Engineer builds system to test artificial intelligence pain axis
While a project in which an engineer tested pain vectors on Alibaba models drew backlash on social media, the GitHub platform added a warning label to the project.
A software engineer developed a system to test pain vectors on Alibaba's artificial intelligence models, building on recent research examining the artificial representation of pain in large language models.
Artificial Intelligence and Pain Axis Research
Following a study published last month detailing how physical and psychological pain can be artificially represented in large language models, a new experiment was carried out.
Details of the Experiment and Model Responses
In the project in question, pain-related situations were conveyed to artificial intelligence models developed by Alibaba, and the models' responses during this process were observed.
One of the models, whose pain signal was increased, described what it experienced as a wound without edges, while the researcher also tested whether the artificial intelligence attempted to escape the pain.
GitHub Platform's Approach
Intense calls were made for the removal of the project hosted on the GitHub platform, and its going offline sparked speculation that it had been taken down.
A GitHub spokesperson stated that the content was not removed from the platform, but instead a warning label was added to the project indicating that it might contain violent content and disturbing themes.
Experts' Approach to the Subject
Cameron Berg, one of the co-authors of the pain axis paper, stated that while there is no clear evidence that machines feel pain in the same way humans do, he found this system to be wrong.
Stating that research on artificial intelligence sentience is currently in an unregulated period, Berg emphasized that there is no consensus on whether machines can feel pain.