For hundreds of years general encyclopedias like Britannica or Wikipedia have been a human creation that centralized control over knowledge access and content. These works are not part of the usual academic curriculum but a source of reference for ordinary people. The AICYC project recognizes that human editors can no longer keep up with the flood of human created articles let alone the new AI generated ones. On top of the volume problem, centralization of ownership threatens a free exchange of knowledge and ideas by using risky algorithms geared to profit over the general good. In particular there is a disconnect between the rewards for creative work and digital advertising revenue. Ad revenue is not based on quality and even increases by controversy over facts and values. In other words increasing misinformation rewards the owners of centralized platform by increasing traffic. AICYC provides a way out of the fragile future of knowledge by using AI to break information free from government, political or profit control.
PROMPT: Suppose SAM and LLM collaborated to build an encyclopedia of knowledge with SAM providing LLM the prompt tokens and steering while LLM wrote the articles?
CHATGPT-3: That’s an interesting idea! Such a collaboration could potentially lead to a more comprehensive and accurate encyclopedia of knowledge than either model could produce alone. Here’s how the collaboration might work:
1. SAM would provide LLM with a prompt that specifies the topic or concept to be covered in the article, as well as any relevant background information or context.
2. LLM would use its language generation capabilities to write the article based on the prompt provided by SAM. LLM’s output would be in the form of natural language text.
3. SAM would then analyze the text generated by LLM and check it for accuracy and coherence. SAM could also use its structured knowledge graph to identify any missing or incorrect information in the article, and provide feedback to LLM to improve the quality of the article.
4. LLM could then revise the article based on SAM’s feedback, and the process could continue iteratively until the article is deemed to be accurate and comprehensive.
By collaborating in this way, LLM and SAM could leverage each other’s strengths and compensate for each other’s weaknesses to produce high-quality articles that are both accurate and easy to understand.
The AICYC directory is proof of scale
END
The iKNOWit Story 