Last week I attended Google’s annual 3-day developer conference, Google I/O, held in Mountain View, California. This event has frequently been a showcase for new Google products and ventures, some of which persist and grow (Android, the Chrome web browser, Google Photos) and others that fade (Nexus Q, Google Wave, Google Glass).
This year’s event followed a similar pattern, but with less emphasis on new products and ventures. Instead there was a strong emphasis of existing Google capacities and how they have matured and expanded.
I was very impressed by Google’s TensorFlow software library for machine learning and its potential. TensorFlow is an effort that started years ago with the Google Brain project, a deep learning/artificial intelligence effort, and was made available as an OpenSource software library in 2015. Over the years it has become part of Google’s solution for voice and image recognition problems, and other AI applications.
This year, TensorFlow, along with the high-capacity computing platform Google has engineered for it is own systems, has evolved into a new research capacity that can be widely shared. In a clear and concise article about machine learning from earlier this year, Aditya Singh predicted that “tech companies in the next few years will democratize deep learning”. Google appears to be doing that now, as Sundar Pichai, Google’s CEO, highlighted the new potential this offers to the world’s developer community for addressing critical problems in environmental studies, health care, transportation, and more. Several sessions at Google I/O were devoted to helping developers understand how to take advantage of TensorFlow and Google’s computing platform, and Google announced the TensorFlow Research Cloud that will make this platform freely available for important academic research projects.
By way of example, Pichai’s keynote address featured this compelling video about Abu Qader, a 17 year old high school student from Chicago who used TensorFlow’s machine learning tools and existing medical imaging devices to improve early detection of breast cancer. I loved how Abu Qader described the potential:
I’m by no means a wizard at machine learning. I’m completely self-taught. I’m in high school. I YouTubed and just found my way through it. You don’t know about that kid in Brazil who might have a ground-breaking idea, or that kid in Somalia. You don’t know they have these ideas, but if you can open-source your tools, you can give them a little bit of hope that they can actually conquer what they’re thinking of.
Bruce Washburn is a Principal Software Engineer in OCLC Research. He provides software development support for OCLC Research initiatives and participates as a contributing team member on selected research projects. He operates the ArchiveGrid discovery system and provides software development support for selected OCLC Products and Services.