OpenAI’s GPT-4 security programs damaged by Scots Gaelic • The Register

[ad_1] The protection guardrails stopping OpenAI’s GPT-4 from spewing dangerous textual content could be simply bypassed by translating prompts into unusual languages – similar to Zulu, Scots Gaelic, or Hmong. Giant language fashions, which energy immediately’s AI chatbots, are fairly completely happy to generate malicious supply code, recipes for making bombs, baseless conspiracy theories, pretend… Continua a leggere OpenAI’s GPT-4 security programs damaged by Scots Gaelic • The Register