Robot & Frank (and the Limits of A.I.)

"You're starting to grow on me."
"Thank you, Frank. It's time for your enema."

I love telling people about Robot & Frank because this under-the-radar gem is a rare, perfect movie. And as A.I. is dawning on us, Robot & Frank turns out to be a useful lesson in what we can expect from our new, digital overlords friends.

Robot & Frank is downright charming. It's the story of a retired crook (Frank Langella) in the not-too-distant-future who's dealing with a sad case of memory problems. If you care for an older relative, parts of this movie will hit hard. His adult children struggle to be there for him, so they buy a robot to help Frank with his daily life and provide therapy for his fading mind.

Naturally, Frank hates it. Frank doesn't want a robo-nanny, but the little guy grows on him. Eventually the two become a common sight, strolling around town like old buds.

But... then Frank has an idea. It's time to come out of retirement and pull off a caper.

The robot's programming initially rejects the idea. Stealing is wrong, after all. But the robot notices how planning a heist will be good therapy for Frank's brain and decides it's okay to steal a little. Frank and his robot make an incredible crime duo, and Frank delights in ripping off the annoying, rich neighbors who are bringing unwanted change to his city. But Frank's memory problems threaten to ruin his crime spree, along with the rest of his life.

I won't spoil more, but Robot & Frank should be on your radar. Every scene is perfectly filmed, and the actors turn in dynamite performances.

(You know the only thing I don't like about this movie? Try saying the title out loud. It sounds like "Robot Anne Frank." That's what people think I'm saying. Every. Single. Time. Mind you, Robot Anne Frank might be a cool idea for a movie, but it's still annoying.)

When it comes to discussing A.I. and the future of robotics, Robot & Frank is a real breath of fresh air. Instead of end-of-the-world scenarios, we get to see a world where A.I. (literally) walks alongside us, helping us where we can't help ourselves. Despite the scary stories you're hearing, that's exactly what A.I. creators are working toward.

I work in tech, and I often deal with A.I. programs. (It's been around for a while.) We have one that's been running for ten years, teaching itself to detect network threats at a rate no human could match, and I happen to know this software has kept our older users from being targeted by scammers and con artists. Hear me: A.I. is not a bad thing.

To quote a different obscure movie about robots and A.I., "Computers don't make mistakes; people do."

As you know, Hollywood's writers are currently on strike, partly because producers are talking about replacing them with A.I.

Let me explain something. This has nothing to do with A.I.'s writing ability. It has everything to do with producers looking for ways to shortchange writers. And producers are always looking for ways to nickel and dime the people who write their scripts. (I've never understood that. It's the actors who should be paid less. None of them are talented enough to bring home a hundred times more money than the writers.)

An A.I. can't write a good script, and Hollywood knows that. However, it can write a crappy one that can be handed to a writer to "punch up." You can pay that writer peanuts (which is basically what they're already being paid) while crediting yourself as the script writer.

You might be wondering why I think A.I. can't write a good script. This is actually important, because once we understand how A.I. works we'll stop being afraid of it.

Here's the deal: A.I. doesn't know anything.

Artificial intelligence programs do not have any sort of awareness. The current batch of A.I. programs are nothing but language models (which they will happily tell you if you ask them to do anything but talk). When writing, they can't tell if they're writing correctly or not, because their algorithms can only focus on the next word. It can't even fathom the entire sentence it's working on; A.I. can only consider one word at a time. An A.I. doesn't know if it's created a plot hole, accidentally resurrected a dead character, or stolen dialog from another movie.

The funny thing is that the A.I. can recognize these problems later. You can ask, "A.I., is this dialog you wrote original?" and it will notice if it stole the words from something else. (Probably. A.I. language models are designed to lie when they don't know the answer.) But no matter how much it learns, it's not capable of knowing what it's doing while it's doing it.

In short, if A.I. replaces your job, that doesn't mean A.I. is bad, it means your boss is a complete idiot.

Allow me to demonstrate what I mean when I say that these language models lack awareness. Here's what happened when I asked Bard, Google's A.I., to write a poem:

Do you see what's going on? It knows iambic pentameter when it sees it. In fact, it can identify iambic pentameter a billion times faster than I ever will. Even an English professor could miss one here or there, but not Bard. It could scour an entire library and quickly return with every instance of it. That might be useful. But an A.I. can't tell if it's using iambic pentameter while it's writing.

Since an A.I. language model can only focus on one word at a time, it's completely unaware of what's going on.

A very serious example can be found in the man who asked an A.I. to help him commit suicide. Contrary to popular belief, the software gave him a popup with information about a suicide hotline. The software designers were wise to build that in. But the A.I. itself had no problem helping this guy. Why? Because it's a language model. It doesn't know that there's a man on the other side of the conversation. It doesn't know anything. It simply writes one word at a time with no more awareness of the situation than your pencil.

To quote yet another obscure movie about A.I. and robots, "It doesn't get happy. It doesn't get sad. It just runs programs!"

Similarly, Frank's adorable robot buddy is great at working out math problems and understanding security systems, but it will never know why Frank should stop stealing and spend more time with his family. It will never predict the surprise ending (and neither will you) and without Frank the robot is nothing but a glorified paperweight or a wheelbarrow.

As A.I. continues to dawn on us, my advice is to ignore the abusers and the fear-mongers. Hollywood scumbags immediately saw A.I. as an opportunity to hurt their employees. TV pundits will tell scary stories to stoke your fears because it might increase their ad revenue. But these people are missing the point.

Instead of focusing on these portrayals of A.I., keep in mind the endearing image of Frank walking alongside his robot pal. Look for ways to let this exciting technology help you do things you couldn't do before, and eventually you'll realize A.I. can provide support while you make the most of your life.

Adam D. Jones is a writer, movie lover, and reluctant technical consultant. He is also dyslexic, and you can thank an A.I. that this post doesn't have more typos.

• Back to Movie Reviews home