https://openai.com/blog/multimodal-neurons/
ML Training on images and text together leads to certain neurons holding information of both images and text – multimodal neurons.
When the type of the detected object can be changed by tricking the model into recognizing a textual description instead of a visual description- that can be called a typographic attack.
Intriguing concepts indicating that a fluid crossover from text to images and back is almost here.
There are a few potential security concerns to consider when working with language models:
- Data privacy: Language models often require large amounts of data to be trained, and this data may contain sensitive or personal information. It is important to ensure that this data is protected and that appropriate measures are in place to prevent it from being accessed by unauthorized parties.
- Model security: Language models can be vulnerable to attacks such as adversarial examples, in which an attacker intentionally manipulates the input to the model in order to cause it to make incorrect predictions. It is important to consider the security of the model and take steps to protect it against these types of attacks.
- Misuse: Language models have the potential to be misused, for example by generating fake or misleading content. It is important to consider the potential unintended consequences of using language models and to put safeguards in place to prevent their misuse.
- Bias: Language models can sometimes exhibit biases due to the data they are trained on. It is important to consider the potential biases in a model and take steps to mitigate them.
- Intellectual property: Language models may be protected by intellectual property laws, and it is important to respect these laws and obtain the appropriate licenses when using language models developed by others.
Secure Machinery 