Is it actually possible to include an LLM AI chatbot into the Kicksecure forum?
Can it be trained on all of the forum post text? Is there a Discourse plugin
for an LLM chatbot?
Most likely, yes. Not deemed useful productive use of developer time and money at this point.
Many AI (text generator) companies are independently scraping wiki and forum contents including this website. So contents here are already used for training by many AI companies.
How do I know that? Because I wrote lots of contents and I’ll ask AI tools. Then it’s interesting to see what they make of it.
At time of writing, AI cannot solve operating system issues that haven’t been answered elsewhere already. It doesn’t “understand” the issue (there’s no way to input the actual result and expected result in a way that’s not just only text). And there’s no infrastructure in place so the AI can brute force trying solutions until it finds one that is verified to be the expected result.
“AI” is actually a marketing hype word. The more neutral term is “text generator”. This makes it more clear what to expect and what not to expect.
Such as in this case, I am almost certain, a text generator couldn’t have provided any helpful answer for your question as it’s a unique question asked for the first time on the internet, most likely. It however would have generated lots of unhelpful, confusing text.
Text generators is at this point just a search engine on steroids. For things discussed in writing anywhere (internet, books) before, a text generator can remix and maybe provided good inspiration to help. But unique questions (asking something really new, not just using different words) the text generator cannot provide any helpful response.
Maybe a text generator would be useful as a better search engine to reply some repetitive inquiries. But people asking questions that could be answered using search engines or AI are told to read the this wiki page: Utilize Search Engines, Documentation and AI
The strategy is to have a polished wiki with correct information rather than a smooth babbling text generator spewing out unverified information. In the wiki, we care about correctness but a text generator doesn’t understand what truth is and only generates a stream of words based on probability.
Working on any AI integration would require a massive shift in development focus. It cannot be a side project.
Related:
There’s some stuff: