Behind the Scenes

The People Behind AI

It’s crucial for the public to understand how much human labor goes into creating today’s chatbots and other generative AI systems. Much of this labor, often conducted in the global South for low pay, has to do with the data that goes into making these systems and also into keeping these systems safe and trustworthy for users in the global North.

While much of this AI labor can be exploitative or even harmful, a new nonprofit called Karya is finding a way to use AI labor to address poverty in India in an ethical way.

Image: Baby Rajaram Bokale, an AI worker from India interviewed for the exhibit.

The Ideology Machine

I’m fascinated that some of the earliest chatbots were made by psychologists and psychiatrists. In the late 1960s into the early ’70s, psychologist Robert Abelson created a chatbot-like system as part of a project to model political thought and other kinds of motivated reasoning.

The “Ideology Machine” modeled the reasoning of a steadfast supporter of the ultraconservative US senator Barry Goldwater.

Image: Yale psychologist Robert Abelson.

The Trouble with Humans

There is a lot of talk about making chatbots safe for us, but what about the issue of keeping chatbots safe from us? The story of the chatbot Tay by Microsoft really raises this important issue. Designed to learn from its interactions with users, when Tay was connected to Twitter (now X), malicious users quickly trained Tay to produce racist, misogynist, homophobic, and antisemitic language.

Defending chatbots against these types of malicious attacks and hateful language from users has become a major factor in commercializing AI and in AI safety.

Image: Tay webpage.

A Parrot?

Why is there a taxidermy parrot in the exhibit? It represents what I think is the key question about today’s chatbots: “Some experts say today’s chatbots are stochastic parrots, that only give the appearance of understanding things, while other experts say they are emergent learners, that actually do understand things about the world. Which is it?”

Experts have wildly different answers to this question, and a lot about our future depends on which answer turns out to be correct.

Creating Immersive Experiences

CHM Director of Media Jon Plutte says that the chatbot exhibit will feature extensive projection mapping throughout. This technique has taken the museum field by storm, allowing the visitor to “step inside” the story being told.

Creating a projection map experience requires lots of math and calculations, creativity and filmmaking, and especially, says Jon, a willingness to literally think outside the box.

Preparing artifacts for exhibition

CHM Director of Collections and Senior Registrar Aurora Tucker explains that when artifacts go on display, the registrar is responsible for their care throughout the course of the exhibit. When on display, an artifact is subject to more wear and tear than when it’s in storage.

For the chatbot exhibit, Aurora will be utilizing a combination of pre-made acrylic mounts for simple items like books, but she will also hand-solder custom brass mounts for items like figurines, so that each artifact not only looks stunning but is supported and protected for display.