Please enjoy ARK Disrupt Issue 61. This blog series is based on ARK Brainstorming, a weekly discussion between our CEO, Director of Research, thematic analysts, ARK’s theme developers, thought leaders, and investors. It is designed to present you with the most recent innovation takeaways and to keep you engaged in an ongoing discussion on investing in disruptive innovation. To read the previous issue, click here.
1. Cruise Automation Is Cruising Along In Autonomous Cars
This week Cruise Automation released another video demonstrating its autonomous system’s performance on city streets. (GMGM acquired this autonomous car startup last year for $1 billion.) Introducing the video, Cruise Automation’s CEO explained that the engineer in the car selected a random destination with no preplanned route. Also important to note, the engineer operator’s hands never touched the steering wheel during the drive through complicated city streets, according to camera footage shot from a second angle. This video is among the most impressive that we have seen in the autonomous vehicle space.
In reports required by the California Department of Motor Vehicles (DMV), companies testing autonomous vehicles must detail the average number of autonomous miles between disengagements, or time periods when the operator takes control over the steering wheel from the autonomous system. As of November 2016, the number of miles between Cruise Automation’s disengagements was about 380 miles, well below the more than 6,000 miles that Google’sGOOG Waymo reported, as shown in the tables below. For perspective, the average American drives 13,500 miles per year, suggesting that a riders committed full time to Google’s autonomous Waymo car would pull over and ask for help from an operator only twice a year on average.1
While the DMV reports show that Google is miles ahead of Cruise with its technology, GM has a fairly advanced autonomous vehicle compared to other traditional automakers. The exception, perhaps, is Tesla, though Tesla is not a traditional automaker. Purportedly, GM has planned on launching self-driving car tests with Lyft customers some time this year. ARK will be watching closely.
2. Automation and Robotics Has A Software Side
Automation created lots of news this past week. Striking close to home was a report that Goldman Sachs has automated its U.S. cash equities trading desk since the peak of the tech and telecom bubble: from a peak of 600 in 2000, it is down to two traders today. Automated trading programs have replaced the others, supported by 200 computer engineers.
Rethink Robotics, a collaborative robotics company, also made news with a significant software update. The update, Integra 5, gives users the ability to create a behavior tree for robots, minimizing set-up costs. In traditional industrial robot ecosystems, set-up costs can account for a third of the total cost of ownership. Collaborative robots require less upfront capital and with the right software, like Integra 5, can be trained to do a number of tasks in varied conditions. Rodney Brooks, Rethink’s founder, notes that software will be just as important as hardware in the proliferation of automation.
As illustrated in Mobility-As-A Service: Why Self-Driving Cars Could Change Everything, while hardware costs are falling, improvements in software and deep learning are increasing the number of tasks that robots can accomplish. In other words, the power per unit cost of automation is increasing, and with it the addressable market for robots and automation.
3. Machine Learning Meets CRISPR
A word from another time, “convergence” comes back to life and into full focus when the biggest breakthroughs in computer science meet those in molecular biology. A team of researchers from Microsoft, the Broad Institute, and Harvard has developed a new computational model to optimize the CRISPR/Cas9 gene-editing system using machine learning. This partnership is advancing the revolutionary CRISPR technology by predicting the best way to knock-down genes of interest in humans and mice. The promise of CRISPR lies in its ability to correct disease-causing mutations in the human genome, providing a curative class of therapeutics.
While highly promising, CRISPR is in early days and under investigation for further characterization, particularly the selection of a guide RNA (sgRNA) that directs the “molecular scissors” to the correct part of the genome for editing purposes. A suboptimal sgRNA could result either in unintended edits, otherwise known as off-target effects, or in no effect at all.
While screening hundreds, if not thousands, of sgRNAs against the ~20,000 human genes would be a daunting task, machine learning has the potential to collapse both the time and the cost of such research. The research teams at Microsoft and Broad, for example, were able to identify all possible guides with just 20 well-screened genes, and then use predictive machine learning models to isolate the most efficient and effective guides in different situations. In fact, the research teams were able to leverage this model and generalize the guide efficiency for genes that were neither in the initial data set nor ever interrogated using CRISPR/Cas9. Increasingly, we are witnessing the convergence of biology and technology…biology in silica!
4. Deep Learning Is Booming in China
Deep learning, the algorithm taking artificial intelligence to a new level, was primarily a North American focus, until 2014. China has awakened to the potential of deep learning and has stepped up its research and development around deep learning dramatically. As shown in the chart below, as measured by the number of publications mentioning “deep learning” or “deep neural networks,” it has surpassed the U.S. by more than 20%.
Source: Office of Science and Technology / The White House
China also leads in the number of papers published and subsequently cited, highlighting the high quality of its research. Indeed, the winners of ImageNet 2015 and 2016, the prestigious image recognition contest, both hail from China. In 2015, Beijing-based Microsoft Research Asia was the winner while, in 2016, a Chinese government research agency took first place.
Also migrating to China is talent. In 2014, BaiduBIDU hired as Chief Scientist, Andrew Ng, Google’s deep learning expert, and recently, as Chief Operating Officer, Qi Lu, Microsoft’sMSFT artificial intelligence expert.
While cutting edge research still originates elsewhere, China is well positioned to make big strides in deep learning longer run. Among the reasons are the following:
- China has more internet users than anywhere else in the world, 700 million individuals collectively training algorithms for Baidu, AlibabaBABA, TencentTCEHY, government agencies, and others.
- Deep learning breakthroughs in industry and academia are published and disseminated quickly in the global open source ecosystem. China does not have to reinvent the wheel.
- The Chinese government is placing top priority on its country’s leadership in technology. In 2016, China built the world’s fastest supercomputer using chips designed and manufactured domestically.
China has yet to surpass the US in deep learning breakthroughs, though it is well placed for progress. That said, China is one country. We wonder which multinational company will be the first to harness data from 3-6 billion internet users, training algorithms and taking deep learning to yet another level.
5. The People’s Bank of China Freezes Bitcoin and Litecoin
The People’s Bank of China (PBOC) made more news in Bitcoin this week, as it announced further restrictions on some Chinese bitcoin exchanges, demanding more robust anti-money-laundering (AML) procedures and policies. In response, two of the three biggest exchanges in China have frozen bitcoin and litecoin, another cryptocurrency, from leaving their platforms. That said, the customers of these exchanges can withdraw their yuan if they want to sell their bitcoin or litecoin.
While the bitcoin price initially dropped 10% after the news broke, its resilience has been impressive since the announcement, as it has returned to more than $1,000 a coin. Chinese investors do not appear overly concerned by recent events, as bitcoin is trading on China’s exchanges at only a 2% discount to US exchanges as of Saturday night. The PBOC is increasing regulation on bitcoin exchanges methodically and responsibly, first curtailing margin to avoid severe market swings before enforcing more stringent AML policies and procedures. In our view, it is validating and enhancing the cryptocurrency ecosystem.
Meanwhile, Japan has usurped China’s position as the leader in bitcoin liquidity, as many of its exchanges still offer 0% trading fees and margin services. Creative destruction and disruptive innovation are the lifeblood of cryptocurrencies!
- Note that ARK foresees autonomous taxis pulling over to the side of the road and receiving help from teleoperators in times of need, which ARK expects will be infrequently. ↩
ARK's statements are not an endorsement of any company or a recommendation to buy, sell or hold any security. For a list of all purchases and sales made by ARK for client accounts during the past year that could be considered by the SEC as recommendations, click here. It should not be assumed that recommendations made in the future will be profitable or will equal the performance of the securities in this list. For full disclosures, click here.