Photo: Intelligenmer; Photo: Getty Images
In Early August, an anonymous x Account Started Making Big Promises About He. “(Y) ou’re about to taste agi,” it clalimed, adding a Strawberry emoji. “Q* is a project a project. It is a portal. Altman’s Strawberry is the Key. The single isn’t coming. “Tonight, we evolves.”
If you’re not part of the hyperactive cluster of he fans, critics, doomers, accelements, grifters, and rare insists that have congregated on x, Discord, and reddit to specules the futures, this probably tears likensense. If you are, it might have sounded enticing at the time. Some background: agi stands for artificial General Intelligence, a term used to describe humanlic ability in it; The Strawberries Are A Reference to An Internal codename for a rumored “reasoning” Technology Being Developed at Openai (and a post from Openai Sam Altman minutes before that included a Photo of Strawberries); Q* is eather a previous codename for the project or a Related Project; and the singulariy is a theoretical point at which he or technology more broadly Becomes self-improving and uncontrolble. All this coming tonight? Wow.
The specificity of the Account’s Many predits got the attention of he influencers, who Speculated about who might be and whereter it might be an it isself-a next-Gennection openai model-leevel is drumming up publicity for its creator. Its posts Broke Containment, Howver, AFTER AN UNSOLICIED REPLY FROM SAM ALTMAN HIMSELF:
For a certin type of highly receptive he enthusiast, this typped a plainly absurd scenario-an amateurish, meme-still anonymous account with the guy from Time As an avatar announcing the Arrival of the post-Human age-Int plausibility. It was time to celebrate. Or was it to panic?
Neither, it turns out. Strawberry Guy didn’t know anyding. HIS “Leaked” Release Dates Came and Wentand the community started to turn on Him. On Reddit, The /R /Singulary Community Banned Mention of the Account. “I was duped,” wrote seo and peripheral he influencer Marie Haynes in a postmorem blog post. Before she knew it was a fake, she Said, “Strawberry Man’s Tweets were starting to freak me out.” But, she concluded, “It was all for good reason … we truly are not prepared for what’s the coming.”
What was Coming Was another Mysterious Account, Posting Under the name lily ashwood, who starts appearing in live voice discussions about it on x Space. This time, the Account didn’t have to work very hard to get people theorizing. He enthusiates who has Had gathered to talk about what they have thight might be the immenent release of gpt-5-some of the whom were starting to suspect they dupa hersel Be an he, spread uses Openai’s Voice Technology and an unrelessed model. They zoneed in her manner of speaking, her Fluent Answers to A Wide Range of Questions, and Her Caginess Around Certaine Subjects. They attempted to out het, prompting her like a chatbot, and coming away uncertain of who, or what, they were talking to.
“I Think I Just Saw a Live Demo with GPT 5,” wrote one user on reddit, after jaining an x Space with ashwood. “It ‘unbellievable how good she is. It almost fools you into believing it a human.” Egged on by Strawberry Guy, Others Started Collecting Evidence That Ashwood was agi in the wild. Openai Researchers HAD JUST CO-SIGNED A Paper Calling for the Development of “personhood credentials“To” Distinguish Who is Real Online. ” COULD THIS BE PART THAT STUDY? clueright there in the middle: iLea, as in ilya sutskever, the openai co-founder who left the company after clashing with Sam Altman over he safety. Besides, JUST LISTEN TO THE “SUSPICIOUS NOISE-GAting” and the “Unusual Spectral Frequency Patterns.” Superintelligence was right there in front of me, chatting on x:
It wasn’t. Ashwood, who declined a Request for Comment, Described Herself as a single Mother from Massachusetts and Releas a video poking fun at the episode (Which, of Cours, Some observers took as stovence and he was ai):
But by then, evening prominent he influencers had gotten caught up in the drama. PLINY The Liberator, An Account that Gained A Large Following by Cleverly Jailbreaking Various AI Tools – Manipulating say to reveal information about how they work, and breaking down Guardrails put in place by their creaters – was briefly convinced that ashwood have ben he. He Described the Experience as Psychologically Taxing, and Held a Debriephing With His Followers, some of the WHOM Angry About What He, An anonymous and Sometimes Trollish Account that they’d ‘ Allowed say to Believe:
On One Hand, Again, This is Easy to Dysmiss from the Outside: A Loose Community of People with Shared Intuitions, Fears, and Hopes for a vaguely defined technology are Working therselves ancelves a fare in insular spaces, manifesting predictions as the predictions ( Longer Than Expective). But the Very First Chatbots, Who Wold Today Be Instantly Recognizable As Inert Programs, Were Psychologically Disorienting and Distressing Wen They First Arrived. More recently, in 2022, a Google Engineer quit HIS JOB IN PROTEST AFTER BECOMING CONVINCED THAT AN INTERNAL CHATBOT WAS SHOWING SIGNS OF LIFE. Since then, millions of People have interactions with more sophisticated tools than he has had access to, and for at least some of the say, IT’S SHAKEN LOOSE. Pliny, who recently received a Grant in Bitcoin FROM Venture Capitalist Marc andREESSEN FOR HIS WORK, WONDERED IF OPENAI’S RELEASE SCHEDULE HAD BEEN SLOWED “Because Suficiently Advanced Voice it has the potential to induce psychosis,” and made a prediction of His Own: “Fair to Say We’ll See the First MedicaLY DOCUMENTED CASE OF AI-NIGHT Psychosis by December?” HIS FOLLOWERS WERE UNIMPRESSED. “Nah, i was hospitalized back in 2023 AFTER GPT CAME OUT… 6 MONTHS STRAIGHT, 7 DYS A WEEK, LITTLE TO NO SLEEP,” Wrote One. “Hahaha i made it in august,” Wrote Another. Another Asked: “Have you Not Been Paying Attention?”
The fundamental claim here – that he systems that can can talk like people, sound like people, and look like people might be able to fool or manipulative actual people – is fairly unconroversial. Likewise, Its Reasonable to assume, and worry, that comprehensivly anthropomorphized products like chatgt Cynically or inadvertently exploit users’ Willingness to personify say in Harmful Ways. Nor Postible Triggers for Latent Mental Illnessit would be difficult to come with something more on the nose say machines that pretend to be humans by people who talc in riddles about the apocalyps.
Setting aside the plausibility or human-level autonomous intelligence, there are plenty of narrow contexts in which he is already in simillary misleading ways, to defraud People by Cloning Family Members’ Voices, Or by Simply Posting Automated Missinformation on social media. Some Openai Employees, spread sensing that things are getting a little bit out of control, have posted calls to, basically, calm down:
Whether Companies Like Openai Are on the Cusp of Releasing Technology That Matches Their Oracular Messaging is Another Matter Entirely. The Company’s Current Approach, Which Involves Broad Claims that Human-Level Intelligence is Immint, cryptic Posts from Employees Who Have Become Micro-Celebrities Among AI Enthusiasts, and Secrets About their Actual Product Roadmaps, has been effective in building hype and anticipation in ways as materially benefi openai. Much Bigger Models Ares, and they’ll Blow Your Mind is exactly what increasingly skeptical investors need to be hear in 2024, especally now that gpt-4 is approaching two years old.
For the online community of People unsetled, excited, and obsessed by the last few years of he Development, though, this growing space between big he is creating a Speculative void, and it is a starting, and a little bit of Fraud. They were promised robots that coulud trick us into thinking they’re human, and that the world would bes the sun. In the meantime, they’ve settled on the next best thing: tricking each Other.
اترك تعليقاً