#63 Decoding the Google Memo: The Moat-less Landscape of Generative AI?

<< Previous Edition: From Endings to Beginnings

The industry is buzzing with the recent leak of a Google memo titled "We Have No Moat, And Neither Does OpenAI." While the lack of a protective "moat" around generative AI could be a concern, it doesn't necessarily present a problem for Google or OpenAI. ?Instead, the real concern lies in the chaos and unpredictability of open-source initiatives.

The Invisible Hand and Open-Source Movement

The open-source movement's inception was a tactical move, instigated by the formation of the Apache Software Foundation (ASF). IBM silently but influentially aided in offsetting Microsoft's dominance, nurturing the essence of altruism. This was a masterful strategy, where IBM succeeded in creating a new battlefield, one that Microsoft was unaware of.

Interestingly, Microsoft held a scornful outlook towards open-source. This was epitomized by their ex-CEO, Steve Ballmer, who went as far as likening Linux to cancer. This contentious view led to a fervent competition, with IBM and the open-source community on one side, and Microsoft on the other, forever changing the software industry's landscape.

The Need for a Strong Backer

In the realm of generative AI, an open-source alternative needs the support of a heavyweight patron to thrive. Currently, Facebook's Llama model is leading the open-source race, positioning Meta as a potential game-changer. With substantial resources, Meta could potentially render the open-source option feasible. This role of the patron is essential in driving innovation and fostering competition in the AI space.

However, the question remains: will Meta rise to the challenge? This is yet uncertain. Meta's history doesn't indicate a proclivity for such initiatives. Despite having a significant advantage in hyperscale, they chose to stay away from the cloud infrastructure business, indicating a cautious approach towards new ventures.

Understanding the Varied Nature of Moats

In the tech sector, it's easy to believe that intellectual property is the sole form of a moat. However, moats can assume diverse forms. In the B2C segment, the network effect forms the most potent moat. This is likely the type of moat ChatGPT will possess for conversational interfaces, necessitating a tremendous effort from challengers to overcome it.

Brand - another form of moat - is more discernible in the consumer goods realm. Though not as prevalent in tech, there was a time when the phrase "no one ever got fired for buying IBM" rang true. The switching cost moat, on the other hand, is quite conspicuous in the tech industry. It escalates to a seemingly insurmountable barrier, likened to a black hole's event horizon, when dealing with systems of record management products like Oracle and SAP.

Conclusion: The Future of Open-Source and Generative AI

I previously exhibited a bullish stance towards open-source in generative AI, as evident in my prior writings. However, given the fresh information and lackluster performance of existing open-source models in our labs, the fear that open-source might dominate the generative AI movement seems baseless. The current state of the open-source movement in generative AI is in disarray, and without the backing of one of the current era's titans, this is unlikely to change.

>> Next Edition: Harnessing the Power of Fine-Tuning