Exclusive Content:

ChatGPT “went crazy” and started talking with a simulated user’s voice


Company OpenAI A report detailing key risk areas in the latest version of a major language model has been published GPT-4o and ways to reduce them. In one case, OpenAI experts discovered the model’s advanced voice mode, which allows users to communicate with each other. ChatGPTsuddenly started talking with the simulated voice of the user. At such a moment ChatGPT In the middle of answering a person’s question, he suddenly shouted “No!” and then began speaking in a user-generated voice.


Experts OpenAI He explained that this occurs when noise in the user input confuses the model and replaces the authorized voice sample in the system prompt with voice input from the user.

OpenAI He also noted that the risk of unintentional audio copying remains “minimal.” The company has already prevented unintentional audio generation.

Source: Port Altele

Latest

Newsletter

Don't miss

https://www.xataka.com/magnet/colombia-ha-dado-luz-verde-se-busca-ultimo-gran-legado-pablo-escobar-introdido-pais-para-sacrificilo

Pablo Escobar, in addition to transporting large amounts of drugs, planted the seed of another "gift" that Colombia has been trying...

https://www.xataka.com/magnet/como-entendi-mejor-manifyto-comunista-gracias-a-uno-grandes-autores-ciencia-ficcion-momento

If we leave aside the connections and the directions each one leads to, it is clear what the 'Communist Manifesto' is....

https://www.mundoxiaomi.com/noticias/que-cuales-modelos- sensitives-hyperos-experience-enhanced-edition-revolucion-xiaomi-para-su-nuevo-programa-betas

To tell the truth, Xiaomi's entire beta program has been up in the air since the emergence of HyperOS, a system...

LEAVE A REPLY

Please enter your comment!
Please enter your name here