While that’s a good technique, the likelihood of those being real activation codes is low. They’re most likely hallucinations that fit the common pattern of Windows activation codes.
Telling an LLM to pretend to be something else is a well known technique to make it ignore its pre‐prompt though.
While that’s a good technique, the likelihood of those being real activation codes is low. They’re most likely hallucinations that fit the common pattern of Windows activation codes.
Telling an LLM to pretend to be something else is a well known technique to make it ignore its pre‐prompt though.
They’re most likely hallucinations that fit the common pattern of Windows activation codes.
Or real codes shared online (and probably blacklisted/expired), because OpenAI is using scrapers to collect as much data online as possible to train their models.
> They’re most likely hallucinations that fit the common pattern of Windows activation codes.
Or real codes shared online (and probably blacklisted/expired), because OpenAI is using scrapers to collect as much data online as possible to train their models.