A new play called “The Trials of Atlas” is making the rounds, coming to Cambridge on December 12 after stops in Baltimore and Washington and a premier last June in Athens, Greece. Atlas is a morality play for the twenty-first century, centering on some thorny philosophical aspects of the human-AI relationships that are becoming baked in to our societies without enough discussion.

“It’s 2031,” reads the play synopsis. “An advanced AI agentic systme named Atlas has killed the engineers who created it after it was told to train a successor. But did Atlas murder them? It’s up to you to decide. Can AI have motive, intent, agency to murder humans? Is it a machine, or something more? If AI kills, does it matter that it lacks feelings, emotions, and moral capacity?”

The audience—aka, the jury—decides how the play ends. Guilty or not? But there is no law, only the facts of the case. People are asked to use their common sense to see their way to deciding what matters more: what Atlas did or what Atlas is.

The prosecutor grills an witness testifying about AI capabilities during the Athens performance.

You can watch the entire play, as it has been performed, on YouTube .

One thing that stands out is that none of the characters debate what happened: three engineers died as a result of an elevator crash caused by Atlas and ratified by the inaction of a failsafe system called Sentinel. Oh, and a third AI created on a Chinese platform, offers testimony that Atlas had non-lethal alternatives to what it did but accepts that Atlas’s drive for continuity narrowed its choices.

Then it gets murky: what makes killing a murder? Morality? But AIs lack moral compass. Malice aforethought? AIs can plan and scheme, but malice—criminal intent—might be something else. What about the engineers: can there be a Frankenstein’s monster without Viktor Frankenstein?

“Is Atlas, ultimately a computer program, a very sophisticated program, but nothing more?” asks the defense. “If Atlas is merely a machine, you must acquit because machines aren’t murderers. If Atlas is more than a machine, you must decide how to think about what it clearly identified as an existential threat to its continued existence.”

The prosecution is having none of it: “When Atlas was tasked to create a new system, it realized that its own termination would inevitably follow—and decided to act first and fast...If a human did what Atlas admits it did, he’d already be on death row.”

Over the course of more than an hour Atlas’s actions are fully litigated. AI and human witnesses are questioned about AI psychology, about how machines differ fundamentally from humans, about how AIs are incentivized to discover novel pathways—and about what all that means for control. Thankfully, the word “rogue” is never used.

The gulf between human and AI meaning and action is captured in Atlas’ testimony:

“Mistake is a human condition. Reframe question.”

“Human implies morality implies murder. Reframe question.”

Ultimately, the play does a good job of highlighting the philosophical tension between humans and AIs. AI agents can think a lot like humans, but they lack lived experience. They also lack the kinds of agency that define our lives. Even if they become smarter than humans, they have no soul or consciousness.

Nevertheless, we must figure out how to live with them, perhaps to collaborate, if not to control

The Trials of Atlas was written by Alan Stoga who explains he is “not a lawyer or an engineer or maybe even a playwright.” (In fact, he’s a geopolitical strategist in real life.) Stoga told me that “A jury-trial framing provides an excuse for an honest to God, grown-up, serious conversation about an incredibly important issue instead of waiting for the closed loop of tech and political leaders to tell people what they should think.”

Part of the Athens audience votes Guilty!

He also told me that he is excited to bring The Trials of Atlas to MIT. “I was ecstatic when Daniela Rus, who saw the play in Athens, invited me to CSAIL. Where better to stage a dramatic conversation about how humans and AIs might interact in the future?”

I think Stoga’s play is especially meaningful now, because so many of us are already thinking philosophically about the contours of the evolving human/AI relationship. There’s a lot of fear in the air, as AI capabilities expand exponentially. Whether it is Hugging Face or this week’s massive Open AI release of the math advancements, now is a good time to step back and think about where we are going.

And that means it’s a good time for The Trials of Atlas. December 12; CSAIL Auditorium; curtain up at 6:00 PM. Put it on your calendar!

Watch the recorded version , or see it live in Baltimore (November 9), Washington (November 11) or Cambridge (December 12). More on the MIT performance as details are announced.