A Paradigm Shift: From Generative AI to Smart Research Tools

🌈 Abstract

The article discusses the importance of being a critical consumer of information, especially when dealing with AI-generated answers that lack proper explanation and justification. It highlights the need for an AI-powered "smart" research tool that can provide comprehensive information, including references, confidence levels, and credibility markers, to support the user's investigation process.

🙋 Q&A

[01] Why is the asthma inhaler dose the same for the author's 7-year-old son as for the author?

• The doctor explained that the dose is the same, but the amount of medication the son gets is less because his lungs are smaller.
• The author found this explanation "such a cool answer" as it shows that the dose is "self-adjusting by lung volume".

[02] Why is it important to ask "how do you know that?" when presented with information?

• The author argues that with the abundance of dubious information, it is critical to always ask for justification and reliable sources.
• This is based on the theory of "virtue epistemology", which says we should learn and strive to follow good practices surrounding knowledge.
• Insisting on reliable sources is a central epistemic virtue.

[03] What is the problem with current AI products like ChatGPT and Google's SGE?

• They provide answers without any explanation or support, similar to the doctor who just said the dose was right without further justification.
• The author doesn't want to just take the AI's word for it and wants to see references, assessments of source quality, and markers of confidence.

[04] What kind of AI-powered "smart" research tool does the author propose?

• A tool that summarizes what is known and what is not known, with visible markers of confidence level for each point.
• It should provide references, references for the references, and credibility markers for each source.
• The tool should also be designed based on user testing to encourage the use of well-supported sources.

[05] How is the proposed "smart" research tool different from "explainable AI"?

• Explainable AI focuses on showing the AI's work for a given answer.
• The proposed tool assumes the human is conducting an investigation and aims to provide comprehensive support for that process, not just explain a single answer.