Google's CEO, Sundar Pichai, announced at the Google's I/O developer conference the launch of a new division that will be dedicated to AI in both services and research. This new division already has a new site www.google.ai and will be focused on AI, deep learning as well as building the services and tools to make it all possible.
Pichai also announced new Cloud TPUs (Tensor Processing Units) which uses to train machine-learning models is available in the Google Cloud Platform for anyone to use via the Google Compute Engine today. In his own words"We want it to be possible for hundreds of thousands of developers to use machine learning".
Each device delivers up to 180 teraflops of floating-point performance, and these new TPUs are designed to be connected into even larger systems. A 64-TPU pod can apply up to 11.5 petaflops of computation to a single ML training task.
Pichai actually referenced this architecture is like the movie "Inception" but without Leo DiCaprio. The system takes a set of candidate neural nets (Pichai called them baby neural nets) and iterate them using a reinforcement training approach until the best one is found. The CEO said the results are promising.
All of this work will be used in Google's own products as well as a special focus to help medical researchers. Pichai talked about using AI to help sequence DNA and helping pathologists locate things like the spread of cancer. System is not flawless yet "There are important caveats, we do have higher false positives," Pichai said. "But already getting this into the hands of pathologists they can improve diagnosis."
Google announced that it's already partnering with health care providers to put the technology into action to help improve care and prevent medical incidents.
Hopefully we won't get to this on 2021...