Post 34: An Angel on your Shoulder
A while-hat LLM? Part 4 of the “LLMs, a solution in search of a problem?” series
This week's cramming of posts ended up going in a direction I didn’t expect. In this conversation ChatGPT and I discuss a virtual “angle on your shoulder.” An idea for an AI that is always with you when you are online and warns when the content you are receiving is deceptive or manipulative.
But, as discussed below, it is also more than that. With an LLM integrated it is also a companion that can tell you in plain language why what you are seeing is deceptive/manipulative. . . something that can answer your questions but is fundamentally on your side, something that proactively works to protect you from all of the online forces that are, and I don’t want to sound too paranoid, fundamentally out to get you, whether by design or side-effect. (For example, by design, the LLM integrated AI trying to convince you that you need a certain product, or by side-effect as we have discussed this week, the optimization algorithm on your favorite social media app that keeps pushing you towards the fundamental unreality of political extremes).
The discussion below follows from where we left off yesterday and continues to refine what such a tool could look like, how it could be created, and also by whom. (The irony of discussing a defensive LLM AI with an LLM designed by OpenAI is not lost on me nor is it, somewhat amusingly, lost on ChatGPT.) We discuss the merits of open source, (seems like an absolute must) and non-profit vs some kind of hybrid funding model and the potential pitfalls of each, (I don’t think this one is as obvious as it seems).
I don’t pretend to have any idea whether such a thing could be created, (although ChatGPT seems to think it is technologically feasible… I am certain ChatGPT is a techno-optimist by design 🙂), but given the need, I certainly intend to continue exploring the idea.
Finally, one more thing before the conversation. I am pretty amused that this whole idea is falling under the “LLMs, a solution in search of a problem?” series. After all, this idea very much looks like a solution to a problem that was created by the solution. A beneficent AI to contend with malignant AI. Very poetic.
