexier sound than a soprano. Bach's 6th Brandenburg for instance.

Me too. I think the self-loathing relates more to the feeling that it got less repertoire from composers and less glamorous parts. I'm told it was sometimes also seen as a consolation prize for musicians who couldn't gain mastery of the violin. It is a lovely sound, and I imagine a lot of the oniony jokes are mostly in jest, and maybe self-deprecation.

13 minutes ago, John Cuthber said:

What?

The writer is advancing the opinion that overuse of a slang or diminutive form of a word can debase proper usage. We all pushed back and pointed out how the word cello had long ago become accepted, and the usage should be seen as a normal part of the evolution of language. The same with pianoforte becoming piano a couple centuries ago.

2 hours ago, TheVat said:

Me too. I think the self-loathing relates more to the feeling that it got less repertoire from composers and less glamorous parts.

A couple of pieces where I think the viola gets some of the best lines (and that's coming from a brass player!)

I think they're both gorgeously played, though the AI may disagree.

23 hours ago, TheVat said:

The writer is advancing the opinion that overuse of a slang or diminutive form of a word can debase proper usage. We all pushed back and pointed out how the word cello had long ago become accepted, and the usage should be seen as a normal part of the evolution of language. The same with pianoforte becoming piano a couple centuries ago.

Yes, but "So improper wording rules by accepting the wrong slang." doesn't seem to parse in English.
Did you not see the irony?

3 hours ago, John Cuthber said:

Yes, but "So improper wording rules by accepting the wrong slang." doesn't seem to parse in English.
Did you not see the irony?

I did and it was pretty funny. I nearly split my pantaloons. ;)

23 hours ago, sethoflagos said:

A couple of pieces where I think the viola gets some of the best lines (and that's coming from a brass player!)

Nice, and now I know Prokofiev wrote klezmer - wouldn't have suspected. Now I'm wondering what American bluegrass and traditional music would have sounded like if all those fiddles had been violas...hmm, maybe such questions shouldn't be asked.