The combination of cloud, crowdsourced information, and machine learning “will be the basis for every fundamental and hugely successful IPO win in the next five years,” he said during this morning’s keynote at GCP NEXT, the developer conference for Google Cloud Platform.
Schmidt is prone to sweeping statements. But having watched computing transform many times in 45 years — he was a Sun Microsystems bigwig when the company launched Java — he said he felt qualified to predict that machine learning could lead to truly new innovations, the kind that can’t yet be envisioned.
If Schmidt is right about machine learning, then adding the technology to Google Cloud could be a nice competitive wedge against Amazon Web Services (AWS), which leads the public cloud market by a large margin.
But Google is working on catching up. Walt Disney Co. was announced as a customer today, as expected, and the GCP NEXT keynote also included executives from Coca-Cola Co. and Spotify.
Machine learning has been used inside Google since 2012, but it’s mostly in the past year or so that usage has taken off, corresponding with that machine learning technology getting better, said Jeff Dean (pictured above), the Google Fellow who’s spearheading the Google Brain project.
To help programmers tap the technology, Google is providing three APIs: Google Translate, Google Vision, and Google Cloud Speech. The first two had already been released. Cloud Speech, providing speech-to-text conversion in more than 80 languages, is being announced today.
And Google has already seeded the open source community. Cloud Machine Learning is based on a library called Tensorflow, which Google open-sourced in September. Tensorflow lets developers build machine learning models and then scale them to production.
You can see the results in applications such as Neural Art, which converts a photo into the style of a famous painter of your choice. Here’s Dean again, with a sample: