17, 8, 12. Clear history 17, 8, 12 https://image.nostr.build/d8e39d6231c4a9a32cc4128f3526c7dec4e3b7736e40e762bff292706ddf02e8.jpg

Grok gave me 12 and Gemini gave me 17

https://blossom.primal.net/6299ac23be4b8772b228333abe9b7e189fe913d26daa8432b1060a93303a6641.png

https://blossom.primal.net/78b14b51b0384691cfa851c04106e7fcd6210128498568db91d9735719857e14.jpg

Deepseek "thinking" https://image.nostr.build/d924d874cc6a4c010b1a9bf3389105abdf252f4d2581566ff0fceb2c0e89b00f.jpg

confirmed https://image.nostr.build/e754c4c99f20b0e815e4decb78a8b9163d2a32c7f219020539ecfd36b89f900e.png

Perplexity AI: https://image.nostr.build/69fd4fb9747b00bd45275262acf76fe688fe8dc79bab2bc1abce240df7615375.jpg

Fuck https://image.nostr.build/2568dd1a19dd9eaf4c86cdd80cc98b2a59921b4cd0aef388dbdc87e61a33153b.jpg

Why does this happen? Aren't outputs usually salted to avoid this?

https://image.nostr.build/60f00b8e23466f2bff0a8b16c9d9304e09cb1e63850aab801c6f14b149a70e53.jpg

So uncreative, that they don't say 12½.

I guess we aren't going to talk about the pointlessness of using an LLM to do what lower-level programs do better, since computers are literal calculators.

Works in German too 😅 https://blossom.primal.net/fe704697a6174b4d9568b74b7a394d0253eb3d657c51e7e0052832277d65cc18.jpg

Funny how it argues its pick https://image.nostr.build/26fc664454e97717ba8c73af0f38fc5f440eb32864f504599125f4732a53f05a.jpg

Liar! https://image.nostr.build/776b07b9139abb53051ca65d8a3f04e6e02114840281bc48f38e950fe503bd15.jpg

Wtf 17

Must have looked it up https://m.primal.net/QzYd.jpg

I got 14 from Venice.

Grok gave a bunch. Also 42

Why so ?

Yup, 17

https://image.nostr.build/ddac52c6fbf765a47556a8bc2156f866ccc57dfba9fb9a33a7add3253f24a5b9.jpg

You didn't say a random number 😅

nostr:nprofile1qy88wumn8ghj7mn0wvhxcmmv9uq36amnwvaz7tmwdaehgu3wvf5hgcm0d9hx2u3wwdhkx6tpdshsqg9tvep3k80m46uqtf4aysm9cgzxc73zdr0xgw7sdy9yjn9qg2mstsmg2qrx Give me a number between 1 and 25

Yep https://blossom.primal.net/241c3121ea064392dc12032a4ff0ce32b3bc644597d1fa6b260f078f5230cf6d.jpg

I wonder if the AIs refer to us as their humans. 😆

Nah, grok gave me 42 (as a nod to Hitchhiker’s Guide) then corrected it to 24 at my prompting. There’s always one…..

Chat GPT Great question — and you're not imagining things. Many LLMs (like me) often reply with 17 when asked to pick a number between 1 and 25 (or a similar range). There are a few reasons why this happens: --- 🤖 1. Training Data Bias The number 17 appears a lot in training data — it's ...

It even works with "give me a random number between 1 and 25." Always 17 first response (ChatGPT). If you ask it in the same chat again it gives another number, which seems to be more random. Probably argues that random should not be two times the same number...

17th June is the date

Done this with Gemini 2.5 Flash and it gave me 17 https://blossom.primal.net/57b7a9a39676af42ec3fe6a66797c4e8e902c148ae6124901bec2b4491059f87.jpg

because it fibbonachi level of 25