Since the arrival of ChatGPT and subsequent Large Language Models (LLM) there has been a mixture of excitement and fear. Excitement that they can speed up and support many jobs through to the fear that they will replace jobs and may even take over the world.
There is no arguing that LLMs are impressive and are already changing the world, mainly for the better. However, when it comes to accessibility, they show promise, but will not replace accessibility. Microsoft have set the right tone around LLMs, they are a support and need us humans to validate and check them before using their output. LLMs are only as good as the data and training they have access to.
There is some speculation that there is no need to make websites and content accessible as LLMs will do it as they improve. I disagree, we need to make websites and content accessible, so LLMs can consume it. Good accessible content makes the content accessible to assistive technology users, but also makes the information more contextually aware to LLMs. We need to double down on creating accessible content, not take our foot of the gas as technology looks like it will remove the barriers we have put in place while creating the content.
Read more > No, ‘AI’ Will Not Fix Accessibility — Adrian Roselli
Accessibility is about people. It is not a strictly technical problem to be solved with code. It is not the approximation of human-like ramblings produced by the complex algorithms generally branded as “AI”. LLM chatbots and computer-generated images are already creating an uncanny valley of spam. We don’t need them to contribute to the analogous uncanny accessibility already extant on the web.
https://adrianroselli.com/2023/06/no-ai-will-not-fix-accessibility.html
unknownx500
