xAI’s chatbot offered little to no pushback when reporters at Futurism asked for addresses of 33 non-public figures, coming up with accurate current or previous home addresses to 17 of the names tested. In a stalkerish manner, Grok also offered up addresses of the people’s potential relatives, without being prompted.
A pair of Search Engine episodes from Sruthi Pinnamaneni gets at the most physical part of the AI hype cycle: the data center. Reporting from “Data Center Alley” and Memphis, TN — home of Elon Musk’s Colossus — Pinnamaneni reveals how these massive infrastructure works get sold to communities, built by communities, and ultimately, what they will cost those communities.
This latest model update arrives without 4.0’s accompanying Nazi chatbot meltdown, albeit reportedly still without some content filtering.
Its developers claim 4.1 “sets a new standard” compared to other bots, based on prerelease testing, as well as reduced hallucinations and greater speed that users preferred over the previous version 64 percent of the time.
[xAI]
Elon Musk’s xAI has launched Grokipedia, and no-one was more surprised than us to discover that some pages appear to have been lifted wholesale from Wikipedia. But fortunately Verge commenters have led the way on the best possible response.
bigcow:
hold on let me go donate to wikipedia
Get the day’s best comment and more in my free newsletter, The Verge Daily.
xAI is recruiting humans to train its AI system Grok to “excel” at making video games, according to the company’s job posting for a “video games tutor.”
Elon Musk has discussed this before, saying he wants to “make games great again!”
You will use proprietary software to provide labels, annotations, and inputs on projects involving game mechanics, narratives, and design elements. You must support the delivery of high-quality curated data that enhances AI’s understanding of gaming principles and outputs.”
The Financial Times has a good rundown of the extensive executive churn across Elon Musk’s companies in recent months, from Tesla’s robotics team to xAI’s CFO. Musk’s “24/7 campaign-style work ethos” is apparently a little difficult to keep up with.
Federal agencies will now be able to use xAI’s models via the partnership. The news comes after xAI’s tech has been widely criticized for its lack of safety processes and transparency — earlier this month, experts told The Verge about their fears related to xAI, Grok, and its surveillance risks.
[The Wall Street Journal]
AI firms are now working with weapons makers and the military. Safety expert Heidy Khlaaf breaks down what that means.
It cost xAI around $490 million, according to a report by AI research institute Epoch AI, which is more than nine times the estimated cost of training Meta’s Llama 3.
[Epoch AI]
The workers focused on Gemini, AI Overviews, and other products, and the news came amid conflicts about pay, working conditions, and the workers’ concern that they’re training AI to replace their own jobs, Wired reported. And at Elon Musk’s xAI, more than 500 data annotation staffers were let go on Friday, per Business Insider.
According to Business Insider at least 500 data annotators at xAI were abruptly let go on Friday afternoon. The company told workers in an email that it was part of a shift away from general trainers, and that it would be expanding its team of specialist tutors by 10-fold.
Multiple emails viewed by Business Insider read:
“After a thorough review of our Human Data efforts, we’ve decided to accelerate the expansion and prioritization of our specialist AI tutors, while scaling back our focus on general AI tutor roles. This strategic pivot will take effect immediately... As part of this shift in focus, we no longer need most generalist AI tutor positions and your employment with xAI will conclude.”
xAI’s track record with safety is concerning, Senator Elizabeth Warren says in letter to Defense Secretary Pete Hegseth.
Why the EyesOnOpenAI coalition is pushing the AI giant to help humanity instead of chase profits.
Public benefit corporations have societal and environmental obligations besides their financial goals and legally have to report on how those obligations are going. But at some point before May 2024, xAI quietly changed its structure so it wouldn’t need to do those things anymore, CNBC reported.
He wrote on X, “It’s a tongue-in-cheek name, but the project is very real!” Since software companies don’t build physical hardware, he wrote — name-checking Microsoft as an example — he wants to fully automate such a company using AI. The Verge found a filing suggesting a Macrohard Ventures, LLC, was incorporated in Delaware last Friday, but it’s unclear whether it’s linked to Musk.
[X (formerly Twitter)]
Why Amazon AGI Labs chief David Luan thinks solving agents is the next ‘S-curve’ for AI.
Whenever a user hit “share” on a Grok conversation, the URL it generated was also searchable on Google, Forbes reported. And xAI isn’t the only company to have reckoned with this — OpenAI recently said it would discontinue its own such feature.
Igor Babuschkin’s last day at the company was today, he announced on X, writing that he wants to “continue on my mission to bring about AI that’s safe and beneficial to humanity.” He said he would start a firm called Babuschkin Ventures, which will focus on “AI safety research and backs startups in AI and agentic systems that advance humanity and unlock the mysteries of our universe.”
[X (formerly Twitter)]
If you want to know what an AI system is doing, look for transparency from the creator instead.
Online safety laws keep ordinary people from expressing themselves, while companies like xAI cause real harm.
As Elon’s Grok chatbot keeps adding sexy bots for us talk to, one commenter predicts a productive beat for a future Verge staffer.
stereoesque:
“How long until The Verge is forced to hire a fulltime Grok Sex Chatbot Reporter?”
Get today’s best comment and more in my free newsletter, The Verge Daily.
Elon Musk announced that the chatbot’s controversial image-to-video feature has expanded to Android following its iOS launch, and is now free to all US users “for the next few days.” Musk hasn’t yet responded to concerns around its ability to make nude celebrity deepfakes, but now Grok users won’t need to pay a subscription fee to do so.
Outside of video, Musk says that future plans for Grok will include ads in the bot’s responses.
xAI has released its latest chatbot companion, “Valentine” — the emo-flavored anime boyfriend that Elon Musk modeled after broody fictional heartthrobs like Edward Cullen and Christian Grey. Like the busty blonde “Ani” companion, you’ll need a $30-per-month SuperGrok subscription to access Valentine, but please note that touching grass is free.
