AI is the hottest media

Marshall McLuhan’s ever prophetic book Understanding Media: The Extensions of Man¹ discusses two types of media—hot and cold. Cold media is low definition and hot media is high definition. Cold media requires the user to supplement the lack of fidelity, to mentally fill in the blanks. Hot media fills your senses and leaves less to the imagination. AI, or more specifically human-AI interaction, is especially comprehensible in terms of McLuhan’s hot and cold media. Which is it?

Indeed, there may be many modalities in which AI-based information is channeled. In terms of sensory information, some are hotter (VR headsets) and some are colder (chatbots) than others. But the fundamental component of interactive AI is that we are allowing the machine to think for us. By definition, this makes AI white hot—AI is the hottest media. Essentially, we are asking the computer to fill in the blanks for us.

To some extent, all media thinks for us. Even a blank piece of paper is used as a tool for thought, and this extended and externalized view of the mind has been intriguingly argued.² But a piece of paper doesn’t talk back to you. Even a video game or interactive video has a predetermined set of responses. AI seems to supersede Alan Kay’s idea of a “reactive engine” as described in the first sentence of his PhD thesis—”The design of machines which can participate in an interactive dialogue is the main topic of this thesis.”³ AI is a “proactive” engine.

All of this begs the follow-up question: can AI be made cold to demand more of us? That, to me, is the most interesting research question in human-AI interaction. Even if AI is made to be more demanding—perhaps through cognitive forcing⁴—we are still leasing the direction of our thoughts to a proactive entity. It’s direct manipulation vs. interface agents all over again.

References

1. McLuhan, M. Understanding Media: The Extensions of Man. (MIT Press, 1994).
2. Clark, A. & Chalmers, D. The Extended Mind. Analysis 58, 7–19 (1998).
3. Kay, A. C. The Reactive Engine. (University of Utah, 1969).
4. Buçinca, Z., Malaya, M. B. & Gajos, K. Z. To Trust or to Think: Cognitive Forcing Functions Can Reduce Overreliance on AI in AI-assisted Decision-making. Proc. ACM Hum.-Comput. Interact. 5, 1–21 (2021).