AI, the complicated life, and the moral question

Life is too damn complicated. Personal AI assistants could simplify decision-making - and marketing of the future might primarily target machines.

My colleague Bruno Giussani recently posted a video clip from a TV show called "The Good Place." I've never heard of the show before, but after watching a few clips, I'm intrigued by it.

"The Good Place" revolves around the question of whether someone goes to a nice, heavenly place after death. However, to get there, you have to have lived a morally righteous life. To measure that, a point system exists that tallies up all the actions you took in your life (yes, the Chinese social scoring system says hi).

Life is complicated

There’s a scene where Michael, the main character, explains to a judge from the good place that the point system is broken because nowadays, life is too complicated:

Michael: “Your Honor, I once stood in front of you and said I thought there was something wrong with the point system. I finally know what it is. Life now is so complicated. It's impossible for anyone to be good enough for the good place. […] These days just buying a tomato at a grocery store means that you are unwittingly supporting toxic pesticides, exploiting labor, contributing to global warming. Humans think that they're making one choice, but they're actually making dozens of choices they don't even know they're making.”

Judge: “Your big revelation is life is complicated? […] I mean, this guy chose this tomato. Those are the consequences. You don't want the consequences? Do the research. Buy another tomato.”

(emphasis added)

I think this scene brilliantly captures our current moral dilemma. Life is indeed complicated. Just buying a tomato can be a tough decision.

In another scene, Michael describes how someone intending to do good - by buying his grandmother roses - ends up facing a series of unintended negative consequences:

“In 2009, Doug Ewing of Skagsville, Maryland, also gave his grandmother a dozen roses. But he lost four points. Why? Because he ordered roses using a cell phone that was made in a sweatshop. The flowers were grown with toxic pesticides, picked by exploited migrant workers, delivered from thousands of miles away, which created a massive carbon footprint. And his money went to a billionaire racist CEO who sends his female employees pictures of his genitals.”

What are we to do about this? It would be easy to just adopt the judge’s stance: “do the damn research.” However, this view is shortsighted. No one has the time to do the research; it’s difficult; data may be hard to come by, and so forth. We hope for regulations, we rely on labels to protect us from some of these issues, but it’s an ongoing battle.

Life is complicated… for humans

This is where AI comes in. I, like many others, see the ultimate killer app of AI as the personal assistant. And it's clear that’s where things are heading.

Current AI systems are very limited because their context windows are short, and they only have access to a bit of data. Imagine a context window that holds your key life experiences and preferences. Imagine the AI with access to detailed data about products, supply chains, political dynamics, and so on.

Which tomato should I buy to do good (or at least, do little harm)? Don’t make me do the research. The AI should do the research and then give me a few options to let me decide. In many cases, I might even want the AI to decide for me, given my preferences.

Marketing for the machines

The implications are stark. Today, when I walk into a store or browse online, I'm bombarded with marketing messages. From the presentation to package information, everything is designed to make me buy it. But what are the unintended consequences? It’s too stressful. I'm simply looking for a healthy meal without doing harm down the line.

Me: “AI, I want a healthy meal, what should I buy?”

AI: “Here’s a healthy choice without unintended negative consequences:…”

Today’s shopping feels like email without a spam filter - impossibly hard to navigate. In the future, we might not have to make these choices. I might not be looking at products directly; the AI assistant will sift through them for me. If you have something to sell, you might first need to market it to the AI, not to me.

Does this sound dystopian or utopian to you?