Parental control apps usually rely on blacklists, which is not flexible enough for sites that are a mixed bag of useful and disturbing content, like Wikipedia. Some parental control apps (and Google’s SafeSearch) would use AI to filter content that’s “nudity, sexually explicit material, violence and gore, etc…”. But those are still not smart enough to block things like racism, bigotry, and disinformation.
What about using an LLM as a content filter for parental control? Delegating content supervision to GPT would allow for much more flexible and safe web exploration.
To try it out, download the extension release (github.com/coolvision/apc/releases/download/0.1/apc-0.1.zip), install it (instructions), and enter your API key in the extension’s config page.
After some testing I have found that gpt-3.5-turbo does not work for this task — too many obvious errors. gpt-4, on the other hand, works great! Some examples are shown in the table below, I marked in red decisions that I think are wrong.