Will Knight comments on how the recent Tesla autopilot crash highlights that the increasing complexity of modern AI systems potentially outstrips our ability to comprehend them:
Tesla hasn’t disclosed precisely how Autopilot works. But machine learning techniques are increasingly used to train automotive systems, especially to recognize visual information. MobileEye, an Israeli company that supplies technology to Tesla and other automakers, offers software that uses deep learning to recognize vehicles, lane markings, road signs, and other objects in video footage.
Machine learning can provide an easier way to program computers to do things that are incredibly difficult to code by hand. For example, a deep learning neural network can be trained to recognize dogs in photographs or video footage with remarkable accuracy provided it sees enough examples. The flip side is that it can be more complicated to understand how these systems work.
Fortunately, the industry is already starting to respond:
As these algorithms become more common, regulators will need to consider how they should be evaluated. Carmakers are aware that increasingly complex and automated cars may be difficult for regulators to probe. Toyota is funding a research project at MIT that will explore ways for automated vehicles to explain their actions after the fact. The Japanese automaker is funding a number of such research projects related to challenges with self-driving cars.
From BIO (a biotech trade organization), new data shows that over all trials, the 2006-2015 clinical development success rate was only 9.6% for phase I through approval. Some disease areas have had better luck than others, but overall the report is a very sobering reminder that clinical research is really, really, hard.
From John Ioannidis, a recent essay: “Why most clinical research is not useful.” Ioannidis has long been critical of research reproducibility and has advocated for reform. From the summary:
- Many of the features that make clinical research useful can be identified, including those relating to problem base, context placement, information gain, pragmatism, patient centeredness, value for money, feasibility, and transparency.
- Many studies, even in the major general medical journals, do not satisfy these features, and very few studies satisfy most or all of them. Most clinical research therefore fails to be useful not because of its findings but because of its design.
Perhaps current practices actually make clinical research harder than it needs to be…
Great observations from Arnold Kling, promoting his new ebook, Specialization and Trade: A Re-introduction to Economics.
I don’t buy into everything Arnold writes, but I do agree that there are problems with the paradigm of macroeconomists treating the economy like a homogenous GDP factory driven by an engine that you rev up with more government spending. Such an approach ignores specialization and patterns of trade which don’t necessarily respond significantly or rapidly to increases in government spending. “Specialization is subtle, deep and highly dependent on context.” There’s a lot more complexity and non-intuitive feedback loops in a modern economy compared to 1930 and it’s not clear how useful Keynesian stimulus is today.
Construction projects take too long to get started, and automated factories can ramp up production too easily (without hiring) to meet the increased demand from policies that stimulate consumer spending, weakening the link between increased spending and increased wages. Real changes in the economy take years to unfold as new pattens of trade need to be identified and developed. Another short term factor in the globalized world is how increased spending in one country may simply lead to higher purchases of imports (for the U.S. think consumer electronics, toys, clothes, etc.). A big chunk of your own government stimulus spending therefore goes toward stimulating another country’s economy (e.g. China).
Arnold’s new book is a different take on macro economics, and it’s much more optimistic than this bit of Keynesianism criticism. In fact, he argues the complexity and productivity that arise from specialization and patterns of trade are at the heart of why modern economies are so successful.
Interesting take on the somewhat overhyped knowledge economy. It’s a pretty accurate assesment in terms of the types of jobs we have in the economy (even in the 21st century) — but it neglects that fact that knowledge management and data analytics are today embedded in everything.
About a dozen trucks from major manufacturers like Volvo and Daimler just completed a week of largely autonomous driving across Europe, the first such major exercise on the continent.
The trucks set off from their bases in three European countries and completed their journeys in Rotterdam in the Netherlands today (Apr. 6). One set of trucks, made by the Volkswagen subsidiary Scania, traveled more than 2,000 km and crossed four borders to get there.
The trucks were taking part in the European Truck Platooning Challenge, organized by the Dutch government as one of the big events for its 2016 presidency of the European Union. While self-driving cars from Google or Ford get most of the credit for capturing the public imagination, commercial uses for autonomous or nearly autonomous vehicles, like tractors from John Deere, have been quietly putting the concept to work in a business setting.
Read the whole article at Quartz.
SpaceX Dragon 9 booster lands safely on an autonomous floating platform, after launching a payload into space.
Potentially a 10-fold reduction versus current launch costs. Game. Changer.
Video at: https://youtu.be/sh8V0COrrzE?t=2129
And Blue Origin isn’t far behind.
Update: The payload was the inflatable Bigalow space capsule, scheduled to be deployed on the International Space Station for two years of testing. Another potential game changer.
Everyone knows Florence Nightingale as a dedicated and courageous caregiver, the founder of modern nursing. But she was also quite a good statistician. In 1850’s, while she served at the British military hospitals in Turkey, she instituted rigorous data collection and analysis for medical records. She collected new types of data, filling notebooks with tables and graphs. In fact, she popularized a kind of infographic, the “coxcomb” chart (essentially polar area plots). Because of her data analysis, she was able to demonstrate seasonal trends in mortality and illustrate the large fraction of preventable deaths caused by disease — supporting the idea that poor sanitary conditions were the main cause of hospital deaths. In fact, in 1859, Florence Nightingale was elected the first female member of the Royal Statistical Society.
One of the more interesting topics this week at Apple’s product event was the introduction of Liam, a robotic system for disassembly of iPhones:
Click here for a cool YouTube movie.
When an iPhone is finally discarded, Liam detects, disassembles, and separate parts for recycling. Cobalt and lithium are recovered from the battery, gold and copper from the camera, platinum and silver from the main logic board, etc. Check out apple/recycling for more information.
Makes you wonder how far automated assembly of iPhones has progressed…
Update: Mashable has a detailed article about Liam. The full system is about as big as a medium-sized warehouse, has 29 different arms, and can process around 11 million iPhone 6S devices a year. Apple is continuing to fund significant R&D in automated recycling — apparently controlled disassembly makes it much easier to recycle the materials compared to conventional “shred-and-separate” methods.
From an analysis done by USA Today: Just 28 companies in the S&P 500 index generate 50% of all the net income. Of those, Apple and JPMorgan Chase together make 10% of the profits. Berkshire Hathaway, Wells Fargo, Gilead Sciences and Verizon round out another 10%. That’s one fifth of total profits from only six companies.
Think about that for a moment. If you believe profits drive stock market value, that means the overall returns of large scale investments such as 401k’s and pension plans have to be driven by these large cap stocks — because that’s where the profits are. Taken as a whole, there aren’t enough profits in small caps for everyone’s investments to do well. Another strong argument for market cap weighting, it seems to me.
Here’s a chart from the article:
As #blizzard2016 passes over #Chicago, the #EastCoast seen in distance clearly has a long way to go. #YearInSpacepic.twitter.com/qMrkTXo9ie
— Scott Kelly (@StationCDRKelly) January 23, 2016