“The cumulative effect of this innovation will continue to change society.”
It’s been a busy year for UCSC Silicon Valley Instructor Oswald Campesato, who expects to publish a book on TensorFlow 2 any day now, marking his 23rd book in an ongoing “Pocket Primer” technology series.
“I don’t think they’ve turned down a book proposal yet,” he says, remarking on an amiable, 18-year-long relationship with his publisher, Mercury Learning & Information.
Earlier this year the Pocket Primer series added Campesato’s C Programming Pocket Primer, and more recently added his TensorFlow Pocket Primer and Python for TensorFlow Pocket Primer.
The Potential of AI
It’s an amazing time to be involved in machine learning, deep learning and reinforcement learning, says Campesato who is teaching courses on all of those topics.
“There’s nonstop innovation in multiple areas of machine learning, with recent breakthroughs for NLP (neuro-linguistic programming),” he says. “Problems that were maybe once considered practically impossible—now, they’re finding solutions. The cumulative effect of this innovation will continue to change society.”
Changes to TensorFlow
In the midst of it is TensorFlow, which with Keras, is reigning as most popular framework for deep learning and machine learning.
TensorFlow 2.0 alpha, released in March, added new features and an improved user experience. It more tightly integrated Keras as its high-level API. TensorFlow 2.0 Beta, which marked “a significant milestone” in the technology, was announced in June.
TensorFlow 2 is essentially a platform that includes mobile, the web, and the ability to deploy large scale models in the cloud, Campesato says. By way of comparison, working with TensorFlow 1.x is more like driving a stick shift transmission while TensorFlow 2.0 is like an automatic.
“It’s a nice way to do a lot of things,” he says. “It’s so much more convenient.”
The Deep Learning/TensorFlow course curriculum uses TensorFlow 1.x. It will probably switch to 2.0 some time next year as the latest technology gains traction in new projects.
“It’s like turning an ocean liner,” Campesato says. “TensorFlow is going through a big change. When they release the production version of 2.0, then 1.x will become legacy code. Although version 1.x will be supported for a year beyond the production release of 2.0, the clock is ticking and the needle is moving toward version 2.0.”
Keeping Pace with Change
Just a few years ago, Campesato was focused on mobile applications and web development, but the dramatic potential of deep learning and machine learning beckoned him. He became a sought-after speaker for deep learning presentations and has been an integral instructor as UCSC Extension builds out its AI curriculum.
“For me, the switch to AI was necessary and inevitable, exhilarating, inspirational—all those great words. That might sound corny to some people, but I really believe it’s true.”
Who can afford to wait? he asks.
TensorFlow 2 Pocket Primer
Keep your eye out for Campesato’s next book in the Pocket Primer series: TensorFlow 2 Pocket Primer which was recently released.