Sunday, April 27, 2014

Movie / Transcendence (2014)

MOVIE / TRANSCENDENCE (2014) with Johnny Depp

As an AI researcher and ethicist, I found this film an interesting "first stab" at addressing some issues of strong AI, but with enough problems that experts would find it infuriating, while the general public would likely get bored. I hope that other film makers can learn from this initial attempt and make a better, more informative film.

SPOILER ALERT: THIS REVIEW CRITIQUES MOST OF THE PLOT!

The film does a decent job of setting up the basic issues. The AI researcher, an academic divorced from reality, makes a bunch of loose (albeit potentially true) public statements about how strong AI will eclipse humanity, triggering the rise of an anti-AI terror group that kills him and many other researchers. However, as he is taking a while to die, he and his wife decide to "upload" his mind, just as they recently did with a monkey.

Minor nit: Why would the monkey supposedly (as told by the terrorist) have been screaming in pain, from some electrodes sucking up his or her neural impulses? (Maybe it was freaked out to find itself inside the computer. If so they should have waited for it to calm down.)

Problem 1: Is uploading going to be a practical first pathway to strong AI? I doubt it. Uploading remains highly speculative and futuristic. The camera pans over whiteboards of equations, but there is zero insight into how we can effectively mimic the organization of the mind, much less immediately go way beyond it.

Minor nit: Satellite communication speeds are too low to transfer his mind and related software before the terrorists shoot the place up. His wife should have whipped out a credit card and FTP-ed him to the Amazon EC2 cloud!

Problem 2: Once his mind is uploaded and "reorganizes" itself, it immediately begins to display super intelligence, illegally infiltrates the entire Internet, makes millions in the stock market, becomes a master of medical and environmental nanotech, and so on.

But wait, the financial markets already have tons of trading programs, the Internet has some degree of security (and can't just be taken over), and reorganizing a super-AI is a hard problem. Just because our hero has been uploaded doesn't mean he can suddenly solve or break into anything.

Minor nit: Altho there is media coverage of the assassinations, the film punts on trying to depict any further media or public response.

Problem 3: Once he has built his own data center with stock market loot, his wife, who had been his most ardent booster, singing the praises of bright futures from AI, also has (IMHO) some kind of brain change, and turns into a dull witted love interest, with the usual shots of her looks. She should have been saying holy cow, look at all these life affirming miracles, let's get on the road and promote them!

Problem 4: She and everyone else become so dull witted they don't believe what they're seeing, and the super-mind fails to see the storm coming. Now that his mind is way beyond theirs, she starts to feel alienated and uninvolved, and others question whether there's anything left that's really him.

IMHO This is like asking if dogs can really feel pain, while going ahead with torturing them. Of course they can, you moron. But perhaps it is realistic that society will reject anything it perceives as "other," and decide to kill it, no matter how self-defeating that might be.

Finally, despite that the super-mind is literally Jesus Christ, healing the sick, bringing the dead back to life, conferring super abilities, and restoring the planet, just as the wife had extolled in her former life, she and the government (teaming with the terrorists for deniability) decide they've got to terminate him.

There's an homage to "The Day the Earth Stood Still," as he carries her wounded body into the data center, which had me intoning "Gort, klaatu barada nikto."

Problem 5: Another AI scientist takes some of the nano program from one of the healed people and uses it to create a "virus" that can take down the super-mind. a) This conflates biological and computer viruses, and b) if it was based on his own code, why did they need to kill (or even touch) the healed person to get it?

Minor nit: How killing off a rogue AI with a virus will permanently take down the entire US power grid is left as an exercise for the viewer, as there's no justification whatever for such an outcome.

Okay, that's enough criticisms for one review. No doubt other AI and nanotech experts can find many more things to complain about. But if you've ever wondered if Ray Kurzweil might get shot for predicting that AIs will eclipse humanity, we get an exploration of that scenario.

0 Comments:

Post a Comment

<< Home