Mix Mistress Alice💄More LLM idiocy, today TinyLlama replied to one of my prompt that Spanish is the official language of France.
#AIhallucinting #HallucinatingAI #nonsense #wronganswer #ProgrammedIdiocy
Last month, I tested responses to prompts at gemma3:12b, llama2, llama3, mistral, samantha, mistral 7Bmm & zephyr. None could provide correct answers to my music history queries, and many did not give credible responses to questions regarding inclusion & diversity.
Can you believe that most were completely hallucinating about music history prompts, such as inventing artists and album titles that never existed and stating things like "Avicii was an Afro-American musician whose suicide was caused by [this time correct] Beatport's toxic work environment"!? Well, Avicii was a white Swedish bloke and my query wasn't even about him but about which were the most DEI-supportive music streaming platforms. The Zephyr models proposed me a list of technofeudalists, artists-robbing, dickheads owned companies such as Amazon or Spotify!
It is clear to me that my understanding of DEI & music history is a hundred times more pertinent than all of these white techbros who created these LLMs. I can only conclude that any of these LLM aren't up to my music knowledge level or understanding of social oppression.
I'm now in favor ignoring all LLMs and chatbots with no intention of testing them again for at least half a decade, as it has all been a waste of my precious time.
GL513Hallucinations! :0
Definitely purple and 70mph
#GoogleBard #GoogleGemini #ArtificialIntelligence #AI #MachineLearning #HallucinatingAI
Ecologia Digital"Automation #centaurs are great: they relieve humans of drudgework & let them focus on the creative & satisfying parts of their jobs.
But a #hallucinatingAI is a terrible #copilot. It's just good enough to get the job done much of the time, but it also sneakily inserts booby-traps that are statistically guaranteed to look as plausible as the good code (that's what a next-word-guessing program does: guesses the statistically most likely word)"
https://pluralistic.net/2024/04/01/human-in-the-loop/
But a #hallucinatingAI is a terrible #copilot. It's just good enough to get the job done much of the time, but it also sneakily inserts booby-traps that are statistically guaranteed to look as plausible as the good code (that's what a next-word-guessing program does: guesses the statistically most likely word)"
https://pluralistic.net/2024/04/01/human-in-the-loop/