Chatbots Decoded: Exploring AI (Review)

2024-11-15

Chatbots Decoded: Exploring AI is a new exhibit from the Computer History Museum (CHM) opening November 20th, 2024. In development for more than a year, the exhibit covers both the history of chatbots and conversational interfaces as well as the social ramifications of the technology. While there is not an explanation for why GPUs rather than CPUs are so useful for learning or what “attention is all you need” means, both are alluded too and I think, many audiences, including children, will enjoy it.

Conversational interfaces, as typified by chatbots, have had a long parallel development in computing with more traditional interfaces. Science fiction has featured robots and automated creatures controlled by voice or written words as far back as the Golem and R.U.R. Star Trek’s main computer presents a mixed interface, both tactile graphical displays and voice control. The series differentiated between the computer’s intelligence and sentience. Although the computer could flexibly carry out tasks, it lacked any introspection, motivations, or free will. This distinction is one of the subtleties explored in the exhibit.

The exhibit begins with fictional robots that, by obeying voice commands, helped shape both our expectations as well as fears. The first real grounded example is ELIZA, the conversational system which was typically used to imitate a Rogerian therapist. Visitors have access to a terminal where they can interact with an implementation, hopefully experiencing some of the magic feeling of interactivity but also witnessing the limitations of the technology.

Developments in symbolic logic and improvements such as A.L.I.C.E. are recounted. However, the next big revolution is deep learning where a timeline describes the exponential growth in the models, the amount of data, and the amount of funding. This leads into probably the central showcase: the Ameca exhibit.

Ameca is a high-fidelity robotic face joined to a ChatGPT model. Visitors can step right up and start talking to Ameca, which tries to spark some conversation on its own. According to a staff member, children feel an immediate emotional connection to the robot, laughing and giggling. For my part, I felt self-conscious, wanting to say something profound amongst all the humans in ear-shot. In order to prompt conversation, Ameca will urge visitors on. This can end up interrupting the visitor’s attempts at interaction, but that ends up increasing the humanity of the experience.

As a topic that is more current event than historical, Chatbots Decoded concludes with a large section of open questions. For instance, CheatGPT raises both the potential of LLMs as education aids, the concerns of students using LLMs as labor-saving devices, the battle to detect LLM generated text versus student authored text, and how relevant the older debate about student use of calculators is to the current fight. With no planned closing date for the exhibit, CHM will need to gauge when new developments warrant an update to the display.

All photos copyright of the author.

Note: As a Supporting Member of the CHM, I gained early access to the exhibit. I was not compensated for the review.