Google dominated much of the news this week. A few days ago, the company just finished its big developer lovefest, the annual I/O conference. It is no big surprise that Sundar Pichai, CEO of Google, declared the company will use an AI-first approach in all its products. Google Assistant is now available on iOS. The company is building its own AI chip, the Tensor Processing Unit (TPU), to run neural network. Since they do not want to rely 100% on outside vendors, it makes sense to integrate vertically at this early stage of the game. Wire has a nice summary (see below) on the conference. The most interesting announcement to me is obviously Google Jobs. Industry participants have seen this coming for some time. In other verticals such as consumer review, Yelp is fighting back by integrating downstream and complaining loudly everywhere. Industry boundary clearly doesn’t mean much these days and we will see how this will play out in the next few quarters
Google has also unveiled a new approach to machine learning where neural networks are used to build better neural networks – essentially teaching AI to teach itself. If computer becomes fast enough to teach itself, what will happen next? Based on the results Google has seen, this new approach might be even be smarter at recognizing the best approaches to solving a problem than the human experts. This has the potential of automating the workflow of building AI algorithm in the coming years. This obviously requires intensive computing power but the company also wants its AI technology to cover devices with less horsepower. TensorFlow Lite, a streamlined version of the open source software engine that drives neural networks inside Google’s data centers, will provide AI to your smartphones. It’s designed to be fast and small while still enabling state-of-the-art techniques. Google will make TensorFlow Lite part of the primary TensorFlow open source project later this year. We all can play around with it soon. Running AI on your phone. Google is certainly not leaving any computers behind.
WSJ had a great front-page article on Google’s tangled relationship with its former star employee, Anthony Levandowski, who was currently sued by the advertising giant for stealing trade secret to its self-driving car competitor, Uber. The newspaper also featured a special series detailing how machines and their masters are the undisputed new kings of Wall Street. The company ended the week by beating Chinese go champion, Ke Jie, who said, “For the first time, AlphaGo was quite human-like. In the past it had some weaknesses. But now I feel its understanding of go and the judgment of the game is beyond our ability.”
Lyrebird, a Montreal-based AI startup, has just revealed a voice imitation algorithm. The company says the technology can not only mimic the speech of a real person but shift its emotional cadence. The most impressive part of the technology is that it requires very small amount of training data. While it may not replace the SNL cast soon, it may get much better as the algorithm improves itself. Currently, Lyrebird still retains a slightly robotic tone. You can hear the clip of Trump, Obama and Clinton generated by Lyrebird. It is not bad but I still like Alec Baldwin much better.
Can any job be immune to AI? A computer science professor at Georgia Tech is using artificial intelligence to
replace augment his lowly-paid teaching assistants. Next time you have a questions on your algorithm homework, an AI bot may be giving you advice.
Speaking of job destructions, share prices of staffing giants Adecco (AHEXY) and Ranstad (RANJY) fell Wednesday after being downgraded by Credit Suisse research analyst. The Swiss bank warned investors of potential threat posed by automation to both companies. Credit Suisse estimated that as much as two thirds of temporary occupations will eventually be automated. If the analyst is right, this is bad news as much of new job creation in the last decade is coming from temporary jobs.
New York Times has a great article on the concept of Fat Start-Up yesterday. In the past, venture capitalists typically made relatively small investments in startups in hope of huge payout from the winners. The distribution of startup success is highly skewed but VCs only need a couple home-runs to keep their investors happy. The NYT article profiled Opendoor, a startup with the goal of transforming the real estate market, by offering sellers certainties in selling their houses. The idea is simple and real estate is one of few sectors largely untouched by technology. If Opendoor has enough data to perform market making, it can earn enough profit by the spread and provide important liquidity to homeowners. However, it requires lots of capital and sophisticated risk management. To me, this sounds like investment banking on steroid. The company obviously has many believers as it has already raised $300 million. Another startup, Grammarly, just raised over $100 million in their first institutional round. Staggering amount of capital in private market. Amazon only raised $56 million in their IPO twenty years ago.
Walter Mossberg is retiring as he just posted his last weekly commentary. Mossberg is arguably the best known technology journalist and has been with WSJ forever until he and his partner parted way with the newspaper giant.
This is my last weekly column for The Verge and Recode — the last weekly column I plan to write anywhere. I’ve been doing these almost every week since 1991, starting at the Wall Street Journal, and during that time, I’ve been fortunate enough to get to know the makers of the tech revolution, and to ruminate — and sometimes to fulminate — about their creations.
His last commentary is vintage Mossberg, full of insights and analysis. I will certainly miss his column.