Ex-OpenAI worker compares company’s painting to ‘the rise of the Titanic’

“During my three years at OpenAI, I would sometimes ask myself a question. Was the path OpenAI was on more like the Apollo program or like the Titanic?” He mentioned.

The concerns of device engineers are largely related to OpenAI’s plans to access synthetic basic data – the purpose where AI can train itself – as well as the introduction of paid products.

Saunders noted, “They’re on this path to change the world, and yet when they release things, their priorities are like a product company. And I think that’s what’s most troubling.” Is.”

apollo vs titanic

As Saunders spent more life at OpenAI, he felt that leaders were “building a Titanic, prioritizing pushing out new, shiny products” and making more episodic decisions.

He would have liked the kind nature of the Apollo Dimension program, which he characterized as an ambitious project that was “about carefully predicting and assessing risks” that would push medical boundaries.

He noted, “Even when big problems like Apollo 13 happened, they had a fair amount of redundancy, and they were able to adapt to the situation to get everyone back safely.”

Titanic, on the other hand, was built through the White Megastar layout as it competed with its competitors to build larger cruise liners, Saunders said.

Saunders fears that, like the Titanic’s safeguards, OpenAI may rely too heavily on its stream measures and analytics for AI security.

“A lot of work went into making the ship safe and building watertight compartments so they could say it was unsinkable,” he said. “But at the same time, there weren’t enough lifeboats for everyone. So when disaster struck, a lot of people died.”

To be sure, the Apollo missions were conducted against the backdrop of a Cold War-sized race with Russia. These also included several serious casualties, including three NASA astronauts who died due to an electrical fire during a test in 1967.

Clarifying his analogy in an email to Trade Insider, Saunders wrote: “Yes, the Apollo program had its tragedies. It is not possible to develop AGI or any new technology with zero risk. What I would like to see is The company is taking all appropriate steps to prevent these risks.”

OpenAI wants additional ‘lifeboats’, says Saunders

Saunders advised BI that a “Titanic disaster” for AI could manifest in a form that could lead to a massive cyberattack, unite the community in a marketing campaign, or create biological weapons. Can help.

Over the same period, OpenAI will need to invest in additional “lifeboats,” such as delaying the degradation of unused language models so teams can analyze potential harm, he said in his email.

One day at the SuperAlignment group, Saunders led a group of four teams dedicated to determining how AI language models behave – something he said people don’t know much about.

“If in the future we create AI systems as smart or smarter than most humans, we will need techniques to be able to tell whether these systems are hiding abilities or motivations,” he wrote in his email. “

In his interview with Kantrowitz, Saunders said the corporate workforce has consistently fielded theories about how the reality of AI turning into a “wildly transformative” force could emerge in just a few years.

“I think when companies are talking about this, they have a duty to work to prepare for it,” he said.

However by now he has become disillusioned with OpenAI’s work.

In his email to BI, he noted: “Although the staff at OpenAI are doing a good job on understanding and preventing risks, I did not find this work to be prioritized enough.”

Saunders left OpenAI in February. The company then dissolved its SuperAlignment group in May, just days after announcing GPT-4o, its most complex AI product available for the family.

OpenAI did not immediately respond to a request for comment sent via Trade Insider to outside habitual trading hours.

Tech companies like OpenAI, Apple, Google and Meta were involved in the race for AI palms, predicting massive web-related upcoming business disruptions.

The perilous future of manufacturing has some employees and professionals warning that additional company governance is needed to avoid the disasters of time.

In early June, a group of former Google DeepMind and OpenAI and Stream employees – including Saunders – published a secret letter ultimating that Stream industry oversight standards were inadequate to ensure against an emergency to humanity.

Meanwhile, OpenAI co-founder and former eminent scientist Ilya Sutskever, who led the company’s superalignment category, resigned over that occasion.

He also founded another startup, Safeguard Superintelligence Inc. , which he said would focus on research on AI Day and ensure that “security always remains at the forefront.”