Google co-founder Sergey Brin has admitted that the tech giant was overly cautious in deploying language models, despite being pioneers in the field. While speaking at the All-In Summit, Brin discussed Google’s hesitance, acknowledging they were “too timid to deploy” language models, even though they essentially invented them with the publication of the “Transformer” paper seven years ago.
Brin attributed this hesitancy to a fear of making mistakes and causing embarrassment. He stressed the importance of risk-taking in the tech industry, highlighting that while these models can make “really stupid mistakes,” they are also “incredibly powerful” and can significantly simplify tasks that would otherwise require extensive learning.
Despite the potential for errors, Brin believes these models should be released to the public for experimentation rather than being kept under wraps until they are considered perfect. He advocates for a more open approach to AI development, allowing for public exploration and refinement.
Google’s cautious approach to AI deployment has been a topic of discussion within the tech industry. In May, Gene Munster, managing partner of Deepwater Asset Management, suggested that Google was six months behind OpenAI, the company behind ChatGPT, in AI development but still about five years ahead of the rest of the industry.
However, Google has faced criticism from within. A 16-year Google veteran claimed that the company’s AI projects were driven by “panic” rather than user needs. An internal email from Microsoft revealed their concerns about falling behind Alphabet’s AI capabilities, indicating that Microsoft’s investment in OpenAI was fueled by a fear of losing ground to Google in the AI race.
Brin’s statement reflects a shift in Google’s stance towards AI, potentially signaling a more aggressive approach to deploying its language models in the future. This could be a significant development in the ongoing race for AI dominance within the tech industry.
