Google Losing AI Edge To Open-Source Community: Internal Memo Reveals Concerns

A secret internal Google paper has been widely circulated in the technical community. According to a top Google software engineer, Google is losing its edge in artificial intelligence (AI), but the winner will not be OpenAI, the creator of ChatGPT – it ought to be AI developed by the open-source community.

According to the engineer, Google’s battle with OpenAI has diverted the company’s attention from the rapid development of open-source technology. Many independent researchers in the open-source community have made surprising and rapid progress using AI approaches.

“The uncomfortable truth is, we aren’t positioned to win this arms race and neither is OpenAI. While we’ve been squabbling, a third faction has been quietly eating our lunch,” the engineer wrote.

A Google spokeswoman confirmed that the memo was written by one of the company’s top software engineers, but stressed that it was not an official company strategy document. The engineer has not yet responded.

Google Is Not The Creator Of The Next AI Model

Employees at Google and other digital giants are buzzing internally and publicly about the technology that is transforming their businesses as research into generative AI advances. Google is under particular scrutiny as the success of ChatGPT has raised concerns that Google is losing its edge in AI. Google has long been a leader in the field.

But in the paper, the aforementioned senior Google engineer says that the real threat to Google comes from the open-source community, where programmers are rapidly inventing models that match the quality of the big tech companies’ wares, but at a lower cost. He claims their models are faster, more configurable and more helpful than Google’s.

“We don’t have a secret sauce,” the employee said. “Our best hope is to learn and work with what others outside of Google are doing.” He is concerned that buyers may refuse to pay for a model that offers such high-quality technology for free.

A large language model created by Meta was released to the public in February, accelerating the growth of generative AI in open-source forums. The model, called LLaMA, is smaller and, therefore easier to use than those promoted by Google and OpenAI. To gain access to LLaMA, researchers currently have to apply to Meta.

The Google engineer argued that the company would be better off focusing on a smaller, more flexible model. “Huge models have slowed us down,” the engineer wrote. “In the long run, the best models are those that can be iterated quickly.”

A Google official declined to comment on the content of the post. Google CEO Sundar Pichai said on a recent earnings call that “our investments and breakthroughs in AI over the past decade have positioned us well”, emphasizing that the company is building models and working with developers and other partners. Pichai has previously advocated for AI regulation, warning that the technology could be “very harmful” if not used carefully.

Source