Google, for years, has been known for its research in Artificial Intelligence (AI) and regularly published its findings in academic papers. However, in February, Jeff Dean, the head of Google’s AI department, announced a significant policy change. Researchers could no longer share their work with the outside world until it had been turned into a product. The launch of OpenAI's ChatGPT, inspired by Google's work, caused this shift.
The change in policy forms part of a greater shift inside Google. Its AI department, previously considered a leader in the field, has been forced into a defensive mode, defending against a wave of nimble AI competitors. Moreover, it needs to protect its core search business, stock price, and future, all of which executives have said are intertwined with AI.
The CEO of Google, Sundar Pichai, has urged caution on AI, warning that it can cause substantial societal harm and can supercharge the creation of fake images and videos. However, Google has also overhauled its AI operations, with the goal of launching products quickly.
Google has lowered the bar for launching experimental AI tools to smaller groups. It has developed a new set of evaluation metrics and priorities in areas like fairness. Google has also merged its AI organizations to accelerate progress in AI. This new division will be run by Demis Hassabis, the CEO of DeepMind, a group known for its fresh brand and hard-charging approach.
This acceleration of innovation comes as experts and industry veterans are calling for AI developers to slow down, warning that the technology is developing faster than its creators anticipated. Additionally, an AI arms race continues without oversight, and companies are becoming less cautious in the face of competition.
Google insists that its efforts to speed up do not mean cutting corners. Pichai has stressed that the company is creating a unit that will help it build more responsible and capable systems. Google has also established an internal governance structure and a comprehensive review process, with hundreds of reviews across product areas.
Google employees have mixed views regarding the accelerated approach. Although the need for additional approval before publishing could mean researchers get "scooped" on their discoveries, others acknowledge that Google has lost many of its top AI researchers in the last year to start-ups seen as cutting edge.
Overall, the policy change to stop sharing AI knowledge until it has been turned into a product reflects the change within Google, from "peacetime" to "wartime." The company is now in a defensive position, and executives want to protect their core business and AI's future. This shift is part of a more significant trend across the AI industry to balance rapid innovation with responsible development.
