What an AI expert thinks about the killer AI in ‘Mission Impossible’
Home/Technology / What an AI expert thinks about the killer AI in ‘Mission Impossible’
What an AI expert thinks about the killer AI in ‘Mission Impossible’

This article is a preview of The Tech Friend newsletter. Sign up here to get it in your inbox every Tuesday and Friday.

The latest “Mission: Impossible” movie features a villainous artificial intelligence called the Entity. Spooky, right?

Like many fictional AI including Skynet in the Terminator films and HAL 9000 from “2001: A Space Odyssey,” the Entity is up to no good.

There’s a race between the good guys and the bad guys to take command of this all-knowing, out-of-control AI before it destroys the world in some unspecified way.

Sure, sure, sure, the new “Mission: Impossible” has amazing stunts and action sequences. I had fun watching it.

But my technology-obsessed brain wondered: How plausible is the Entity? How does it stack up against other supercomputer baddies in popular culture? And how seriously should we take fake AI in movies, anyway?

To explore these questions, I spoke with Alex Hanna, director of research at the Distributed AI Research Institute and co-host of the "Mystery AI Hype Theater 3000″ podcast. (An edited and condensed version of our chat is below.)

Maybe it seems silly to have deep conversations about movie AI villains, but fiction matters. Made-up artificial life forms going back to “Frankenstein,” Isaac Asimov’s robot novels and Iron Man comics have influenced our beliefs about real world AI.

This piece comes with an obligatory spoiler alert: It discusses some plot details of “Mission: Impossible — Dead Reckoning, Part One.”

But reading this won’t ruin the movie for you. The plot is not the point. (The Washington Post’s Ann Hornaday gave it three out of four stars.)

Q: Does the Entity accurately portray the real capabilities of AI?

Q: Why? What's implausible about the Entity?

The idea that the Entity can hack into basically any electronic system — that it can listen and see everything happening in the world and learn from it — is pretty ridiculous. None of the current AI technologies can do this autonomously.

The Entity has such foresight that it can predict who is going to do what in the future down to the second, and game every possibility with humanlike ingenuity. That’s not possible.

The Entity also has intent. [Tom Cruise’s character] Ethan Hunt wants to shut down or control the Entity, and it is like the robots from Isaac Asimov’s fiction in that it’s trying to protect itself.

My colleague Emily M. Bender says that these machines don’t understand anything. Understanding is exclusive to sentient beings.

Q: Are there more realistic AI technologies in this movie?

Yes. Facial recognition is so commonplace in this movie that it's not even seen as AI. It's used by the American intelligence agencies, by the Italian police and by Hunt's team.

The depictions track with a lot of representations of facial recognition in popular media — that the software has true sight and can identify who is a terrorist. But this technology in the real world messes up all the time.

There’s also a moment in this “Mission: Impossible” movie in which [a character named Benji Dunn] is in a self-driving car and he hops over to the passenger seat. He puts on his seat belt as if to say, I don’t trust this thing.

That’s an existing AI that’s still questionable, and the movie makers had enough thought to make a character insecure about it. That was great.

Q: Are there examples of other AI in popular media that you think are accurate?

The perennial problem is that technology and computing are portrayed in popular media as magic.

Even in this “Mission: Impossible” movie, the idea is once the good guys get a key to access the Entity’s source code, the AI can be controlled. That’s a misunderstanding. Even if you had the actual source code of an AI, it wouldn’t tell you what you need to know.

You’d also need the model weights in a neutral network to be able to replicate any of its decision-making. You would need to know which data it has access to.

[Here’s a glossary of artificial intelligence terminology]

Q: Do you like any technology-related entertainment?!

I love [the android character] Data in “Star Trek” because he aims to be more human to improve his relationships with his crew mates — as opposed to his evil twin, Lore, who wants to kill humans and accrue power.

The movie “Hackers” isn’t realistic, but I love it for the motley crew of people who are trying to take down a mega corporation. Plus a young Jonny Lee Miller and Angelina Jolie are fun.

Q: This is all fictional. Why do fake depictions of AI matter?

Science and science fiction co-create each other.

Representations of AI can influence how we think about real-world AI because they can assign those real-life models too much agency or power.

But real-world AI is much more narrow, and destroys lives with much less public carnage — for example, facial recognition and predictive policing software.

Q: Do you believe that some technologists’ current fears of AI triggering human extinction are influenced by fictional killer machines?

I don’t know. Fears of out-of-control artificial entities have a long history. Frankenstein’s monster is a cautionary tale about man’s hubris.

[Read more: Why are we so afraid of AI?]

Q: If you were writing a fictional AI villain for a movie, what would it be?

It might be AI systems used as a cudgel by bosses to break the backs of workers. Those are the kinds of AI uses that Hollywood writers’ and actors’ unions are talking about now.

Those depictions would be more like labor movies than “Mission: Impossible.”‘ But I’d watch that.

If you have an Android phone, you will soon be notified if someone is stalking you with Apple’s AirTags location-tracking devices.

Relatively inexpensive Bluetooth trackers such as AirTags can be useful for keeping tabs on your keys and luggage, but they have also been abused by stalkers to covertly track their victims.

Apple’s iPhones have added notifications when unknown AirTags are moving along with you, but the notifications weren’t built into Android phones. (Apple has an Android app to scan for AirTags nearby. It’s a good bet that not many people use that app.)

Google said on Thursday that most Android phones over the next few weeks will update automatically to include anti-stalker notifications for AirTags. The notifications don’t work for other brands of Bluetooth trackers.

Once you have the updated Android software, you’ll be able to tap on the phone notifications for more information and advice on what you might do.

These automated alerts work for the majority of Android devices dating back to 2015, Google said.

Google also included a feature to manually scan for Bluetooth trackers that are near you.

Go to Settings → Safety & emergency → Unknown tracker alerts and tap the “Scan Now” button. (This feature wasn’t available for my Android phone as of Friday morning.)

Google and Apple are working on a separate technology that could automatically detect major brands of Bluetooth trackers near you including AirTags, Tile, Chipolo and eufy Security.

That technology isn’t available yet. Yes, this is all confusing.

Read more: People are using AirTags to track children and give them freedom

Source link