Posts tagged ai
Hiding messages in machine-generated text
On a whim I thought I’d try getting ChatGPT to do a bit of steganography for me. There are a bazillion ways (give or take) for hiding a secret message in unrelated cleartext, where there’s a trade-off of secret bandwidth against cleartext flexibility. I chose Morse code encoded in the last letter of each word, because it’s obvious and easy to express as rules that anybody can follow.
Poisoning badly-behaved AI crawlers
Regardless of how you feel about generative AI stealing your job or filling creative spaces with slop or whatever, one thing I hope is less debatable is that their crawlers should still obey the rules of the road.
Machine innervation
Looking at the faults my espresso machine develops over time, I’ve often wondered if fault detection and error messages could be better. It’s probably safe to assume that the cost of including sensors for all possible faults and developing code to check and understand all those sensors is prohibitive and unreliable, so I’ve wondered about a different approach.