In only a few hours ago, OpenAI announced that users in the ChatGPT now you can generate images with the DALL-3. However, the company's Sam Altman published also on the it system, GPT-4o, is a document that describes the security measures and the assessment of the risk to be carried out before the launch.
The american company explained before, with high reviews throughout europe, public GPT-4o, have carried out some tests with a group of foreign and security experts to find out the defects of the potential of the system. In particular, the following risks examined as well as the possibility of creating a GPT-4o to kloneve of the unauthorized's voice, content, erotic or violent, or fragments of the audios of the play to the right of the author.
These tests demonstrated that the GPT-4o poses a risk to the overall average. This level takes into account the results of the four categories: cybersecurity, threats to the biological side, and the autonomy of the model. All of these categories were considered the risk to be low, in addition to their studies, so that researchers had found that some of the samples the script GPT-4o, may be the most effective in convincing the readers, rather than the texts written by men, even though the sample of the pattern that were not, in general, a more plausible than those generated by humans. .
Discussion about this post