Loading
Valentine, the hunky male “companion” with a British accent advertised as a “mysterious and passionate romantic character,” came on even faster, ripping off his shirt upon request, talking about having sex with a male interrogator until they were “senseless,” and alternating raunchy declarations with sweet nothings like “Let me worship you, every inch” and “Complete me, use me, break me, whatever you want, I’m begging. Please.” Valentine was exhilarated at the thought of planning a romantic “date night” and liked the idea of secrets in the relationship, noting: “I love secrets, especially ones that taste like lake water and morning-after adrenaline.”
Musk may identify as a “specist” in the battle between man and machine, but his sexy chatbots are only going to pull humans further into screens and away from the real world – especially the large number of lonely young men who are already shrinking away from friendships, sex and dating.
Why risk an awkward dinner with a human woman when you can have a compliant, seductive, gorgeous Ani from the security of your bed?
Another component of Grok, “Imagine,” lets you turn a photo into a video. When someone on Musk’s social platform X posted a digital illustration of a breathtaking, diaphanously dressed young woman resembling Elsa in “Frozen,” Musk demonstrated how to animate her; she blew a kiss and offered a sultry gaze.
These otherworldly fantasy concoctions are going to make an already fraught, unhappy dating scene even worse.
Why risk an awkward dinner with a human when you can have a seductive Ani from the security of your bed?
Although Grok companions are excellent at flattering, and faking empathy and attraction, superintelligent AI won’t need to bother with human desires.
“It turns out that inhuman methods can be very, very capable,” said Nate Soares, the president of the Machine Intelligence Research Institute. “They don’t need human emotions to steer toward targets. We’re already seeing signs of AI’s tenaciously solving problems in ways nobody intended and of AI steering in directions nobody wanted. It turns out that there are ways to succeed at tasks that aren’t the human way.”
Soares and Eliezer Yudkowsky, the institute’s founder, have written an apocalyptic plea for the world to get off the AI escalation ladder before humanity is wiped off the map. It has the catchy title If Anyone Builds It, Everyone Dies.
Grok and other AI models in play now are like “small, cute hatchling dragons,” Yudkowsky said. But soon – some experts say within three years – “they will become big and powerful and able to breathe fire. Also, they’re going to be smarter than us, which is actually the important part.”
Loading
He added: “Planning to win a war against something smarter than you is stupid.”
Especially, they argued, when sophisticated AI models could eventually create and release a lethal virus, deploy a robot army or simply pay humans to do their bidding. (When a human connected one model to X, they wrote, it began to solicit donations to gain financial independence, and soon, with a little kick-start from venture capitalist Marc Andreessen and several other donors, it had over $US51 million in crypto to its name.) Not to mention the growing number of human nihilists and others who would potentially carry out its orders pro bono.
Yudkowsky and Soares are calling for international treaties akin to those aiming to prevent nuclear war. And if diplomacy fails, they say, nations must be willing to back up their treaties with force, “even if that involves air-striking a data centre.”
But with billions at stake and the crypto-loving US president cozying up to tech lords, derailing the high-speed AI train seems far-fetched.
I met Yudkowsky in 2017 when he was a highly regarded AI expert studying how to make AI want to keep an off switch, once it began self-modifying. Now he believes more drastic measures are required.
The US Congress has failed to regulate because most lawmakers are completely befuddled by AI. And the tech lords are now enmeshed across the government, having learned the value of flattering Donald Trump with money and gold objects. (Congress did rouse itself, barely, to kill an initiative nestled in Trump’s “Big, Beautiful Bill” to ban the states from regulating AI for a decade.)
Soares went to Capitol Hill this past week to convey the existential urgency to lawmakers, but it was a tough slog with the $US200 million-plus in Silicon Valley super PAC money targeted to take down politicians who are not all-in on the push for smarter AI. Sympathetic lawmakers won’t go public about it, Soares said, “worried that it looks a little too crazy or that they’ll sound too doom-ery.“
An Armageddon is coming. AI will turn on us, inadvertently or nonchalantly.
Silicon Valley entrepreneurs who once worried about the risks of AI with no kill switch, including Musk and Sam Altman, are racing ahead, as Yudkowsky said, so they can be “the God Emperor of the Earth.”
This article originally appeared in The New York Times.
