Monthly Archives: February 2017

Forbes

Technology Hits The Fields Forbes During a recent blizzard in Massachusetts, Sonia Lo, CEO of FreshBox Farms, was in a grocery store suggesting to skeptical patrons that they sample her leafy greens. They were picked yesterday, is what she told tasters. She also told them no, they ...

See more here:

Technology Hits The Fields - Forbes

On the surface,a Google subsidiarys blistering accusation last week that Uber has stolen its driverless car technology looks like any of the thousands of patent lawsuits piling up in Silicon Valley court dockets.

This one is different, however. And its different in ways that could spell bad news for Uber.

The lawsuit was filed Thursday in San Francisco federal court by Waymo, a subsidiary of Alphabet Inc. devoted to developing self-driving technology. (Alphabet is the new name for Google.) Waymo is responsible for those bug-shaped cars and other vehicles testing the technology around Northern California. Theyre equipped with sophisticated laser systems that create a 3-D picture of the landscape, allowing the vehicles to navigate around obstacles.

Waymo says that Anthony Levandowski, who was once Googles driverless-car guru, downloaded 14,000 proprietary company files onto his own computer and absconded with them when he left Google to found his own company in 2016. That company, Otto, was soon acquired by Uber for $680 million. Not long after that, Google says, it discovered that Ottos technology was largely identical to its own. The lawsuit seeks damages for alleged infringement of three Google patents and an injunction barring Uber from using any of the technology.

One aspect of the lawsuit that struck some Silicon Valley observers from the first was the extensive detail in the accusations.

Normally in a case like this, theres a lot of innuendo in the early stages, says Eric Goldman, a patent law expert at Santa Clara University law school. But Waymo specified how and when it alleges Levandowski downloaded the files, the breadth of his alleged theft and efforts to conceal his actions, and how it discovered them from an email a supplier sent to members of Levandowskis team and mistakenly copied to a Waymo employee.

Google spent a lot of time and money investigating before it filed the lawsuit, Goldman told me.

Of course, the lawsuit represents just one side of the story. Uber hasnt yet responded in court. In a statement, the company said it had reviewedWaymo's claims anddetermined them to be a baseless attempt to slow down a competitor. Uber added, We look forward to vigorously defending against them in court.

Another unusual aspect of the case, Goldman says, is that Google is bringing it at all. Intellectual property lawsuits are out of character for Google, he says, even though it has so many former employees that a large amount of its IP must be at large in the technology community. They just dont show up as a plaintiff, he says. Moreover, Google is an investor in Uber with a stake of at least $250 million; a Google executive sat on Ubers board until just after Uber acquired Otto.

Google hasnt said much about why this episode should be different from any others, beyond a Waymo blog post that implies it was just too gross an offense to ignore.

These actions were part of a concerted plan to steal Waymos trade secrets and intellectual property, the post reads. Given the overwhelming facts...we have no choice but to defend our investment and development of this unique technology.

I asked Google to elaborate, but havent heard back.

Goldman conjectures further that this might be an asset especially valuable to the Google family, something extra important.If thats so, it underscores the grand expectations for driverless technology, despite indications that it may be oversold. Google has been among the most enthusiastic developers in the field; just last year, Eric Schmidt, its executive chairman, crowed that "the technology worksbecause, frankly, the computer can see better than you can, even if you're not drunk in a car.

The lawsuit says the markets for self-driving cars are nascent and on the cusp of rapid development. It asserts that the companys fleet of self-driving cars has logged 2.5 million miles on public roads, which it says equates to over 300 years of human driving experience. Its arithmetic is murky, however, since Americans alone log more than 3 trillion miles everyyear. In any event, some experts believe that a transition to fully autonomous cars the ones you nap in, rather than paying at least some attention to the road could be decades away.

Obviously, theres a lot at stake in the case for big, brash Uber. The company has built its reputation as a juggernaut by flouting local car-hire regulations and bullying municipal officials who dare to stand in its way. Google may not be as inclined to back off as your city alderman.

As my colleague Tracey Lien observed Friday, the lawsuit capped a bad stretch for Uber. That started with a boycott of the firm after it was perceived to have taken advantage of a taxi drivers strike at New Yorks JFK airport to protest President Trumps immigrant ban. It was followed by a devastating picture of a sexual harassment culture at Uber headquarters posted online by a former engineer, Susan Fowler Rigetti. Now comes Waymos unusually detailed accusation of intellectual property thievery.

The case may also underscore the weakness of Ubers claim to a $70-billion valuation in the private venture market. That valuation had been based on the expectation that Uber was poised to radically reform the transportation-for-hire economy by shouldering vehicle-owning taxi companies and individuals out of the way, replacing them with independent drivers using their own cars. If Google is to be believed, Uber now puts such stake in owning its own capital assets that it waswilling to pay $680 million for the necessary (allegedly stolen)technology.

Moving from a business model in which the company essentially owns nothing but skims a vigorish of 25% or more off the fares paid to its independent contractors, to one in which it owns and must continue to develop a fleet of its own vehicles represents a major change of direction. As transportation expert Hubert Horan observed in a detailed critique of Uber last December, it is unclear why investors would wager billions on the prospect that it will eventually be able to design and build highly sophisticated vehicles more efficiently than competitors such as Google, Tesla, Toyota, Mercedes-Benz, Ford and General Motors.

The allegations in the lawsuit imply that Uber went to great lengths to obtain its driverless technology. The core allegation concerns its light detection and ranging system, orLiDAR, which coordinates and interprets in real timethe signals returned from laser beams bouncing off objects in the real world. Waymo says its LiDAR is the most advanced in the fieldanda trade secret.

According to the lawsuit, while Levandowski was managing Waymo, he was plotting to start his own competing company. Starting in December 2015, he downloaded 14,000 Waymo files, including specifications for its LiDAR system, from a company laptop, the lawsuit alleges,then he erased and reformatted the laptop to eliminate evidence of what he had done. Within weeks he resigned from Waymo and launched Otto. Other Waymo employees soon followed him out the door, taking other trade secrets, the lawsuit says.

Then, last December, a Waymo employee was sent a copy of an email destined for the Otto team. It happened, the lawsuit said,to include a rendering of an Otto circuit board that bore a striking resemblance to a circuit board design that Levandowski had downloaded.

Keep up to date with Michael Hiltzik. Follow@hiltzikmon Twitter, see hisFacebook page, or emailmichael.hiltzik@latimes.com.

Return to Michael Hiltzik's blog.

See the original post:

Facing a lawsuit from Google over driverless car technology, Uber may finally have met its match - Los Angeles Times

Manufacturing small proteins known as peptides is usually very time-consuming, which has slowed development of new peptide drugs for diseases such as cancer, diabetes, and bacterial infections.

To help speed up the manufacturing process, MIT researchers have designed a machine that can rapidly produce large quantities of customized peptides. Their new tabletop machine can form links between amino acids, the buildings blocks of proteins, in about 37 seconds, and it takes less than an hour to generate complete peptide molecules containing up to 60 amino acids.

You can dial in whatever amino acids you want, and the machine starts printing off these peptides faster than any machine in the world, says Bradley Pentelute, the Pfizer-Laubach Career Development Associate Professor of Chemistry at MIT.

This technology could help researchers rapidly generate new peptide drugs to test on a variety of diseases, and it also raises the possibility of easily producing customized cancer vaccines for individual patients.

Pentelute is the senior author of a paper describing the new system in the Feb. 27 issue of Nature Chemical Biology. The papers lead authors are graduate students Alexander Mijalis and Dale Thomas; other authors are graduate student Mark Simon, research associate Andrea Adamo, Ryan Beaumont, and Warren K. Lewis Professor of Chemical Engineering Klavs Jensen.

Fast flow

Using traditional peptide manufacturing techniques, which were developed more than 20 years ago, it takes about an hour to perform the chemical reactions needed to add each amino acid to a peptide chain.

Pentelute, Jensen, and their colleagues set out several years ago to devise a faster method based on a newer manufacturing approach known as flow chemistry. Under this strategy, chemicals flow through a series of modules that each perform one step of the overall synthesis.

The teams first version of a flow-based peptide synthesis machine, reported in 2014, sped up the process to about three minutes per peptide bond. In their latest effort, the researchers hoped to make the synthesis even faster by automating more of the process. In the earlier version, the person running the machine had to manually pump amino acids out of their storage bottles, but the new machine automates that step as well.

Our focus when we were setting out to design the automated machine was to have all the steps controlled by computer, and that would eliminate a lot of the human error and unreliability thats associated with someone doing this process by hand, Mijalis says.

Once a user enters the desired amino acid sequence, the amino acids are pumped, in the correct order, into a module where they are briefly heated to about 90 degrees Celsius to make them more chemically reactive. After being activated, the amino acids flow into a chamber where they are added to the growing peptide chains.

Its a very iterative process, where youre building up this molecular chain, one piece by one piece, Mijalis says.

As each amino acid is added to the chain, the researchers can measure how much was correctly incorporated by analyzing the waste products that flow into the final chamber of the device. The current machine attaches each amino acid to the chain with about 99 percent efficiency.

In my view, this approach opens up the field to the generation of peptide libraries that enable more complete structure-activity relationships of bioactive peptides in a matter of days, as well as extending this chemical approach to the synthesis of small proteins and protein domains, says Paul Alewood, a research group leader in chemistry and structural biology at the University of Queensland Institute for Molecular Bioscience.

It will be used in both academia and industry when commercially available instruments for this chemistry become widely available, says Alewood, who was not involved in the research.

Personalized chemistry

Once synthesized, small peptides can be joined together to form larger proteins. So far, the researchers have made proteins produced by HIV, a fragment of an antifreeze protein (which helps organisms survive extreme cold), and a toxin secreted by snails. They are also working on replicating toxins from other animals, which have potential uses as painkillers, blood thinners, or blood clotting agents. They have also made antimicrobial peptides, which scientists are exploring as a possible new class of antibiotic drugs.

Another possible application for the new machine is generating peptides that could be used as personalized cancer vaccines targeting unique proteins found in individual patients tumors. Thats exactly what our machine makes, and it makes them at scales that are all ready to meet this demand for personalized cancer vaccines, Pentelute says.

The MIT team is also interested in adapting this technology to make other molecules in which building blocks are strung together in long chains, such as polymers and oligonucleotides (strands of RNA or DNA).

We can start thinking about a personalized chemistry machine, Pentelute says. Its modular and its adaptable to all sorts of other chemistries.

Go here to see the original:

New technology offers fast peptide synthesis | MIT News - MIT News

The United States shares 5,525 miles of land border with Canada and 1,989 miles with Mexico. Monitoring these borders, which is the responsibility of U.S. Customs and Border Protection (CBP), is an enormous task. Detecting, and responding to, illegal activity while facilitating lawful commerce and travel is made more difficult by the expansive, rugged, diverse, and thickly vegetated geography that spans both often-crossed borders. To help mitigate the challenges to border surveillance, a group of researchers at MIT Lincoln Laboratory is investigating whether an airborne ladar system capable of imaging objects under a canopy of foliage could aid in the maintenance of border security by remotely detecting illegal activities. Their work will be presented at the 16th Annual IEEE Symposium on Technologies for Homeland Security to be held April 25-26 in Waltham, Massachusetts.

Requisite for effective border protection is timely, actionable information on areas of interest. Leveraging the laboratorys long experience in building imaging systems that exploit microchip lasers and Geiger-mode avalanche photodiodes, the research team developed and tested two concepts of operations (CONOPS) for using airborne ladar systems to detect human activity in wooded regions.

"For any new technology to be effectively used by CBP, an emerging sensor must bring with it a sensible deployment architecture and concept of operation," said John Aldridge, a technical staff member from the Laboratory's Homeland Protection Systems Group, who has been working with a multidisciplinary, cross-divisional team that includes Marius Albota, Brittany Baker, Daniel Dumanis, Rajan Gurjar, and Lily Lee. The CONOPS that the engineering team focused on were cued examination of a localized area and uncued surveillance of a large area. To demonstrate the approach, the engineering team conducted proof-of-concept experiments with the laboratory's Airborne Optical Systems Testbed (AOSTB), a Twin Otter aircraft outfitted with an onboard ladar sensor.

For cued surveillance, the use of an airborne ladar sensor platform (whether a piloted or unpiloted aircraft system) might be prompted by another persistent sensor that indicates the presence of activity in a localized area at or near the border. "The area of coverage for cued surveillance may be in the 1 km2 to 10km2 range, and the laboratory has already developed and demonstrated sensor technology that can achieve this coverage in minutes," Albota said.

Uncued wide-area surveillance sorties might be flown long distances and over timelines of days or weeks to establish typical activity patterns and to discover emerging paths and structures in high-interest regions. "The area coverage required under such a CONOPS may reach as high as 300 to 800 km of border, depending on the Border Patrol Sector and vegetation density," Aldridge explained, adding, "Although the current AOSTB's area coverage rate is limited by the aircraft's airspeed, the sensor can image such a region in a matter of hours in a single sortie."

As a start to their field tests to assess their CONOPS, the team flew data collection runs over several local sites identified as representative of the northern U.S. border environment. The sites contained a variety of low-growing brush, thin ground vegetation, very tall coniferous-trees, and leafy deciduous trees. For the tests, the team positioned vehicles, tents, and other camp equipment in the woods to serve as the targets of interest. "We made 40 passes at an altitude of 7,500 feet to allow for a spatial resolution of about 25 centimeters," Dumanis said. "In between each pass, we moved the concealed items so that we could perform post-process analysis for change and motion detection," Baker added.

In this post-processing stage, the team members enhanced the data captured during the flights so that human analysts could then inspect the ladar imagery. They digitally removed ground-height data to reveal the three-dimensional ladar point cloud above ground and then digitally thresholded the height (erased 3-D points above a certain height) to eliminate the foliage cover. The resulting images gave analysts Gurjar and Lee a starting point for approximating the locations of both the planted objects as well as objects that were already on scene.

Searching through vast quantities of ladar data to spot areas for careful inspection is a labor intensive task even for experienced analysts who can recognize subtle cues that direct them to the possible presence of objects in the imagery. For the ladar data to be efficiently mined, an automated method of identifying areas of interest is needed. "One of the ways to alert analysts to potential targets is to track changes in the 3-D temporal data," Lee explained. "Changes caused by vehicle movements or alterations in a customary scene can indicate uncharacteristic activity."

To begin a change detection approach to the discovery of potential targets of interest, the research team registered the before and after ladar data and then subtracted the before data from the after dataset. This process allowed some improvement in the visual identification of vehicles that appeared where there had been none before; however, even a skilled human analyst would find it difficult to spot the small changes that signaled the presence of a vehicle.

A change detection approach, therefore, must compensate for the challenge posed by clutter in the ladar data. This clutter comes from the nature of ladar collection in densely foliated environment. As light travels through gaps between foliage, it bounces off a surface of leaves, ground, or human-made objects. The returned light is collected by the ladar sensor to form the 3-D point cloud. Because the motion induced by a flying platform causes each ladar scan to travel through different configurations of gaps between leaves, different parts of the canopy and shrubbery are sensed by the ladar. "Much of the clutter in our change detection output is from the different levels of canopy detected from different ladar scans," explained Gurjar.

To make the ladar change detection data easier for analysts to search, the team looked to automated object detection, a well-established field in computer vision that has been applied to images and radar data. Since ladar data presents in three dimensions and has unique noise characteristics, the team had to enhance the established automated detection approach with a sum of absolute difference (SAD) technique that factors in the height differences used to construct 3-D ladar imagery. Trials of the SAD technique applied to simulated vehicles in a foliated environment demonstrated that the approach yielded high detection rates and has potential as an automated method for reducing the huge amount of ladar data analysts would have to scrutinize to discover objects of interest.

"Looking forward, we hope to improve the capabilities of automated 3-D change detection to be more robust to natural temporal changes in foliage, expand the number of automatically detected object classes, and extend automated detection capability to full 3-D point clouds," said Lee, with Aldridge adding that they are also interested in exploring alternative aircraft for hosting the ladar system.

In its strategic plan "Vision and Strategy 2020," the CBP has expressed the need to apply advanced technology solutions for border management. Continued development of Lincoln Laboratory's automated approach to using a low-cost ladar system for surveillance of foliated regions may in the future offer another tool that the Department of Homeland Security's CBP can deploy to monitor the growing volume of land border activity.

Continue reading here:

Foliage-penetrating ladar technology may improve border surveillance - MIT News