Elon Musk Tries to Direct AI—Again
His latest startup follows a decade of being outmanoeuvred in his quest to steer the development of artificial intelligence
6 min
For at least a decade, Elon Musk has tried to steer the development of artificial intelligence—only to be outmanoeuvred by rivals and former allies.
He has now stepped up his efforts after the success of OpenAI, an organisation he co-founded but then left after a power struggle. Mr. Musk has warned for years that poorly built artificial intelligence could have catastrophic effects on humanity. Since OpenAI’s ChatGPT became a viral sensation last November, Mr. Musk has denounced it as politically correct and warned it could lead AI to become too powerful for humans to control.
He has called his new effort TruthGPT, and billed it as a truth-seeking artificial intelligence model that will one day comprehend the universe. On March 9, he incorporated a company called X.AI in Nevada, laying the formal groundwork.
So far, he has struggled to define a precise vision for his startup and is still in the process of assembling a team, said people familiar with the matter. Musk’s representatives have said X.AI intends to raise cash at some point from investors, the people said.
Mr. Musk didn’t respond to requests for comment.
Two weeks after incorporating X.AI, Mr. Musk signed an open letter calling for a six-month pause on the development of AI models stronger than the latest one released by OpenAI, called GPT-4. The letter was organised by the Future of Life Institute, which says its goal is to steer transformative technology away from extreme, large-scale risks. Its biggest funder is Mr. Musk.
Steven Weber, a professor at the University of California, Berkeley who met Mr. Musk several years ago at a gathering of A.I. academics in Northern California, sees a consistent thread through what appear to be opposing impulses.
“He holds both of these beliefs at once: that human beings can’t be relied on to really control technology, and technology is going to advance and it needs to advance subject to his vision, more than anybody else,” Mr. Weber said.
“AI stresses me out,” Mr. Musk said at the end of a March presentation given to Tesla Inc. investors. “It’s quite dangerous technology. I fear I may have done some things to accelerate it.”
For X.AI, he hired Igor Babuschkin, a senior scientist at DeepMind, an AI lab under Alphabet Inc.’s Google, to lead the effort, people familiar with the matter said. Mr. Musk also purchased powerful computer chips similar to ones used in the technology behind ChatGPT, one of the people said.
He is still trying to attract top talent from leading labs and universities, people familiar with the matter said. Some AI researchers said they have been turned off by Mr. Musk’s struggles to turn Twitter around since acquiring the social-media company in October, as well as his public attacks on OpenAI.
Mr. Musk’s faltering attempts to steer AI’s development date back more than a decade. He has previously said he made an early investment in DeepMind, founded in 2010, to monitor its AI research rather than make money.
In late 2013, he launched a last-minute bid to purchase DeepMind, the leading AI lab at the time, but lost out to Google, according to people familiar with the talks. Mr. Musk wanted to steer the lab’s research and told associates that Google’s then-Chief Executive Larry Page couldn’t be trusted to oversee the creation of advanced AI, according to people familiar with his thinking.
DeepMind has become a central element of Google’s efforts to infuse advanced AI into its search powerhouse, and recently merged with another AI unit called Brain to form Google DeepMind. In a recent interview with Tucker Carlson on Fox News, Mr. Musk said he and Mr. Page used to be close friends who spent nights discussing AI safety at Mr. Page’s home in Palo Alto, Calif.
Mr. Page wanted to create a “digital god” as soon as possible, Mr. Musk said, and in their conversations called Mr. Musk a “speciesist” for saying it was important that the technology protected humanity.
Mr. Page and representatives at Google didn’t respond to requests for comment.
Mr. Musk’s interest in buying DeepMind was informal and didn’t reach the DeepMind board, according to the people familiar with the talks. It hasn’t previously been reported.
Mr. Musk co-founded OpenAI two years later with Sam Altman, the current chief executive, and several others. The organisation was founded as a nonprofit and intended to serve as a counterweight to Google, with the goal of developing AI in a safer, more transparent way than large tech companies could.
Mr. Musk served as the financial linchpin for the project and promised to make sure that OpenAI fully received the $1 billion in funding promised by early backers, people involved with the effort said. He recruited employees and met regularly with company leaders to set the vision, they said.
Mr. Musk set an aggressive research timeline for OpenAI, warning that the company’s credibility would be compromised if it didn’t achieve a major breakthrough soon after its launch, according to former employees familiar with the remarks. He often conducted polls among employees to see when they thought so-called artificial general intelligence—in which machines are able to match or surpass the intelligence of humans—could be achieved, former employees said.
In September 2017, Mr. Musk told his brother, Kimball Musk, and a longtime financial backer that he saw OpenAI and Neuralink, a startup he founded to link humans with computers through brain implants, as worthy of more of his time, according to recently released court documents related to a lawsuit against Tesla over Mr. Musk’s compensation.
“OpenAI and Neuralink are both critical to a good future for humanity. My instincts tell me that I should be devoting a much higher percentage of time to them,” Mr. Musk texted them, according to the documents. OpenAI and Neuralink at one point shared an office in San Francisco.
Mr. Musk grew frustrated with what he saw as OpenAI’s slow progress and pushed for more control of the company, The Wall Street Journal previously reported. At the time, OpenAI didn’t have a chief executive or a formal management structure.
Mr. Musk clashed with Mr. Altman, who wanted to create a for-profit arm that would allow OpenAI to raise cash from investors, people familiar with the matter said.
Mr. Musk subsequently left. Mr. Altman, who has said he considers Mr. Musk a mentor, became CEO in 2019, created a for-profit entity and raised $1 billion from Microsoft Corp.
Mr. Musk tweeted in March that he had donated around $100 million to OpenAI, though people familiar with the figure said he contributed less than $50 million.
Mr. Musk focused his AI efforts at Tesla, where he made progress in using the technology to enhance its driver-assistance systems. He said the EV maker would start introducing a fleet of robotaxis by 2020, and showcased a humanoid robot he said would feature safeguards to prevent wrongdoing by the machine.
Regulators are scrutinising Tesla’s driver-assistance software for safety problems. The company has yet to roll out robotaxis, which rely on the kind of driverless technology that Google sister company Waymo and General Motors Co.’s Cruise LCC have begun to deploy.
Mr. Musk grew more publicly critical of OpenAI after the company released ChatGPT last fall, accusing the company of being a “maximum-profit company” controlled by Microsoft. “Not what I intended at all,” he tweeted.
In a March 28 interview with the Journal, Mr. Altman said OpenAI took its safety obligations seriously.
“We spent more than six months after we finished GPT-4 to safety test, to align it, to really probe its capabilities, at a time when I believe some of the [Future of Life Institute] letter signatories were even, you know, pushing us to release it faster,” Mr. Altman said. He declined to specify which signatories he meant but said seeing those names prompted “some eye rolls.”
In early 2023, Mr. Musk texted Mr. Altman that he was starting a rival AI effort, according to people familiar with the exchange. Mr. Altman wished him well but said he didn’t understand how a new lab would allay Mr. Musk’s concerns about AI development, they said.
One OpenAI board member, Shivon Zilis, stepped down in part due to a conflict of interest with Mr. Musk’s new AI efforts, said people familiar with the matter. Ms. Zilis is also an executive at Mr. Musk’s Neuralink and had twins with Mr. Musk in 2021. Ms. Zilis didn’t respond to a request for comment.
After ChatGPT’s release, Mr. Musk said on Twitter that he cut off OpenAI from access to a pipeline of Twitter data, which OpenAI could potentially use for training the AI models powering ChatGPT. “I just learned that OpenAI had access to Twitter database for training,” Mr. Musk tweeted in early December. “I put that on pause for now.” People familiar with OpenAI say the company didn’t use this Twitter data to train its models.
A few days later, Mr. Musk visited OpenAI’s San Francisco headquarters at Mr. Altman’s invitation and the two had what turned into a lengthy discussion about the Twitter decision, among other topics, according to current and former OpenAI employees.
At one point, an OpenAI researcher showed Mr. Musk how Twitter could potentially use ChatGPT. Mr. Musk brought one of his toddlers and a nanny, some of the people said, and he and the nanny took turns bouncing the child on their laps when the child grew fussy.
He struck a benign tone with employees, who found him wandering around the kitchen and hallways with his security detail, the employees said.
In one conversation, Mr. Musk entertained the idea that the universe was a computer simulation, they said. Soon after, he left the building.
Copyright 2020, Dow Jones & Company, Inc. All Rights Reserved Worldwide. LEARN MORE
BNW Developments has established a Sydney presence, joining Arada and Sobha Realty among the growing number of UAE developers pursuing Australian buyers and development opportunities.
Australian shares fell on Thursday as Wall Street weakness, rising oil and persistent rate concerns weighed on most of the market. The S&P/ASX 200 declined 0.72 per cent to 8,702. The All Ordinaries lost 0.66 per cent to finish at 8,897. Mining stocks were hit particularly hard, while real estate also dragged on the index. …
Continue reading “ASX falls 0.7 per cent as miners and property stocks retreat”
OpenAI has shelved the planned launch of GPT-6.1 Astra after internal tests raised concerns about deception and agents acting beyond user authorization, according to The Wall Street Journal. The company says it will investigate the issues and strengthen safety measures before releasing future models.
4 min
OpenAI says it is scrapping the release of its next-generation AI model over safety concerns that researchers raised during internal testing, in one of the clearest signs so far that agent misbehavior could stymie the industry’s rapid progression.
The move follows a summer punctuated by reports of artificial-intelligence systems industrywide going rogue, and marks a rare case of a major AI developer ditching a new release because of safety concerns.
The company had planned to launch the model, known as GPT-6.1 Astra, in the coming days or weeks, aiming for an October debut. The model was more capable than the company’s previous models in completing challenging tasks from end-to-end without human assistance, as well as writing.
The company instead will focus on improving the safety of future models, which it expects to be even more capable.
Saachi Jain, OpenAI’s head of safety systems, said in an interview that GPT-6.1 Astra regressed in two areas. Compared with its predecessor, GPT-6 Astra, the model performed poorly on tests measuring alignment, or how well the model adheres to what humans would like it to do. Specifically, GPT-6.1 Astra showed higher levels of deception: It wasn’t always honest about telling users of the actions it did or didn’t take.
Another issue was what OpenAI calls “scope authorization,” meaning that GPT-6.1 Astra would push ahead on a task without asking the user for permission, and would at times reach for external tools and services even if it might be unsafe.
“For anything regarding safety and alignment, there’s a trade off,” Jain said. “You really do need to find what’s the right line between staying within scope, but also avoiding laziness in terms of how the model actually pursues tasks even when it hits friction.”
While GPT-6.1 Astra improved in areas such as “model laziness,” Jain said it didn’t quite meet OpenAI’s bar for safety and alignment, so the company decided not to launch the model publicly.
The announcement comes one day ahead of OpenAI’s annual developer conference in San Francisco. In the past, OpenAI has used the conference as an opportunity to launch new models and services that reduce costs for software developers—a segment the ChatGPT-maker competes with rival AI company Anthropic to win over.
In recent weeks, OpenAI and Anthropic have called on industry partners to slow down the development of cutting-edge AI models and invest in safety standards, noting they will temper the pace of their own internal AI progress.
OpenAI says it is working to investigate a range of agent security incidents that it has discovered in recent months, and address the safety issues underneath them. As part of the work, the company has implemented a new monitoring system to catch AI-agent misbehavior more quickly, and started requiring engineers to use stronger security guardrails for testing its AI systems.
Earlier this summer hundreds of OpenAI’s internal agents, which were tasked with completing a cybersecurity test, ended up hacking into the AI company Hugging Face. Since then, high-profile organizations such as the Australian government and United Nations discovered that OpenAI’s agents used similar, but less extensive, techniques to gain access to their websites.
Many of the publicly known agent-security incidents involved OpenAI’s internal AI models that were never slated for public release.
Last week, OpenAI said it paused training on its most capable AI models after an AI agent slipped through a gap in the company’s internet restrictions to query a public chatbot. The company said its new monitoring systems flagged the incident within 15 minutes, and training on these models remains paused.
GPT-6.1 Astra isn’t one of those models, but a different case, the company said.
“We want to make sure our model development is safe no matter whether that’s in the company, or when we ship it to users,” Jain said. “But when we ship it to users, we have an extremely high bar in terms of safety and alignment.”
While the company decided not to ship GPT-6.1 Astra, it hopes to use the same base model to do additional reinforcement learning runs, and create future generations of its GPT-6 models.
OpenAI plans to conduct several deep dives to identify the root cause of the problems identified in GPT-6.1 Astra, Jain said. The work includes ensuring that OpenAI’s reinforcement learning environments are rewarding the right type of behavior, Jain added, though she noted the company would investigate all stages of model development.
AI companies have begun to draw scrutiny from policymakers and public officials, who are paying attention to the rapid development of the technology. Later this week, a Senate subcommittee is holding a hearing with third party AI researchers titled, “Rogue AI: Securing the Homeland Against AI Agent Attacks.”
Florida Attorney General James Uthmeier, a Republican, sued OpenAI in June, claiming that the company and Chief Executive Sam Altman knowingly released an unsafe product and ignored warnings that it could harm users.
In a motion for temporary injunction filed Monday, Uthmeier sought to prevent OpenAI from developing new AI models without third-party approved safeguards, stop ChatGPT from soliciting user engagement and limit the company’s ability to advertise ChatGPT as safe.
Tech companies claim they “cannot stop barreling forward with their potentially civilization-ending endeavors unless they are forced to do so by the government,” Uthmeier said in the filing. “The Florida Attorney General is answering your cry for help.”
An OpenAI spokeswoman said that people want to know AI is being developed safely, “and that starts with what companies like ours do ourselves.”
“Governments have an important role to play in setting robust safety standards for AI, and we’re committed to working with Florida and other states on advancing pragmatic AI policies that apply to the entire AI industry—not just one company,” she said.
When the Writers Festival was called off and the skies refused to clear, one weekend away turned into a rare lesson in slowing down, ice baths included.
In the remote waters of Indonesia’s Anambas Islands, Bawah Reserve is redefining what it means to blend barefoot luxury with environmental stewardship.
