Memo from Google Acknowledges Defeat to Open Source AI and Proposes New Strategy

 

Google has admitted it cannot compete with open source AI and proposed an unexpected strategy to restore dominance, according to a leaked letter. The document provides a bulleted list of all the ways in which Google is falling behind open source AI and advises how the company might get back in the driver's seat.

With Open Source AI, We Can Now Compete with Google

The message begins by saying that they have always known that Open Source would be their main rival and not OpenAI. It also recognises that Google has already lost the battle for AI domination and is in no position to compete with open source.

The document primarily focuses on open source's superiority over Google. The author of the memo concedes that Google has a modest advantage over open source, but that it is rapidly diminishing and can never be recaptured. The introspection into the metaphorical hand they've been dealt is gloomy.

The Larger the Language Model, the Worse It Performs

The memo expresses the unsettling realisation that Google's massive size is no longer an asset. They have realised that the enormous scale of their models is a drawback, not an overpowering advantage.

The Open Source Community's Projects Can Grow at Any Size

The ability of the Open Source movement to scale its projects at a rate that closed source cannot is particularly worrisome to Google. Dolly 2.0, an open-source ChatGPT clone, relies on tens of thousands of user contributions to its question and answer collection. Both Google and OpenAI used data taken from Q&A sites like Reddit to help answer questions.

Open-Source Model Development Procedures Are Top-Notch

One more thing that has Google's engineers worried is how well open source projects lend themselves to the rapid iteration and low cost of development that characterise open source software.

Join Open Source If You Can't Beat Them

The author of the memo's only positive takeaway from open source is that Google can benefit from its discoveries without spending a dime. Finally, the author comes to the conclusion that Google's only option is to monopolise the platform in the same way that it does with the free and open-source software Chrome and Android.

According to the document, open source is the best choice. If Google wants to be taken seriously as a leader in the open source community, it needs to stop isolating itself and start working with others to shape the discourse. Some uneasy measures, such as making public the model weights for minor ULM modifications, will be required. Consequently, we will have to give up some measure of command over our models. However, a middle ground must be reached. We can't be at the forefront of innovation and also keep everything under wraps.

The AI community is in upheaval after the Google document was revealed. It recognises the superiority of open source AI and recommends that Google join the open source community as a way forward. Whether or not Google takes this advice to heart is unclear, but it's becoming increasingly apparent that open source projects are where AI's future lies.

Results to Come:

The ramifications of the leaked Google letter, in which the company concedes defeat to open source AI and admits it cannot compete against it, for the future of AI development are profound.

It's possible that closed-source systems like Google and OpenAI will fall behind as open-source AI takes over the market. This could result in a shift away from constructing massive language models with billions of parameters in favour of developing more targeted, application-specific models.

One additional result could be a greater willingness on the part of Google and other closed-source firms to embrace open-source AI and work together with the broader community. As a result, the industry as a whole might gain from the creation of more novel and varied models.

The document also hints that Google could try to reclaim prominence by acquiring full control of the platform, much as how they did with the open-source Chrome and Android systems. This may increase competition and new developments, but it also risks creating an AI ecosystem that is more divided and compartmentalised.

Regardless of the conclusion, the leaked email shows that the AI sector needs to keep pushing the boundaries of innovation and working together. As AI grows more prevalent, it is essential that we create systems that are open, fair, and ethical, and that put people's needs ahead of corporate profits.

Source: https://www.semianalysis.com/p/google-we-have-no-moat-and-neither