. . . was enlightening. It confirmed my knowledge of artificial intelligence and reinforced the need for assessing trustworthiness of Chat content.
So that I’d be able to evaluate the tool, I asked a question in the center of my religious and theological scholarly expertise. I asked “what’s the Biblical book of Jeremiah all about.” The errors were immediately obvious to my scholarly eye—and there were plenty of errors. So, I want to talk about how to assess just a few of the elements of a product or service.
To assess the value of any product or service, you always begin with the fundamental proposal—which usually appears in its advertising or in the sales pitch. In argumentation, the proposal is typically called the “claim,” which is built on specific data and oriented to an objective.
AI development folk have already created development guidelines for human system interaction. Although the guidelines are exceptionally specific for AI development, I’m going to work from the more standard argument process which works across all contexts and can be applied fairly simply—and then added to for specific contexts like digital chat. My basic approach—the Toulmin argument model--applies to absolutely every context imaginable and makes it easy to understand the issues. Then, if you so desire, you can apply the context specific translation of guidelines.
The underlying approach for assessing an AI app is like any other product or service--whether for an auto, a new house, a frozen vegetable or even a condom. I happen to drive a Mercedes C300. The Mercedes proposal is very clear-cut: “we are the defining class since 1886.” The Trojan condom, in contrast, proposes that they are the defining “instrument of closeness.”
The Chat GPT, too, makes a very clear proposal: “we communicate with enterprise data across industries, languages, and specialties. . . driving a new era of convenience, productivity, and connectivity.” Although generative AI is a unique species, it’s spread is moving across industry, healthcare, natural language processing, space exploration and science.
The fundamental issue in all these AI products is trustworthiness.