do you think it's acceptable to use LLMs to make an operating system more accessible to those with disabilities (for example, for image description, text to speech, etc.)?
my ai take@mjdxp i believe that locally hosted ai models with the purpose of accessibility (image description, text to speech, speech to text) are actual positive uses of the tech. i think "agentic" systems could be a potentially useful acessibility feature but with the current quality it is incredibly irresponsible to make them widely available. i do also have all of the usual ethical concerns about ai such as how the models are trained and how most of them don't have actual uses. to be very clear i am very sgainst a 99% of ai use and am very very upset about how much software i liked or rely on has been affected by it (see: waybar, scalar, asp.net core, linux kernel, curl, and i could go on)
1
