Google’s biggest opponent in the AI race is not Open-AI – t3n – digital pioneers
Google is still ahead in the AI race. (Photo: Shutterstock / Rawpixel.com)
A Google employee believes that open-source developers can create and adapt AI models much faster and more efficiently. He wants to encourage the company to learn from them.
A leaked document from a Google researcher shows he is concerned that Google could lose the AI race. However, he does not see OpenAI, the company behind ChatGPT, as the biggest competitor, but open source developers.
The document is on the website Semi-Analysis popped up. According to this, it comes from a public discord server where it was shared by the Google employee. The site has confirmed the authenticity of the document.
In the document, the researcher clearly articulates his concerns: “While our models still have a slight advantage in terms of quality, the gap is closing surprisingly quickly. Open-source models are faster, more customizable, more private, and many times more powerful. They do things with $100 and 13 billion parameters that we do with $10 million and $540 billion [Parametern] have to fight. And they do this in weeks, not months”
Editor’s Recommendations
According to the researcher, open source developers are far more efficient than big tech companies like Google and competitor OpenAI. Below, the Google employee mentions some problems that open source developers have already solved, but which Google is still struggling with.
Open source developers have already managed to run large language models on a smartphone like the Pixel 6. In addition, the open source AI is much more flexible. They can be personalized and customized in an evening with a simple laptop.
In addition, open source developers do not care so much about responsible release of the AI. As a rule, they have no reputation to lose.
The Google researcher is of the opinion that Google is currently still ahead in AI development. However, he warns that open source developers are catching up rapidly. “People will not pay for a restricted model when free, unrestricted alternatives are qualitatively comparable,” it said.
As a solution to the problem, he suggests working with open source developers, as that would help the development of Google’s AI models. Google could learn from these developers and use the methods to make its own models scalable and more customizable.
