The dust has settled on Google I/O 2023, revealing a plethora of exciting announcements ranging from physical products like the Pixel 7a and Pixel Fold to AI updates such as PaLM2. However, one notable absence stands out—an overarching vision that connects these products together: a next-generation AI assistant capable of spanning Google’s vast ecosystem.
It would be unfair to accuse Mountain View of neglecting AI advancements, especially with PaLM2 powering Google Bard and its growing integration with Search, Gmail, and other services. Google continues to make rapid strides in narrowing the gap with ChatGPT, leveraging its well-established product ecosystem as a platform for integrating advanced AI features. However, what remains unclear is Google’s vision for AI concerning its physical product portfolio, if such a vision exists at all.
While Search remains a significant revenue driver, chatbots like Bing Chat have already outperformed Google Assistant in handling mundane queries often directed at smart speakers. Integrating these capabilities into Google’s expansive Home ecosystem seems like the logical next step, greatly enhancing the usefulness of smart speakers and displays. Surprisingly, no such announcement or forward-looking roadmap was revealed alongside the new Pixel Tablet. The Tablet could have been a far more enticing prospect if it incorporated Bard or similar AI capabilities at the heart of our homes, but instead, it appears to be an expensive, dockable, yet generic Android tablet.
Undoubtedly, Google is still fine-tuning Bard, and a swift rollout to various related products would have been atypical for Mountain View’s pace of development. After all, Bard’s waitlist only opened in March, and the primary focus is currently on the immense power of these online-only language models due to the immediate use cases they offer. However, this situation may need to change sooner than expected, and Google should adopt a forward-thinking approach.
Performing an individual query currently costs mere fractions of a cent, making it potentially uneconomical to scale up to the equivalent of the 8.5 billion daily Google Searches. While Google intends to integrate generative AI into Search, the impact on the profitability of its crucial ads business remains uncertain. This is where the significance of slimmed-down, on-device models has yet to be fully acknowledged.
We are still a long way from witnessing the impressiveness of Bard or ChatGPT running on our phones without an internet connection. However, deploying lower-accuracy models directly on devices is undoubtedly a crucial aspect of the AI future, both in terms of cost-effectiveness and security. Qualcomm’s compression of Stable Diffusion to run on its Snapdragon 8 Gen 2 processor has already demonstrated the possibilities.
In line with this trend, Google has developed its own custom silicon, the Tensor G2 processor, designed specifically for on-device machine learning tasks, including advanced image processing. This chip powers various recent hardware releases and AI tools like Magic Eraser. It’s evident that custom silicon with ML capabilities will continue to play a central role in future product launches. Hence, it is puzzling that Google, at least publicly, has no immediate plans to enhance Assistant and leverage this investment to bring broader generative AI to the most convenient place—our pockets.
We can still look forward to the arrival of the Tensor G3 processor and Pixel 8 series later this year, which may unveil more advancements in pocketable AI capabilities. Typically, new hardware leads the way for software innovation. Nevertheless, the absence of any AI-related announcements at Google I/O regarding its smart home, smartphone, and other product ecosystems suggests that we may need to wait for at least another twelve months before the company attempts to push the AI envelope.
In the fast-paced world of AI, a year is an incredibly long time. Google was clearly caught off guard by the explosive arrival of ChatGPT. Let’s hope that Google is not overlooking the broader AI landscape and its potential impact as well.