Season Finale of the OpenAI Saga -- recap, lessons, and what it means for LegalTech?

It has been a wild last few days. Frankly, something better suited for an episode of Silicon Valley, and what will most likely make for a thrilling segment in an upcoming movie, than a company valued at $90bn!

In case you missed all of the fuss, I've provided a summary of the story (so far...) and some thoughts about winners/losers and what this might mean for [LegalTech] startups.

Quick catch-up - a whirlwind four days!

On Friday, Sam Altman, the charismatic leader, Co-founder, and CEO of OpenAI (OAI), is ousted from his position. He was fired by the board, stating that he was “not consistently candid in his communications" with the board of directors. “The board no longer has confidence in his ability to continue leading OpenAI". Mira Murati is named as the interim CEO.

Greg Brockman, Co-founder and President of OpenAI is asked to step down from the board. Later that evening, Greg resigns from his position.

Going into Sunday, confusion around the decision remains. All the while, there was an outpouring of support for Sam, with several notable resignations at OAI.

Reports emerge that investors are urging the OpenAI board to reverse its decision and reinstate Sam as CEO.

Towards the end of the day, it looks like the board will stick with the decision. They announced Emmett Shear (Co-founder and previous CEO of Twitch) as interim CEO.

On Monday, Microsoft's Satya Nadella shared that Sam and Greg will be joining Microsoft as part of an advance AI research team.

Over 95% of OpenAI's 770 employees sign a letter threatening to quit and join Microsoft if the board refuses to resign and reinstate Sam as CEO. This includes Mira Murati and Ilya Sutskever (see below).

By the end of Tuesday, OpenAI announced Altman would be returning as CEO and introduced a new board. Greg Brockman is back as well.

Interesting finds

• ❗OpenAI Approached Anthropic for Merger: Amidst the turmoil, OpenAI's board made a significant move by approaching Dario Amodei, Co-founder and CEO of Anthropic, proposing a merger. The proposal included potentially elevating Amodei to the CEO position of the merged entity. Amodei turned it down.
• New Board Members at OpenAI: Following the tumultuous events, OpenAI has partly reconstituted its board of directors. The new board includes Bret Taylor, former co-CEO of Salesforce, as the chair, and Larry Summers, former U.S. Treasury Secretary, along with Adam D'Angelo, CEO of Quora. The initial board will also vet and appoint a new formal board of up to 9 people. Microsoft and Sam Altman will likely have seats. As Azeem Azhar notes in the Exponential View newsletter (paid), there is room to grow:

It needs to grow to have more of the characteristics of a strong board: cognitive, professional, personal and cultural diversity would be an excellent place to begin. It might also benefit from board members with expertise in technology risk and the sociotechnical dimensions of technology. OpenAI will likely continue on its roadmap to build and ship more robust and helpful enterprise software.

• Microsoft wins overall, but Teams maybe not so much: Microsoft, after investing over $13 billion, holds a 49% stake in OpenAI​​​​​​​​​​. Despite this significant investment, Microsoft was reportedly informed just minutes before the public announcement of Sam Altman's departure. Adding salt to the wound, many on social media highlighted OpenAI's use of Google Meet for its meetings, ironic choice given Microsoft's heavy investment.
• Emmett Shear's 30-Day Plan: Shear acted with speed and confidence, laying out a 30-day plan (the highlights of which he posted on X), which included governance changes as needed:

• Internal Tensions and Power Play: The reported clash between Sam Altman and Ilya Sutskever, Chief Scientist of OpenAI, revealed deeper tensions over safety, the commercialization of AI, and a broader cultural schism within the organization and AI community.

Winners/Losers

• Satya Nadella/Microsoft: Satya Nadella demonstrated a mastery of corporate strategy by acting swiftly amidst the chaos. By offering roles to Sam Altman and others from OpenAI, Microsoft created a win-win scenario for itself. Whatever happens, they will have the best talent, partnerships, and goodwill. Additionally, Nadella spoke candidly on news shows and social media. My favorite was providing reassurance to partners and customers while owning that "Suprises are bad." All this was executed before the markets opened on Monday, which resulted in MSFT stock hitting a record high. NY Times has good coverage.
• Sam Altman: Despite the initial setback, Altman emerged as a winner by returning to his CEO position at OpenAI. His swift reinstatement reflects his pivotal role and influence within the organization.
• Ilya Sutskever: Initially, it appeared that Sutskever might have played a role in the internal dynamics at OpenAI. However, his public expressions of regret and efforts towards reconciliation played a key role in mending the rifts, potentially avoiding awkwardness in future leadership meetings.
• Anthropic and Open-Source LLMs: the importance of an LLM agnostic strategy, leveraging alternatives (e.g., Anthropic, Cohere, and open-source models) has never been more evident. Somewhat timely, Anthropic released Claude 2.1, which can now process up to 200,000 tokens of context (approx. 150,000 words/500 pages of text), has 2x reduction in false/hallucinated statements, and includes early support for connecting to existing company APIs, databases and web services.

What it means for LegalTech

Thousands of startups and applications are built on OpenAI infrastructure (including some tools we've developed at Fringe Legal), and those purely reliant on OpenAI likely had a bit of a panic over the weekend. For startups/companies heavily reliant on OpenAI's API or those who have built their services around it, the recent events should signal a review.

Uncertainty and Dependability: The turmoil at OpenAI highlights the risks associated with dependence on external APIs for core functionalities. LegalTech companies must consider diversifying their technological partnerships or developing in-house capabilities to mitigate these risks.

Strategic Autonomy: one option is to prioritize creating strategic autonomy by developing or adopting a range of AI technologies. The OpenAI situation exemplifies the volatility that can come from relying on a single provider. Diversification can come from integrating multiple AI models, including open-source options, to avoid being at the mercy of external company shifts.

Shaw Curran (Director of Legal Technology at Travers Smith) puts it eloquently in a recent LinkedIn post highlighting a deviation from the "inverted pyramid approach":

At Travers, we quickly understood that this AI revolution would result in most firms allocating their limited capital for AI to the presentation layer - a variety of applications aimed at distinct use cases, yet fundamentally performing the same abstract functionality.

Eventually, cost conscious law firm firm leadership would realise that they had simply re-purchased the GPT family from various resellers. We termed this phenomenon as the 'Inverted Pyramid', where the collective expenditure for firms is concentrated on the presentation layer - the top of the pyramid, leaving very little capital to allocate to the computational layer - the bottom of the pyramid.

Our strategy deviated from the norm. We didn't invert the pyramid. The majority of our capital was invested in the diversification of our base model architecture, with the remainder dedicated to two abstract cross-practice user interfaces that we built ourselves.

Risk Mitigation: The upheaval underscores the need for LegalTech companies to have contingency plans in place, especially for larger providers. This includes contracts with multiple AI providers or setting up systems that can quickly switch between different AI services without significant downtime or loss of functionality. This post from Kelvin Legal captures the sentiment well, highlighting that you should "design for change." I would be curious to learn how services like Harvey, Casetext and others fared during API downtime earlier this week.

Has your AI strategy (or execution against the strategy) shifted as a result of the recent events?