OpenAI Logo

OpenAI: You’re not Fired!

One of the more dramatic corporate sagas played out at OpenAI over five November days. On the 17th, the AI company sacked talismanic co-founder and CEO Sam Altman. Cryptically, its board members said Altman was not “consistently candid in his communications.” The move was hugely unpopular with OpenAI staff and its largest investor Microsoft. CEO Satya Nadella announced Altman would join his company to lead a new AI research team. However, internal and public pressure saw OpenAI perform a volte face. The board was out instead and a new line-up reinstated Altman as CEO on the 22nd. Adam D’Angelo continues on this new board alongside Brett Taylor and Larry Summers. In a concession to the old guard’s concerns, Altman and co-founder Greg Brockman will not return to the board despite re-joining the company.   

No-one is yet sure what these concerns were that led to Altman’s removal. But speculation has focused on the inherent conflict in OpenAI’s structure – a “capped profit” company governed by its original nonprofit board. Its mission is to develop Artificial General Intelligence (AGI), a system that matches or outperforms human intelligence. The nonprofit’s goal is doing so safely while the business has to do it quickly. Rumours abound that a new development, possibly ‘Project Q*’, spooked the board from this safety perspective. We look at the foundations of this schism and how it affects OpenAI’s race to AGI. Finally, with an upcoming share sale looming, we examine whether the drama has strengthened or weakened OpenAI’s hand.         

THE BEGINNINGS: DEEPMIND AND OPEN AI

Walter Isaacson’s recent biography of Elon Musk traces OpenAI’s beginnings to Musk’s fallout with Google’s Larry Page. Larry Page is described as indifferent to the prospect of machines someday surpassing humans in intelligence. “Musk argued that unless we built in safeguards, artificial intelligence systems might replace humans, making our species irrelevant or even extinct.” Musk was dismayed when Google acquired AI research laboratory Deepmind. In 2015 he co-founded OpenAI with Altman as a counterweight to Google’s acquisition. “The goal was to increase the probability that AI would develop in a safe way that would be beneficial to humanity.”    

In 2018 Musk sought to convince Altman that OpenAI should be folded into Tesla. When OpenAI’s leadership team rejected his proposal, Musk left the venture and Altman started a for-profit-arm. In 2019 OpenAI officially transitioned to a “capped profit” model where profit was capped at 100 times any investment. However, the nonprofit entity remained the sole controlling shareholder of the company. This is what gave the board the power to remove Altman.

DOOMERS VS BOOMERS

To understand why it exercised this power, one has to grasp the ongoing schism in the tech world. We discussed this previously in “AI: Enigma at Bletchley Park”. So-called “boomers” minimise existential concerns about AI development and emphasise the potential for progress. “Doomers” are more concerned about these risks and want stricter regulation. Karen Hao is writing a book on OpenAI and spoke with Azeem Azhar about the implications of this struggle within the company. She describes Altman as somewhat “middle of the road” on this spectrum. It’s not really “his main concern or fixation”. But the board’s duty is to worry about this and act from more cautionary principles. 

Was the board concerned with OpenAI’s general direction of travel or something more specific? Ilya Sutskever, a former board member as well as co-founder and Chief Scientist, voted Altman out but later backtracked. Writing on X, he said he regretted his participation in the board’s actions. He offered no insight as to what influenced his original decision nor whether the issue in question had been rectified. It prompted Musk and venture capitalist Marc Andreessen to unsuccessfully probe what it was that prompted the initial sacking.      

AGI BREAKTHROUGH?

Internet sleuths are pointing the finger at ‘Project Q*’. Reuters reports that some at OpenAI believe the project represents a significant breakthrough in the quest for AGI. It’s reported that the new model can solve certain mathematical problems. This suggests an ability to reason rather than simply form patterns from large data sets. Such an advance would put daylight between OpenAI and its competitors. As a business, it would want to commercialise this discovery as soon as feasible. But its board has to think more deeply about its responsibilities. Is society ready for such a change? 

Others in the know suggest even more seismic developments are at play. Pippa Malmgrem is a technology entrepreneur, economist and former Special Assistant to President George W. Bush. In an article for UnHerd, Malmgrem claims that Altman’s plans may upset the world order. She writes about the partnerships he is forming with the Emiratis and Chinese in developing a “new supply chain” that “would not only challenge America’s IT infrastructure; it could be facilitating the diminishment of US power. As the US tries to slow down international technological innovation through restricting chip sales, Altman is challenging its hegemony as he announces plans to build his own ecosystem. Malmgrem says “Altman does not believe in borders. He has one goal: to build the best AI possible.” She concludes by suggesting Larry Summers’ new appointment on the OpenAI board is an attempt to at least wrestle back some government influence in this venture. 

AI ARMS RACE

Malmgrem’s warning echoes familiar concerns about the power of big tech. These companies are richer and more powerful than most countries. And, like countries, they compete with each other. OpenAI is the most famous name in the AI space. But it has a growing list of competitors. These range from recent start-ups like Stability AI and Anthropic to Google, Amazon and Baidu. OpenAI’s status as the first mover is no guarantee of future success. Friendster was once the biggest craze in social networking while WebCrawler was the go-to search engine years before Google was founded. Altman has commercial obligations and cannot discard these. Unless governments assert their power with more stringent regulations around new developments, the arms race will continue. 

$86BN VALUATION?

An upcoming sale share will reveal how investors think recent turbulence has affected OpenAI’s valuation. The sale was previously expected to value OpenAI at $86bn. However, the leadership battle has made the company look chaotic and highlighted its unusual governance arrangements. Some analysts expect the valuation to take a hit in light of this. They point to more stable offerings provided by Google and Amazon. Others see benefits. The outgoing board members may have secured certain concessions. But ultimately this is a big win for Sam Altman and a more commercial vision for the company. And should there prove more substance to ‘Project Q*’ rumours, that $86bn valuation may soon look hopelessly outdated.       

 

0Shares

Leave a Reply

Your email address will not be published. Required fields are marked *