The First Horseman of the AI Apocalypse

Published: 2023-01-21
Tagged: readings

02/06/2023: It occurred to me that the original didn't treat the topic seriously enough. This, then, is another stab at it. The old version is at the bottom of the page.

Robots writing poetry

"DoNotPay, the company that bills itself as “the world’s first robot lawyer,” is launching a new AI-powered chatbot that can help you negotiate bills and cancel subscriptions without having to deal with customer service. (...) The chatbot cites problems with Xfinity’s services and threatens to take legal action, to which the representative responds by offering to take $10 off the customer’s monthly internet bill."

Finally! Average mortals like you and I can now enlist assertive and tireless digital mercenaries to deal with customer service.

How much of our precious time have we burned listening (always listening) to the metallic, raspy tones of corporate jingles? Or those recordings telling us what an important customer we are and to please (please) hold because the estimated waiting time is 4 hours and 13 minutes? Or the soul-dead voice of overworked call center contractors guiding us through troubleshooting steps that have nothing (nothing) to do with out problem?

If we could really get this time back, it would be a great victory for us faceless Joes and Janes. Such a tool would turn Kafka's "The Castle" into a short, cheery little story.

But all technology eventually leads to unintended consequences. The designers of the Transmission Control Protocol, the protocol that has powered the Internet since about the 70's, never imagined it would be used to give life to lolcats or 4chan.

One such scenario I can imagine robot lawyer AI leading to is this:

Faced with such an epoch-defining event, what would be humanity's collective reaction?

I suspect it would be the usual mixture of fear, anger, curiosity, grief, awe, and ecstasy. Because if you look at the last 100 years or so, where almost every decade produced some sort of civilization-redefining miracle--from flight to nuclear energy, from vaccines to the Internet--you'll notice that it always takes less than a single generation to transform wonders of progress into everyday nothingburgers.

Adaptation. That's what we excel at as a species. It's also the reason why I think that, ultimately, nothing serious will change after robot lawyers and other types of AI become commonplace. But at least we'll be freed from the hellish experience of dealing with customer service.


This is the original post, published on 01/21/23

"DoNotPay, the company that bills itself as “the world’s first robot lawyer,” is launching a new AI-powered chatbot that can help you negotiate bills and cancel subscriptions without having to deal with customer service. (...) The chatbot cites problems with Xfinity’s services and threatens to take legal action, to which the representative responds by offering to take $10 off the customer’s monthly internet bill."

This is pretty cool and potentially very significant. I think we'll see companies switch to AI-powered customer service pretty quickly given that it can be scaled quickly and on the cheap. Then we'll have bots owned by some companies talking with bots owned by some other companies.

Companies that would use chatbots are the type that would also collect data on everything they do. That helps the MBAs and others in charge to decide how to allocate resources. It would be nothing out of the ordinary that whoever is in charge of customer service would review some numbers and seek out ways to decrease costs by, for example, making the bots more aggressive so they hand out fewer discounts.

But the same thinking is going on the other side, the side running the consumer-assistance bots. So these guys will tweak their bots to be a little bit more aggressive in demanding discounts.

It's a race to the bottom. Except, at some point, someone clever will get the bots to tweak their own parameters because that's just more expedient. Then it's a gradient descent to the bottom.

What if, through pure luck, bots from one side discover an advantage by producing rap lyrics? Or sonnets? Or detective novellas? Through relentless adversarial optimization they will start producing increasingly beautiful prose and poetry.

Then one day some bored engineer will decide to inspect what all these bots are writing to each other, only to discover the most inspiring and soul-wrenching writing ever to come into existence. They will, of course, leak it to the media. Briefly, there will be discussion whether this is real or moral or whatever. But then most ordinary people will switch to consuming this new type of content and never look back. The art world will fade away with barely a whisper. War, too, will cease. History will truly end this time.

Or maybe the neural nets and LLMs that power the bots will take a different path and produce increasingly random gibberish. Then the bored engineer who ventures to peek inside the black box will discover, say, endless exchanges composed solely of varying sequences of the letter "K."

Well, whichever way it goes, one thing I'm sure of is that we're on the cusp of a shift as significant as the appearance of semiconductors was in the last century.

Comments

There aren't any comments here.

Add new comment