The two Google Tensor G1 and Tensor G2 chips used in Pixel devices have so far disappointed users who expected high performance to beat the A16 Bionic and Snapdragon 8+ Gen 1. But in the latest Made by Google podcast, Monika Gupta who is the senior director of Google Silicon development gave the reason why this happened.
According to Gupta, at Google they are focused on bringing better artificial intelligence (AI) features to smart devices. Looking at the need for AI on devices in the next five years, a chip that meets Google's needs will have to be built in-house. Good hardware and software integration is done to ensure this vision is achieved.
He added that existing benchmarking software was developed before the era of widespread use of AI on smartphones. Tensor's current performance is sufficient for Google's software needs.
At the same time Google has a vision of the Tensor family is for ambient computing where it makes everyday life easier in assisting photography tasks, making phone calls and how the phone is used.
So in conclusion Google focuses on AI like Duplex that makes phone calls on behalf of device owners and produces better photos through computational photography.