Thursday, 6 June 2019

Introducing Databricks Runtime 5.4 with Conda (Beta)

We are excited to introduce a new runtime: Databricks Runtime 5.4 with Conda (Beta). This runtime uses Conda to manage Python libraries and environments. Many of our Python users prefer to manage their Python environments and libraries with Conda, which quickly is emerging as a standard. Conda takes a holistic approach to package management by enabling:

  • The creation and management of environments
  • Installation of Python packages
  • Easily reproducible environments
  • Compatibility with pip

We are therefore happy to announce that you can now get a runtime that is fully based on Conda. It is being released with the “Beta” label, as it is intended for experimental usage only, not yet for production workloads. This designation provides an opportunity for us to collect customer feedback. As Databricks Runtime with Conda matures, we intend to make Conda the default package manager for all Python users.

To get started, select the Databricks Runtime 5.4 with Conda (Beta) from the drop-down list when creating a new cluster in Databricks. Follow the instructions displayed when you hover over the question mark to select one of the two pre-configured environments: Standard (default) or Minimal.

Why Databricks Runtime with Conda

Conda is an open source package & environment management system. Due to its extensive support and flexibility, Conda is becoming the standard among developers for managing Python packages. As an environment manager, it enables users to easily create, save, load, and switch between Python environments. We have been using Conda to manage Python libraries in Databricks Runtime for Machine Learning, and have received positive feedback. With Databricks Runtime with Conda (Beta), we extend Conda to serve more use cases.

For Python developers, creating an environment with desired libraries installed is the first step. In particular, the field of machine learning is evolving rapidly, and new tools and libraries in Python are emerging and are being updated frequently. Setting up a reliable environment poses challenges, such as version conflicts, dependency issues, and environment reproducibility. Conda was created to solve this very problem.  By combining environments and installation into a single framework, developers can easily and reliably set up libraries in an isolated environment. Building-in first-class Conda support in Databricks Runtime significantly improves the productivity of developers and data scientists on your team.

Our Unified Analytics Platform serves a wide variety of users and experience levels. We enable users migrating from SAS or R to Python but are still new to Python to Python experts. Our intention is to make managing your Python environment as easy as possible. In service of this, we offer:

  • Multiple robust pre-configured environments, each serving a different use case
  • A simple way to customize environments
  • The ease and flexibility to manage, share, and recreate environments at different levels of the Databricks product (Workspace, cluster, and notebook)

Not only do we want to make it very easy for you to get started in Databricks, but also very easy for you to migrate Python code developed somewhere else to Databricks. In Databricks Runtime 5.4 with Conda (Beta), you can take code, along with the requirements file (requirement.txt)  from GitHub, Jupyter notebooks, or other data science IDE to Databricks. Everything should just work out of the box. As a developer, you can spend little time worrying about managing libraries, and focus your time on developing applications.

What Databricks Runtime Conda 5.4 (Beta) Offers

Databricks Runtime 5.4 with Conda (Beta) improves flexibility in the following ways:

Enhance Pre-configured Environments

  • We are committed to providing pre-configured environments with popular Python libraries installed. In Databricks Runtime 5.4 with Conda (Beta), we introduce two configured environments: Standard and Minimal (Azure | AWS). In both environments, we upgraded Python base libraries, compared to Databricks Runtime (Azure | AWS)
  • Databricks Runtime 5.4 with Conda (Beta) allows you to use Conda to install Python packages. If you want to install libraries, you will benefit from the support Conda provides. Please refer to the User Guide (Azure | AWS) to learn how to use Conda to install packages
  • We are leveraging Anaconda Distribution 5.3.1
  • We upgraded to Python 3.7

Easy Customization of Environment

  • Databricks Runtime 5.4 with Conda (Beta) allows you to easily customize your Python environments. You can define your environment needs in a requirements file (requirements.txt), upload it to DBFS, and then use dbutils.library.install to build the customized environment in a notebook. You no longer need to install libraries one by one.
  • You can find sample requirements files and instructions to customize environments in User Guide (Azure | AWS)

Environment Reproducibility

  • In Databricks Runtime 5.4 with Conda (Beta), each notebook can have an isolated Python environment, mitigating package conflicts across notebooks.
  • You can use requirements.txt to easily reproduce an environment to a notebook.

Which Runtime I Should Pick

In the future, the Databricks Runtime for Conda will be the standard runtime. However, as a Beta offering, Databricks Runtime with Conda is intended for experimental usage, not for production workloads. Here are some guidelines to help you choose a runtime:

Databricks Runtime: We encourage Databricks Runtime users who need stability to continue to use Databricks Runtime.

Databricks Runtime ML: We encourage Databricks Runtime ML users who don’t need to customize environments to continue to use Databricks Runtime ML.

Databricks Runtime with Conda: Databricks Runtime 5.4 with Conda (Beta) offers two Conda-based, preconfigured root environments — Standard and Minimal —  that serve different use cases.

  • Standard Environment: The default environment (Azure | AWS). At cluster creation, you select the Databricks Runtime 5.4 with Conda (Beta) in the Databricks Runtime Version drop-down list. Aimed to serve Databricks Runtime users, the Standard Environment provides a ready-to-use environment by pre-installing popular python packages based on usage. A number of base Python libraries are upgrade in the Standard Environment. We encourage users of Databricks Runtime who need these upgraded Python libraries to try out the Standard environment.
  • Minimal Environment: Includes a minimal set of libraries to run Python notebooks and PySpark in Databricks (Azure | AWS). This light environment is designed for customization. We encourage Python users who need to customize their Python environment but run into dependency conflicts with the standard environment to try out the Minimal environment.

To use the Minimal environment, you select Databricks Runtime 5.4 with Conda in the Databricks Runtime Version drop-down list. Then follow the instructions to copy and paste DATABRICKS_ROOT_CONDA_ENV=databricks-minimal to Advanced Options > Spark > Environment Variables, which can be found at the bottom of the Create Cluster Page (see below). In the upcoming releases, we will simplify this step and let you choose the MInimal environment from a drop-down list.

What to Expect in Upcoming Releases

In the coming releases, we plan to keep improving the three key use cases Databricks Runtime with Conda serves.

Enhance Pre-configured Environments

Our ultimate goal is to unify cluster creation for all three runtimes (Databricks Runtime, Databricks Runtime ML, Databricks Runtime with Conda) in a seamless experience. At full product maturity, we expect to have multiple pre-configured environments serving different use cases, including environments for Machine Learning. In addition, we plan to improve the user experience by allowing you to choose a pre-configured environment in Databricks Runtime with Conda from a drop-down list. Finally, we will continue to update Python packages as well as Anaconda distribution.

Easy Customization of Environments

We plan to add support for using environment.yml (environment file used by conda with Libraries Utilities in notebooks. We also plan to support conda package installation in Library Utilities in notebooks and in cluster-installed libraries. Currently both use PyPI.

Easy Reproducibility of Environments

We plan to make it very easy to view, modify, and share environment parameters across users. You can save an environment file in Workspace, and easily switch between environments so that the same environment can be replicated to a cluster at cluster creation.

Upgraded Python Libraries in Databricks Runtime 5.4 with Conda (Beta)

Please find the list of pre-installed packages in Databricks Runtime with Conda (Beta) in our release notes (Azure | AWS).

Read More

  • Databricks Runtime 5.4 with Conda (Beta) release notes (Azure | AWS)
  • Databricks Runtime 5.4 with Conda (Beta) User Guide (Azure | AWS)

 

 

 

 

 

--

Try Databricks for free. Get started today.

The post Introducing Databricks Runtime 5.4 with Conda (Beta) appeared first on Databricks.

Wednesday, 5 June 2019

How to Choose the Right Data Storage System for Team Work

My team works with a big number of media data: video, rendering, photos, illustrations. To ensure the best teamwork, we need constant access to all these files. At some point, we no longer have enough storage available on our own server, and we started looking for cloud storage that satisfies all our needs.

In this post, I will compare popular data storage services for business: DropBox, Google Drive, Citrix ShareFile, and Microsoft OneDrive.

Let’s see our requirements for a cloud storage service:


Unlimited data. We have a lot of data, about 10TB on an average day. I do not want to constantly think about how much space we need to buy this month or why the quota suddenly ended.



Versioning of files and logging. Github has taught us that all changes can be seen and rolled back. Therefore, I want that situation with our files becomes exactly the same: any change or removal should be reversible and easy to control.



Access rights. No more shared folders available to everyone. Each employee must have his own area of ​​visibility and access.



Upload without registration. My clients no longer should look for data hosting services to send us a large file. Files should be immediately uploaded to our repository ...


Read More on Datafloq

8 Industrial IoT Trends of 2019 That Cannot Be Ignored

From manufacturing to the retail sector, the infinite applications of the industrial internet of things are disrupting business processes, thereby improving operational efficiency and business competitiveness. The trend of employing IoT-powered systems for supply chain management, smart monitoring,  remote diagnosis, production integration, inventory management, and predictive maintenance is catching up as companies take bold steps to address a myriad of business problems.

No wonder, the global technology spend on IoT is expected to reach USD 1.2 trillion by 2022. The growth of this segment will be driven by firms deploying IIoT solutions and giant tech organizations who are developing these innovative solutions.

To help you stay ahead of the curve, we have enlisted a few trends that will dominate the industrial IoT sphere.

1. Cobots Are Gaining Popularity

Digitization is having a major impact in the industrial robotics segment as connected cobots or collaborative robots, making their place in the smart manufacturing ecosystem. This trend is improving the efficiency of operations and the reliability of the production cycle.

IIoT is making robots mobile and collaborative, offering technologies, such as self-driving vehicles (mobile collaborative robots), machine vision (part identification), and additive manufacturing that can boost production efficiency and business growth with an excellent ROI. No wonder, ...


Read More on Datafloq

Tuesday, 4 June 2019

HHS Splits Award of $49M for AI Services Between Large, Small Providers

By AI Trends Staff

The Health and Human Services Department awarded 57 spots on its Intelligent Automation/Artificial Intelligence, or IAAI, contract, a $49 million vehicle.

The department awarded spots to 28 small business and 29 large vendors, according to data obtained by Nextgov.

General Dynamics IT is providing artificial intelligence, intelligent automation, blockchain, machine learning, natural language processing and robotic process automation, the company said in a news release on May 30. HHS hopes the partnership will help to improve its performance of key functions, such as fraud prediction and identifying irregularities in HHS’s systems.

“We are excited to join HHS in their journey to a more efficient future,” said Kamal Narang, vice president and head of General Dynamics Information Technology’s health sector, quoted in fedscoop.  “GDIT will utilize our expertise in AI, blockchain and other next-generation technologies to help transform HHS’ business processes and enhance their mission delivery.”

A number of other D.C.-based contractors were also included in the award, including Leidos Holdings, and Booz Allen Hamilton, according to an account in the Washington Business Journal.

The awarded contract is a five-year indefinite delivery/indefinite quantity contract designed to provide the department’s Program Support Center (PSC) with a range of next-generation technologies to pilot, test and implement across a range of shared services functions to help drive more efficiency.

The contract proposal was put out by the PSC, which offers services to 24 government agencies on a competitive fee-for-service basis. The award announced on May 30 allows PSC to compete for orders with more than 50 other companies to provide solutions including AI, robotic process automation, blockchain and machine learning.

Other awardees in the contract include Accenture Federal Services, Parsons Government Services, Northrop Grumman Corp., Unisys Corp. and CGI Federal. It also included tech consulting firms such as Octo Consulting Group Guidehouse and Deloitte.

“PSC believes that IAAI solutions will be doing everything from reducing backlog and cutting costs to performing functions; such as predicting fraudulent transactions and identifying critical suspects via facial recognition, which are considered difficult for an individual to complete on their own,” officials said in the contract’s statement of work.

“HHS is among the largest data producers in the world. The department’s healthcare and financial data alone exceed petabytes per year,” Dave Vennergrund, senior director of data and analytics for GDIT, one of the awardees, told Nextgov. “This data is rich in value and provides the fuel for machine learning models.”

See the source articles in fedscoop, the Washington Business Journal, and nextgov.

Defect Management Tools - Their Importance in the Era of IoT

Bugs and defects appearing in an application are a norm for software testers. Bugs indicate that software development is steering in the right direction. However, detection of bugs is a key to achieving software quality but managing these issues/bugs is quite challenging. There are different bug management tools that are used to gather all the activities related to bug management under one place. An effective tool helps in recording, managing and tracking all the testing efforts while providing its access to software testers and developers.

IoT is the talk of the town in this century and digital transformation efforts have made them commonly available in various sectors. They include home appliances, mobiles, software, electronics, etc. allowing users to connect and share the data. The existence of IoT devices calls for different bug tracking tools that can ensure its efficient use.

Let’s have a look at the importance of defect management tools in the era of IoT:

1. Dashboard Monitoring & Diagnostics

Software testers utilize bug tracking tools that allow easy reporting and tracking of defects by using an interactive dashboard. All defects are detected within the dashboard and software testers monitor and diagnose these defects to improve software quality and effective functioning of the ...


Read More on Datafloq

Human Driver Licensing Versus AI Driverless Certification: The Case of AI Autonomous Cars

Lance’s note to his readers: I recently did an exhilarating podcast with Princeton’s Alain Kornhauser, along with noted syndicated tech journalist Fred Fishkin, covering a variety of current trends and news about autonomous cars, you might find it informative and engaging: https://soundcloud.com/smartdrivingcar/smart-driving-cars-episode-110

By Lance Eliot, the AI Trends Insider

And now, let’s get into the topic of human driver licensing versus the controversial topic of whether or not there should be AI self-driving driverless autonomous car driver certification or licensing.

As they say, on with the show.

Rites of passage. In some countries and locales, you need to hunt down a lion to undertake a rite of passage. Here in California, it seems like our mainstay rite of passage is getting your driver’s license or driver’s permit. Up until getting a vaunted license to drive, you are considered somewhat inessential and possibly even an annoyance, since you need to either harangue someone to give you a lift or you need to find an alternative form of transportation other than a car.

For my children, it was exciting to see them earn their driver’s permit. After years of chauffeuring them around, which was a delight to do and I miss it sorely, there was a great sense of pride that they could now drive a car. They could pretty much go where they wanted and when they wanted, though subject to a personal curfew and a state-based curfew (legally, for the first twelve months, teens cannot drive after 11 p.m. and before 5 a.m., though exceptions such as for work are allowed).

Some teenagers complain about the gauntlet they need to confront when seeking to become a licensed driver. I’d heard scuttlebutt among some teenagers at my children’s school that maybe there was a kind of conspiracy going on, in which adults were scheming to keep kids from driving cars. Rumors were that only a tiny percentage of teens that passed the licensing requirements would actually get anointed with a license.

I mulled over this theory and found out that by-and-large most teens do get their driver’s licenses (assuming they take things seriously), and in terms of the licensing roadblocks trying to keep them from doing so, well, it sure seems prudent to make sure that anyone licensed to drive is properly qualified and ready to do so. That’s not a conspiracy, that’s just plain sensible, I’d say.

Historians believe that Karl Benz in 1888 was the first person to ever receive a driver’s license. Just three years earlier, he had patented what some suggest was the first practical automobile, and after some complaints by those that were annoyed by the stink and smell of his motorcar that he personally drove around, the government decided to establish the need to first obtain a written permit to drive a car. He dutifully got his permit and continued to drive his Motorwagen. In case you are wondering, yes, Karl Benz is the same person as named in the now popular Mercedes Benz moniker.

Nearly twenty years later, the United States put in place a driver’s licensing law that first took hold in New York. It was on August 1, 1910 that the first such licenses were granted, though it was only needed by professional drivers such as chauffeurs. A few years later, starting in 1913, New Jersey opted to force all drivers to become licensed and included the need to pass a mandatory test before you would be granted the license.

A driver’s license has become more than just a signification that you are licensed to drive a car. As many college students can attest, having a driver’s license is the key to survival when being a partying and bar hopping university student. No driver’s license, no means to get that keg of beer for your frat or sorority. Or so it would seem, yet the reality is that there is a vast underground market of fake ID’s that are readily available. In any case, the point is that a driver’s license is often used as a form of identification, not necessarily solely for proof of being able to drive a car.

When you consider the act of driving a car, it becomes apparent that there are two major elements of certification about the matter. There is the aspect of ensuring that the human driver is ready, able, and formally certified to drive, which is achieved by the driver’s license or permit, plus there is the need to make sure that the car itself is also properly certified.

You likely don’t put much thought toward the fact that the car you are driving had to meet various federal regulatory guidelines in order to be on the public roadways. Per the Federal Motor Vehicle Safety Standards (FMVSS), all cars on our roads are supposed to satisfy a rather hefty number of federally mandated requirements. The requirements encompass the full scope of car elements, including the car design, the construction of the car, the performance of the car, it’s safety, its durability, and so on.

Those brakes on your car, yep, they come under FMVSS standards number 106 and other provisions. Even something that might seem to be trivial, such as the windshield wipers, those too come under the FMVSS regulations (item number 104). There are various international standards and other countries also have their standards. If you are an auto maker, you face a rather long list of requirements that your car needs to meet, often times varying by country and forcing you to make some tough decisions about how you design and develop your car, along with whether it is worth the trouble to sell your car in places that might have regulations you aren’t interested in meeting.

In the United States, we have chosen to use the federal government to certify cars, and in contrast chosen to have the states themselves certify the human drivers of those cars. Does it have to be that way? Nope. It’s custom and tradition.

We could certainly decide to have the feds certify cars and certify drivers, doing both, if that’s what we wanted to do as a society. The states though have historically held onto the driver certification aspects. You could assert that it makes sense to have the states certify drivers since each state also comes up with their own driving regulations. The tricky part is that the federal transportation laws regulate travel between the states and in the end, it has fortunately turned out that most states tend to have by-and-large similar regulations, easing the driving across the vast landscape of the United States.

If you are wondering whether perhaps states opt to regulate cars and also regulate drivers, the basis for having the feds regulate cars themselves is to again foster an across the United States ease of using your car. Imagine if every state had idiosyncratic rules about the design and make-up of a car. It could be a nightmare for you when taking a family trip that took you through numerous states. Upon entering each state, you might be violating some car regulation in that state and find yourself never making it across those states unimpeded.

There is also the case to be made that by having one entity, the federal government, regulating cars the matter is simplified for the auto makers (else they’d need to be negotiating endlessly and/or trying to comply with each of the 50 different states individually). One could also claim that there is an economy of scale to be had by embodying all of the car regulations into the hands of the federal government. The states seem to be generally satisfied with the arrangement and aren’t necessarily going out of their way to try and change the status quo on that aspect.

Overall, it seems to be a pretty workable approach. The federal government certifies the cars that are allowed onto our roadways. The states certify the human drivers of those cars. I suppose it is the yin and yang of the arrangement.

Pertinent Aspects Of Human Driver Licensing

Let’s dig a bit more deeply into how the states tend to certify drivers.

I’m going to focus on the licensing of human drivers for conventional cars. I mention this point because there are other kinds of driver licensing needed to drive trucks, or buses, or to drive a motorcycle, and to drive other non-car types of vehicles.

Most of the driver’s licensing requirements are about the same, though the type of vehicle can cause the driver licensing steps to differ, and some would say become more stringent too for when driving a bus or similar multi-person vehicle of a larger size (this seems sensible, since one could argue a bus is a more complicated vehicle and also one that holds the lives of a multitude of passengers at once).

There are about nine major steps when getting a driver’s license. Again, keep in mind that the number of steps and their specific actions will differ by state. I’ve just tried to come up with the major steps and do so to generally cover the gamut of what takes place. As they say, your mileage may vary depending upon which state you live in.

First, you need to achieve a minimum age to seek out getting a driver’s license, which is typically around the age of 16 or 17. Why do we do this? The belief is that only someone that has reached the mid-teens is cognitively ready to drive a car. You need to be able to mentally contend with the driving act, along with the roadways and the crazy antics of other drivers, plus be on the watch for pesky pedestrians. There’s a lot of cognitive work involved.

You also need to be able to appropriately deal with the physical driving aspects, such as being able to use the brakes, the accelerator pedal, and the steering wheel. This requires not just physical size to reach those driving controls, but also a dexterity and a command of your body agility to do so. Your mind has to be able to tell your arms and hands to steer, and your arm and hands need to respond promptly and accurately.

The age requirement is really a surrogate for whether or not you as a human seem ready to tackle the serious and life-or-death decisions that need to be made when driving a multi-ton car on our roads. Admittedly, I’ve seen some 16- or 17-years old’s that are driving a car and for which I would not have trusted them to do something as simple as rake leaves or chew gum. Thus, it is not just the age per se that decides whether you can drive, and there are other steps involved that hopefully wean out those that aren’t ready for the task.

As an aside, my children, now past their teens, interestingly look upon teens in a suspicious manner as drivers. With several years now of driving under their own belts, they have expressed amazement at times that we as a society seem to let many teens drive among us. I’m pretty sure they would support moving up the driving age to at least 18 or even more. Seems that one’s perspective changes over time on such matters, I suppose.

In any case, once you’ve reached the minimum age, you can formally apply to get a driver’s license. You don’t have to do so, in the sense that if you want to wait until you are older to drive, you can wait until then to apply.

I recall debates with other parents that were insistent they did not want their own children to start driving until at least 17 or possibly waiting until the age of 18. The rationale was that it was safer for their own child and they were trying to do the right thing by deferring them driving until a riper age (the offspring did not necessarily see things this way, as you might imagine).

You need to then show proof of residency in the state that you are applying for a driver’s license. This makes sense due to my earlier point about the states being in the driver’s seat when it comes to the licensing of human drivers. States generally will only license those that live in their own state. There are various exceptions related to out-of-state drivers and other complications, but the norm is you need to be a resident in the state for which you are seeking a driver’s license.

Often, you need to show proof that you have taken a driver’s education (driver ed) course. When I was in high school, the school provided a driver’s ed class that the students could take. We even had car simulators that we used in a lab. These were pod-like contraptions that had a steering wheel and the brake and accelerator pedal. There was a film shown on a large screen at the front of the room that portrayed a driver’s view of driving on a freeway or street or wherever. You were supposed to drive your fixed-in-place pod as though you were abiding by what was shown on the screen.

Today’s teenagers would likely recoil in shock and horror at the primitive nature of these simulators. You could actually steer however you wanted, and it had no impact whatsoever on what the film that was showing at the front of the room. The only way that you could get caught messing around was the instructor had a master panel that displayed the actions of each student in their simulator. If you were ramming on the brakes when you were supposed to be accelerating, the instructor could see via red and green lights on their display board what you were doing, and you’d hear the instructor yell out your name (you didn’t want that to happen, I assure you).

One day, we were all properly steering and using our pedals, which after a while became second nature and you went along with it, no whining, no complaining, just do what the instructor expected you to do. I remember this particular day well because of what happened next. The room was usually relatively quiet as all the students sat in their pods and concentrated on following along what was being shown on the front screen. Unbeknownst to us, this particular film involved a car that was going to crash into the back of a truck.

Keep in mind that the crashing of the car would simply be that the film at the front of the room would suddenly turn a corner visually and you’d smash into a truck. There was nothing physical about it for those of us in our pod simulators. No tactile feedback of any kind. Well, the instructor decided that maybe we should have some kind of feedback. Just when the film showed us all ramming into the truck, he had arranged for a bunch of metal trash cans to be stacked up at the back of the room, behind us all, and he kicked over those metal clanking cans. I dare say everyone’s heart stopped. We thought somehow, we had actually hit that truck!

Some states don’t require any driver’s ed, some do. Some require it depending upon the age at which you apply. Of those states that do require driver’s ed, some allow a wide variety of means to fulfill the requirement, while other states are specific about what the content must be, how many hours long it is, and so on.

Another step that you normally must undertake is agreeing to the rules of the road and acknowledging that you are being granted a privilege to drive in that state. This is something we often don’t give much attention toward.

Driving Is A Privilege, Not A Right

Many people seem to think that you have a right to drive a car, as though there is an amendment to the constitution that says every person has the constitutional right to drive. Kind of funny to think that the founders of our constitution might have envisioned a day in which cars would be roaming our land and they might have snuck something such as: We, the People of the United States, in Order to form a more perfect Union, do declare that all shall have the right to drive a car.

Anyway, the state can revoke your privilege to drive. This is handy as a means of trying to enforce the rules of driving. Abide by the rules, and you are allowed to legally drive. Don’t abide by the rules, and you’ll lose those driving privileges you were granted. Obviously, it then becomes crucial that drivers know what the rules are, else they can hardily be expected to follow them.

For most of the states, you need to pass several kinds of tests to be able to get your driver’s license.

There is usually a vision test. This is to make sure that you can see the road and ably navigate the world in which you are driving. When you ponder the nature of the driving task, it is very visually oriented. We depend almost entirely on what we see. Sure, you are supposed to be listening too, such as hearing the sound of an approaching siren of a police car, but overall it is sight that preoccupies our senses when driving a car.

There is usually a written test that questions your knowledge about driving of a car. Most such written tests involve identifying various street signs and roadway infrastructure situations. There are also questions about the rules or laws in that state about driving, which ties back to the earlier step about agreeing to abide by the rules or laws. In theory, the driver’s written test will help ensure that you have studied the rules and laws. To pass the written test, you need to score a certain minimum number right, and it is typically a timed test. If you fail, there are often a limited number of retries allowed right away, and then a waiting period to take the written test again.

These written knowledge tests are often only 20 to maybe 50 questions in size. Can you really test to make sure that a person taking the test knows all of the various street signs (hundreds of those), various traffic or roadway infrastructures (hundreds of those), and the entire body of rules and driving laws (thousands of those), doing so with just a few handfuls of written questions? It does seem suspect.

Admittedly, it is just a random selection of questions and hopefully is sufficient to detect whether someone knows much or not about the whole matter. Regrettably, there are ways to study just for the test, and for which avoids having to know the larger body of elements, but anyway it seems to be sufficient and somehow does the trick. I know adults that take the test and carp that they had to read the DMV (Department of Motor Vehicles) booklet about our driving regulations, insisting that they’ve been driving for decades and must ergo know the rules subliminally (therefore not needing to take the test, while the counter-argument is that you should readily pass the test if the rules are so well-ingrained in you).

There is usually a driver’s roadway test. This involves getting into your car and having a human passenger that is your tester, grading you as you drive around the local area of the test. I’ve often wondered whether these testers enjoy this job or live in continual fear of it. You are putting your life into the hands of a complete stranger. In the case of an as yet licensed driver, you already know that the driver is not yet versed in driving and the odds of something going afoul is sizable. Seems like you would need to have nerves of steel.

The driving or roadway test often involves executing stipulated driving actions. You might need to do several right turns, and several left turns. You might need to showcase a U-turn. You are expected to be obeying all the road signs and paying attention to those road signs. You are likely to be asked to drive in a neighborhood and then onto a busy street or highway. The whole thing is nerve wracking for the driver since they are betting their entire dream of getting a license on a few minutes of driving time.

I remember how nervous my own children were. Not because they didn’t feel like they could drive well, but simply due to the notion of someone eagle-eye watching your every move and dissecting the littlest wrong movement. The situation does not lend itself to calm driving. There is also the random element of the tester and their personality and perspective. I’ve seen circumstances of a “cool” tester that purposely tries to put the driver at ease, while there are other “harsh” testers that seem to go out of their way to intimidate and unnerve the driver.

I remember that in my own case I happened to get one of those harsher testers. I know it sounds like sour grapes, but I swear to you its true that the tester was over-the-top in terms of being deriding. I did pass the driving roadway test, but I lost a point. When I was making a right turn onto a highway, I judged that it was safe to do so, eyeing oncoming traffic and waiting until a good moment presented itself. The tester deducted a point because they felt that though I made a safe choice, in their judgment they would have waited longer. I didn’t dispute the matter, since I had passed, though it has stuck in my craw all these years.

Driver Licensing And Also Certifying The Car

Back to the steps about getting certified as a driver, once you’ve done all of the aforementioned steps, you do a few other paperwork things and ultimately are granted the driver’s license. In some states, there is an initial probationary period, during which the tiniest infraction can get your driver’s license revoked. In some states, you are able to drive before you take the tests, doing so on a probationary basis, though this usually requires that a licensed driver be in the car with you whenever you are driving.

If it’s been a long time since you got your driver’s license, I’m guessing that the aforementioned steps bring back either fond memories or memories you’d just as soon forget.

Recall that I’ve mentioned that there is the car driver certification and there is the car certification. In terms of the car certification, though a car might be certified to be on our roadways, this does not mean that the car is “perfect” in terms of how it works and will perform on the roads. When buying a car or leasing one, most people will often seek out reviews of the brand and model of car, helping them to be aware of the strengths and weaknesses of the car.

You can be a good driver that is driving a “lousy” car. Or, you can be a bad driver that is driving a “good” car. It’s a duality.

The duality is important to keep in mind. Today, as you know, cars don’t yet drive themselves (more on this in a moment). Humans drive cars. If a car is unsafe, the driver can be at the whim of what the car can or cannot do. A seasoned race car driver, presumably an expert-level driver, can be undermined by a car that is not performing well. Likewise, a car that is well-tuned and operating at top shape, can be undermined by the acts of a poor driver, one that is either not versed in the driving of that car brand or model, or one that is distracted while driving, or drunk, etc.

Whenever there is a car accident, you cannot immediately leap to a conclusion that it was caused by the driver per se, since it could be that the car itself failed and there was no action feasible by the driver to avoid getting into the accident. You need to consider the car and what its condition was, along with the driver and their condition. Keep this duality in mind as I progress further into this discussion.

AI Autonomous Cars And The Question Of AI Needing A Driver’s License

What does this have to do with AI self-driving cars?

At the Cybernetic AI Self-Driving Car Institute, we are developing AI software for self-driving cars. One of the most vital and somewhat vexing questions that we face as a society involves the certification of AI self-driving cars.

Allow me to elaborate.

I’d like to first clarify and introduce the notion that there are varying levels of AI self-driving cars. The topmost level is considered Level 5. A Level 5 self-driving car is one that is being driven by the AI and there is no human driver involved. For the design of Level 5 self-driving cars, the auto makers are even removing the gas pedal, brake pedal, and steering wheel, since those are contraptions used by human drivers. The Level 5 self-driving car is not being driven by a human and nor is there an expectation that a human driver will be present in the self-driving car. It’s all on the shoulders of the AI to drive the car.

For self-driving cars less than a Level 5, there must be a human driver present in the car. The human driver is currently considered the responsible party for the acts of the car. The AI and the human driver are co-sharing the driving task. In spite of this co-sharing, the human is supposed to remain fully immersed into the driving task and be ready at all times to perform the driving task. I’ve repeatedly warned about the dangers of this co-sharing arrangement and predicted it will produce many untoward results.

For my overall framework about AI self-driving cars, see my article: https://aitrends.com/selfdrivingcars/framework-ai-self-driving-driverless-cars-big-picture/

For the levels of self-driving cars, see my article: https://aitrends.com/selfdrivingcars/richter-scale-levels-self-driving-cars/

For why AI Level 5 self-driving cars are like a moonshot, see my article: https://aitrends.com/selfdrivingcars/self-driving-car-mother-ai-projects-moonshot/

For the dangers of co-sharing the driving task, see my article: https://aitrends.com/selfdrivingcars/human-back-up-drivers-for-ai-self-driving-cars/

Let’s focus herein on the true Level 5 self-driving car. Much of the comments apply to the less than Level 5 self-driving cars too, but the fully autonomous AI self-driving car will receive the most attention in this discussion.

Here’s the usual steps involved in the AI driving task:

  •         Sensor data collection and interpretation
  •         Sensor fusion
  •         Virtual world model updating
  •         AI action planning
  •         Car controls command issuance

Another key aspect of AI self-driving cars is that they will be driving on our roadways in the midst of human driven cars too. There are some pundits of AI self-driving cars that continually refer to a utopian world in which there are only AI self-driving cars on the public roads. Currently there are about 250+ million conventional cars in the United States alone, and those cars are not going to magically disappear or become true Level 5 AI self-driving cars overnight.

Indeed, the use of human driven cars will last for many years, likely many decades, and the advent of AI self-driving cars will occur while there are still human driven cars on the roads. This is a crucial point since this means that the AI of self-driving cars needs to be able to contend with not just other AI self-driving cars, but also contend with human driven cars. It is easy to envision a simplistic and rather unrealistic world in which all AI self-driving cars are politely interacting with each other and being civil about roadway interactions. That’s not what is going to be happening for the foreseeable future. AI self-driving cars and human driven cars will need to be able to cope with each other.

For my article about the grand convergence that has led us to this moment in time, see: https://aitrends.com/selfdrivingcars/grand-convergence-explains-rise-self-driving-cars/

See my article about the ethical dilemmas facing AI self-driving cars: https://aitrends.com/selfdrivingcars/ethically-ambiguous-self-driving-cars/

For potential regulations about AI self-driving cars, see my article: https://aitrends.com/selfdrivingcars/assessing-federal-regulations-self-driving-cars-house-bill-passed/

For my predictions about AI self-driving cars for the 2020s, 2030s, and 2040s, see my article: https://aitrends.com/selfdrivingcars/gen-z-and-the-fate-of-ai-self-driving-cars/

Returning to the topic of the certification of AI self-driving cars, let’s consider key aspects on this controversial and as yet unsolved problem.

Should We Be Certifying Or Licensing AI Systems For Driving Purposes

First, take a look at Figure 1.

Today, as I already mentioned, the federal government certifies the car, while the state government certifies or licenses the human driver. I’ve earlier provided various reasons why this makes sense, beyond just the tradition of the matter.

Once we have true AI self-driving cars at a Level 5, who should be certifying or licensing the AI driver?

It’s a serious question. I say that it is serious because sometimes people laugh when I ask the question. There laughter stems from the notion that the AI is a kind of robot, though not one you necessarily see in the self-driving car (it’s not a walking-talking physical robot), and they find it hard to imagine that you would give a driver’s license test to a robot.

Yes, go ahead for a moment and make-up your own stories on this. There you are at the DMV, waiting patiently to take the written test to get your driver’s license, and standing in front of you is this six-foot-tall robot, clutching the DMV study booklet, also waiting in line to take the test. The robot looks furtively around.

Glancing at you, the robot whispers to you in a small mechanical sounding voice and asks you what the roadway sign is for a washed-out road. You are torn about whether to give the answer to the robot. Maybe if you help this robot to pass the test, you are helping to put something onto our roadways that you believe shouldn’t be driving on our streets. On the other hand, your worried that if you don’t answer the question, the muscular robot might pick you up by your neck and toss you outside the DMV. Yikes, those robots!

Returning to reality, let’s dispense with the robot imaginary for now.

Assume that an auto maker or tech firm has developed an AI self-driving car. The AI does the driving. There is no human driver for this brand of car. Any humans that get into the AI self-driving car are there strictly as passengers. They can tell the AI where they want to be driven, they can even offer suggestions to the AI about the driving act, yet nonetheless it is the AI that will be driving the self-driving car.

How do you as a human passenger know that the AI can readily, properly, and appropriately drive that AI self-driving car?

And, it’s not just the human passengers that care about this. If I am driving a car on the streets of Los Angeles, and there are AI self-driving cars driving on those same streets, how do I know that the AI can readily, properly and appropriately drive those self-driving cars?

Pedestrians on the streets of Los Angeles want to also know that the AI self-driving car coming down the road is being driven by an AI that can readily, properly and appropriately drive that self-driving car.

We have lots of humans that need to be reassured that the AI of the self-driving car “knows” what it is doing, including passengers in that AI self-driving car, and human drivers of other nearby cars, and human pedestrians that are nearby the AI self-driving car, and so on.

If you like, you could even extend this concern to non-humans, such as dogs and cats that might be running out into the street. Will the AI recognize them and drive in an evasive and safe manner as a human driver might?

You can further extend this to non-human’s non-animals, such as other AI systems that are driving other AI self-driving cars. Allow me to explain this point.

You’ve got auto maker X that makes model Q of an AI self-driving car. Another auto maker Z makes their own model R of an AI self-driving car. The AI that is driving Q is not at all the same as the AI that is driving R. Each of the auto makers has taken their own approach to making their own proprietary AI driving systems.

How does the AI of self-driving car model Q “know” that the model R is being driven by an AI that is readily, properly, and appropriately able to drive that self-driving car of auto maker Z?

The answer right now is that it is a wild west and there is not any as yet a defined or agreed upon means of certifying the AI driving skills of a self-driving car.

As a crude form of analogy, suppose someone took their teenager and opted to teach them how to drive a car, and at some point, the parent figured the teenager was ready to drive on our public roadway unimpeded and unaided. If the parent filled-out some paperwork and submitted it to a governing body, let’s say it was the state in which they were doing the driving, and the parent attested that the teenager was ready to drive, the state would review the documentation and then decide whether to allow the teenager to now be certified or licensed to drive.

There might not be any testing done by the governing body and they would rely solely on the documentation provided by the parent. Substitute the word “parent” for the auto maker, and substitute “teenager” for the AI driving systems, and that’s pretty much the situation we are faced with today as it relates to AI self-driving cars.

Who Will Do The AI Driving Certification

Before I unpack that aspect, let’s revisit the question of whom is going to be certifying the AI driving systems.

It could be that the states each take on the role of doing the driver certification — of course, this has been the role that the states undertake for human drivers. Some though wonder whether the states ought to be doing so. Maybe the federal government ought to be doing the AI driving systems certifications. Why, you might ask?

In the case of human drivers, there are millions of those. Furthermore, those millions of human drivers presumably need to be assessed by the state to make sure that those human drivers meet the requirements of the state and are versed sufficient to drive in that state. It’s kind of a factory of trying to deal with those millions of human drivers.

In the case of AI driving systems, presumably we’ll only have as many as there are AI self-driving car models. Let’s guess that it will be in the low hundreds, using the fact that today there are about 250 or so brands of cars in the United States.

The states no longer would seem to need to be geared up to handle millions of drivers (those humans), and instead would only need to focus on hundreds of drivers (AI driving systems). This also assumes a situation whereby AI self-driving cars become prevalent and there are less and less human drivers to be certified.

Perhaps it would be more economical to have the federal government certify or issue a driver’s license (or equivalent) for the AI driving systems, rather than the states. You could argue that based on my point earlier about duality, the AI is now wrapped into the car. The feds were already doing the certification of the car, and so the AI aspects are simply a natural extension. It would be likened to adding a new bell-and-whistle to a car and expecting that the feds would be certifying the new feature, rather than the states doing so.

You might counter-argue that the AI is not akin to adding a new kind of windshield wiper or anything of that ilk. The AI is driving the car. It is akin to the act of a human driver. Therefore, the certification of the AI needs to be undertaken by the state. Plus, since it is expected by the state that the AI will drive the car as required by the state regulations, the state has a stake in making sure that the AI is indeed ready for and appropriate to be driving in their state.

There are some that even extend this logic and suggest that perhaps the state ought to certify the car and the car driver (the AI) altogether, since the state is getting dragged into the certification of the AI. In other words, if we aren’t going to distinguish the AI from the car, and we consider the AI and the self-driving car to be one integral whole, maybe it is the case that the state should be certifying the entire duality. I mention this notion for completeness, but there are few that take such an extreme view.

It is a conundrum.

How Will The AI Driving Certification Take Place

While pondering this question about whom is going to be doing the certification, let’s also consider an equally juicy and vexing question of how the certification is to be undertaken.

Take a look at Figure 2.

For human drivers, I’ve already mentioned the nine major steps that they must normally perform to be able to get a driver’s license. Of those nine steps, there are some steps that involve testing by the certifying entity, the state, in terms of whether the driver (human) seems ready and able to be a licensed driver.

Humans usually need to pass a vision test. Right now, there is no equivalent to a vision test as administered by a certifying entity for an AI self-driving car. What does a vision test have to do with an AI self-driving car?

You could assert that the equivalent would be a test of the sensory capabilities of an AI self-driving car. An AI self-driving car has cameras and a subsystem devoted to processing visual images. Perhaps that vision processing system should be subject to testing by the entity that might do the certification of the AI driving system. It’s a “vision test” of an AI self-driving car’s vision capability.

Prudently, you would likely enlarge the notion of the vision test to include the other sensory capabilities of the AI self-driving car. We know that vision is crucial, and it is the primary sense used by humans to drive a car, meanwhile for an AI self-driving car it is likely there will also be radar, ultrasonic, LIDAR, and possibly other sensors too, all of which are crucial to the AI driving system. Each and all of those sensors might be broadly placed into the idea of a “vision test” equivalent for an AI self-driving car.

Human drivers usually need to pass a written knowledge test about their driving awareness and understanding. Should there be an equivalent for AI self-driving cars?

Right now, states are tending towards asking for documentation from the auto maker or tech firm, doing so to gauge what the auto maker or tech firm claims their AI system “knows” about. This though does not necessarily involve testing of those aspects. Instead, it tends to be more of a paper-based review, rather than any kind of “test” to ensure that the AI system has or does what the documentation claims that it does.

Should there be a “knowledge” test administered by the certifying entity to an AI driving system for which an auto maker or tech firm is trying to get certified to drive on the public roadways?

And, what about the infamous and terrifying roadway driving test that is administered to human drivers seeking a driver’s license. Should a certifying entity require that an AI self-driving car be taken on a road test by the certifying entity, similar to what is done with human drivers?

At first glance, I am guessing that you might be tempted to say that sure, the certifying entity ought to conduct the vision test, or more broadly tests of the AI sensory systems of the self-driving car, and ought to have the AI undertake some kind of written knowledge test, and ought to have the AI self-driving car undergo a roadway test. This seems like a prudent action. The more the merrier in terms of trying to make sure that the AI driving system is up-to-snuff.

We expect human drivers to pass the battery of tests, why not also expect the AI to do so.

Part of the muddiness involves what kinds of tests you would devise for these purposes. How extensive would the tests be? Where and how would the tests be conducted? Can you do sufficient testing to believe that the AI self-driving car is ready to be on the roadways?

Recall that I had earlier mentioned that human drivers are administered a written test of maybe 20 to 50 questions, and we agreed that this is a rather small subset of the wide range of knowledge that we expect humans to have about driving. Do we feel the same about having such a short shrift when testing an AI driving system?

For the roadway driving test, the human tester of the DMV perhaps observes the human driver for about 30 minutes or so. Would that be sufficient for assessing the capabilities of an AI driving system?

If you are going to do more extensive testing, the odds are that the state testing would require more elaborated testing development, far beyond the kind of testing done today with human drivers. One argument is that maybe have the federal government could do core testing of the AI driving system, and have the states do an augmented testing that examines the state-specifics. Or, another idea is that the federal certification would encompass the state-specifics and indicate which states the AI self-driving car was certified to drive in.

For human drivers, once you are a licensed driver in any of the states, you pretty much can readily drive in another state, doing so without having to get an additional driver’s license.

One concern about AI self-driving cars and testing would be that if a state certified the AI driving system, and suppose that another state does not acknowledge that as a valid certification for AI self-driving cars in their state, it would mean that when your AI self-driving car reached the border of that other state, the AI self-driving car would need to come to a halt and not enter into that state. If the AI self-driving car continued ahead into the other state, doing so would mean that the AI was unlicensed and illegally driving in that state.

Another factor to consider is whether or not we are expecting that the AI of a true AI self-driving car to be able to use common sense reasoning and otherwise have human-like qualities of a sort. If that’s the case, we might need to concoct some kind of Turing Test, which is a long-time and muchly debated means of trying to ascertain whether an AI system is human-like in its cognitive output.

For my article about common sense reasoning aspects, see: https://www.aitrends.com/selfdrivingcars/common-sense-reasoning-and-ai-self-driving-cars/

For the nature of the Turing Test in AI, see my article: https://www.aitrends.com/selfdrivingcars/turing-test-ai-self-driving-cars/

For the matter of the potential of singularity and AI, see my article: https://www.aitrends.com/selfdrivingcars/singularity-and-ai-self-driving-cars/

For my article about safety and AI self-driving cars, see: https://www.aitrends.com/selfdrivingcars/safety-and-ai-self-driving-cars-world-safety-summit-on-autonomous-tech/

More Twists Involved In The AI Driverless Certification

There are additional twists about the testing of AI self-driving cars for certification purposes.

One means of the roadway testing by a human tester would be to do the same as normally is undertaken with human drivers, namely driving on public roadways. This is practical because most DMV’s are not perchance next door to a closed track or proving grounds that could be used for the testing. That being said, some DMV’s require the driver to first do some parking lot driving, essentially akin to doing a limited closed track test.

So, would the AI driving system be tested on the public roadways or instead (or in addition) be tested on a closed track? If it is a closed track, would this be within the confines of state doing the certification, or could be it be outside of that state (since the state might not have such facilities available). Would the closed track be the only done within the proving grounds testing or would it be done in conjunction with public roadways testing?

Another approach to testing could be the use of simulations. The certifying entity might establish a simulation that encompasses driving in their state, and then have the AI driving system have to take that test. The advantage is that the simulation could force the AI to drive millions upon millions of miles, doing so far beyond what might be feasible in roadway driving tests. An obvious downside is that the simulation is not necessarily the same as actual roadway aspects, though you might combine some amount of simulation with some amount of actual roadway testing.

If the AI self-driving car makes use of remote operators, either for controlling the driving of the self-driving car in emergency circumstances or as an augmentation that provides at times guidance to the AI, would this aspect also be encompassed by the testing? One would assume so.

One concern about any of the testing is that the AI you are testing is not necessarily the same AI that will be driving the self-driving car.

If the AI has Machine Learning or Deep Learning capabilities, it is going to be changing over time, doing so by hopefully improving as it “learns” more and more about driving. We don’t know for sure that the AI is going to become better at driving and could actually get worse or introduce nuances that can get itself into trouble.

There are also going to be updates made to the AI system, typically by the use of OTA (Over-The-Air) electronic communications. The OTA is a two-way street. The AI self-driving car can upload data into the cloud such as the collected sensory data. And, the cloud of the auto maker or tech firm can download into the AI self-driving car any patches or updates. This can happen whenever the auto maker or tech firm believes such updates are needed.

Thus, whenever the certifying entity tests the AI driving system, it is only testing at a particular point in time. What about five minutes later? Five days later? Five months later?

Those that aren’t worried about this changing AI system aspect are often quick to point out that humans also change, and yet we don’t seem overly concerned about right away retesting them. Sure, when their driver’s license comes up for renewal every few years, depending upon what their driving records indicates and how long it has been since they were tested, they might be required to do a retest. These such pundits would say that the same can be applied to AI driving systems.

This is not an entirely satisfying argument though, as we aren’t as sure, as one would well argue, about how the AI will be changing versus in the aggregate about human drivers changing over time.

For my article about OTA, see: https://www.aitrends.com/selfdrivingcars/air-ota-updating-ai-self-driving-cars/

For more about Machine Learning and AI self-driving cars, see my article: https://www.aitrends.com/selfdrivingcars/machine-learning-benchmarks-and-ai-self-driving-cars/

For my article about simulations and AI self-driving cars, see: https://www.aitrends.com/selfdrivingcars/simulations-self-driving-cars-machine-learning-without-fear/

For more about closed track testing, see my article: https://www.aitrends.com/selfdrivingcars/proving-grounds-ai-self-driving-cars/

For the aspects of remote operators of AI self-driving cars, see my article: https://www.aitrends.com/selfdrivingcars/remote-piloting-is-a-self-driving-car-crutch/

Edge Or Corner Cases To Be Included

Here’s another interesting twist. When a certifying entity is trying to test an AI self-driving car, how far can they go in doing so? Can they go beyond the normal everyday driving acts and push the boundaries of what might happen while driving a car? These are at times referred to as edge or corner cases.

For example, suppose the front cameras of the self-driving car are suddenly unable to function, maybe obscured by dirt or perhaps they were hit by flying debris. How well does the AI handle this aspect? Is that a proper test by the certifying entity?

I know that some would argue that there is no equivalent for human drivers during their certification testing, namely the DMV tester doesn’t suddenly blind one eye of the human driver or toss mud onto the windshield. Again, I don’t buy into this kind of assertion, and instead believe that we can and should hold the AI to a higher standard of testing.

A somewhat brazen aspect being voiced by some pundits involves having a third party or some set of designated third parties that would do the certification testing of AI self-driving cars, rather than having a governmental entity do so.

On that point, you might be aware that Consumer Reports (CR) recently released their first-ever review of automated driving systems, which is not the same as autonomous driving but nonetheless signifies this notion of third-party testing aspects. Consumer Reports used their Auto Test Center track, along with nearby freeways, and put several of the latest models through a series of tests. They selected four such cars, the Cadillac CT6, the Tesla X/S/3, the Infiniti QX50/Nissan Leaf, and the Volvo XC40/XC60.

Consumer Reports ranked the tested cars based on a variety of factors including automated driving capability, performance, ease of use, and how the respective systems monitored and reacted to human drivers that should be engaged in the driving task. Here’s the Consumer Reports ranking of the four systems tested, listed from highest to lesser on their scale of assessment:

  •         Cadillac Super Cruise
  •         Tesla Autopilot
  •         Nissan/Infiniti ProPilot
  •         Volvo Pilot Assist

If you are interested in the details of how CR did the testing, take a look at: https://www.consumerreports.org/autonomous-driving/cadillac-tops-tesla-in-automated-systems-ranking/

The overall point being that another approach to the certification topic could involve enlisting private companies, non-profits, consortia, or other third parties to perform the certification. This might be done at the behest of a governmental agency or in conjunction with a governmental agency. There are various efforts underway by numerous entities aiming to see if this approach might be viable.

In terms of the states, California is one of a handful of states that has opted to vigorously pursue the matter of AI self-driving cars. Having adopted California Vehicle Code (CVC) section 38750, taking effect on April 2, 2018, the regulations cover the testing and use of self-driving cars in the state.

There are three major portions of the regulation:

  • Auto makers or tech firms can seek a testing permit, requiring the presence of a human driver
  • Auto makers or tech firms can seek a driverless testing permit
  • Auto makers or tech firms can see the deployment license aka public use permit

If you are interested in seeing the details, along with the checklists and forms required for a submission, take a look at: https://www.dmv.ca.gov/portal/dmv/detail/vr/autonomous/bkgd

Conclusion

We collectively need to ascertain how AI self-driving cars should be certified or licensed to drive on our streets. It’s a big deal.

For human drivers, the path to becoming licensed as a driver is well-worn and shown to be relatively reliable and valid.

AI self-driving cars present a novelty that the existing licensing has yet to adapt to. There is an inherent duality of the car and the AI that makes trying to separate out their testing problematic. We need to know that we have good cars and good AI drivers, and it would seem that we should go further than simply having the auto maker or tech firm provide us with their own testing results.

Currently, we all rely upon an entity to certify that drivers in the cars next to us are at least minimally certified to be able to drive a car. When you look over and see that empty driver’s seat and realize the car next to you is being driven by AI, you ought to feel some comfort that one way or another there was a bona fide and independent process of trying to test and ensure that the AI driving system is up to the task at-hand. That’s a rite of passage needed for AI self-driving cars.

Copyright 2019 Dr. Lance Eliot

This content is originally posted on AI Trends.

 

Healthcare Companies Stepping Up Use of AI to Beat Hackers, Eliminate Passwords

By AI Trends Staff

The WannaCry ransomware attack infected hundreds of thousands of computers in more than 150 countries in 2017, bringing several industries to their knees with malicious software designed to block access to files until a “ransom” was paid.

One industry hit hard was health care, including the National Health Service (NHS) in the U.K. and Merck in the U.S. One study found that last year, 78 percent of health-care providers reported a ransomware or malware attack. Health care data is a target for malware bad actors.

But health-care systems like hospitals and insurance providers are starting to fight back with powerful security tools using AI and machine learning, according to an account in The Week.

Healthcare providers are starting to figure out hacker patterns and blocking them before they can get their hands on patient records or data. They’re also saying goodbye to vulnerable passwords and relying on more secure methods, like biometrics such as fingerprints and eyeball scans.

Aetna is one early adopter of machine learning. The health insurance company replaced passwords with a behavior-based security system and some biometric protections in its consumer-facing apps. It put machine learning at the forefront by developing a risk engine that combines multiple pieces of data to authenticate users. This risk engine collects personal information, ranging from the operating system and apps you use to how you text or move within the application. It creates an individual risk score, which can change.

For example, if someone steals your smartphone, hackers won’t be able to log into Aetna’s app because the risk engine knows they’re suspiciously holding the phone differently than you do. It will then demand additional authentication information, like a PIN or fingerprint, which the hacker won’t have.

Aetna has launched a new security system for its consumer mobile and web apps that, in something of a twist, makes passwords optional.

Instead of a password or fingerprint being the only barrier to entry, Aetna’s new behavior-based security system monitors user devices and how and where a consumer uses that machine. Consumers can add biometric protection available on their devices.

“Passwords are a mainstay of conventional online authentication and are considered to be a binary control – if a consumer has the user ID and password, they are enabled to use the application,” said Jim Routh, chief security officer at Aetna, quoted in Healthcare IT News. “Binary authentication controls work well when the assumption is that only the consumer has the password and remembers it. That assumption, however, is no longer valid.”

“Criminals exchange passwords on the Dark Web and use a technique called credential stuffing to apply passwords to targeted web domains and automatically attempt authentication for tens of thousands of compromised passwords,” Routh explained. “Criminals are able to achieve a two percent hit ratio of account takeover using credential stuffing, helped by the fact that consumers reuse passwords across sites.”

Consequently, passwords are becoming obsolete as a primary means of authentication. Enterprises need to change their thinking on the use of binary controls in an authentication event at the initiation of an online interaction with a consumer and consider using many different attributes to confirm authentication, Routh said.

See the source articles in The Week and Healthcare IT News.

Here Are 3 Ways AI is Transforming Business Operations

Constant organization change is the new normal today. Few advancements promise to help organizations cross critical business chasms more profoundly than artificial intelligence. When it comes to business operations, artificial intelligence has enormous potential to catapult companies beyond the status quo and adapt to new dynamics. Here are three ways it does so:

  1. Enhancing the customer experience

According to research by Walker, customer experience will overtake price and product as the key brand differentiator by 2020. The centrality of customer experience in today’s business landscape has motivated many companies to adopt a more holistic approach to customer experience improvement.

Companies such as Intuit, EMC, and Adobe all have core customer experience leadership embedded in their sales, operations, and marketing organizations.

Artificial intelligence can help businesses craft highly-personalized customer experiences. Artificial intelligence can ingest and aggregate all of the various touch-points a customer has with a brand and determine what’s driving customers behavior. Are they eager to buy more? Why? Are they on the verge of churning to a key competitor? Why?

Armed with this information, business operations teams are afforded unprecedented insight into the end-to-end customer experience, far beyond that which can be gleaned from customer feedback surveys. According to Salesforce, 50% of consumers are likely to switch brands if a company doesn’t anticipate their needs.  By providing the right messages and marketing prescriptions to customers at the right time, AI promises to help organizations better understand customers and anticipate their needs.

  1. Enhancing the hiring process

Strategic talent acquisition is central to business operations. Yet the hiring process is fundamentally flawed. The recruitment and interview process is rife with bias and is largely driven by first impressions.

According to one recent study, despite increased awareness of discriminatory hiring practices, hiring discriminating against black workers in the U.S. has not declined in the past quarter century. The fact that employee referrals are the number one source of new hires, representing more than 50% of hires, underscores the limitations, biases, and potentially nepotism ingrained in current processes.

AI is already helping businesses transform their hiring operations and processes. It can help businesses cast a wider net by scanning large volumes of resumes and other talent sources to pinpoint highly qualified candidates for specific jobs. It can review job descriptions, interview transcripts, and call logs to uncover and correct hidden biases in the hiring process and accurately assess candidates.

Several tools already exist on the market. IBM Watson Candidate Assistant, for example, is a talent management solution that helps match potential candidates with the specific jobs that they will thrive in. When implementing AI recruiting tools, it’s critical that we ensure that tools are not prone to the same biases that humans fall victim to.

  1. Enhancing employee engagement and retention

Talent is the lifeblood of every thriving business. Amidst today’s competitive hiring landscape, retaining top talent is top-of-mind for all businesses. Employee engagement must be an integral component of everyday business operations. Yet most businesses are severely faltering in their efforts. According to a recent report by Gallup, the vast majority (85%) of employees are either not engaged or are actively disengaged at work.

AI offers enormous potential to boost employee engagement and retention. Using sentiment analysis technologies, biometrics, and other emerging technologies, AI-powered tools can help businesses confidently understand employee engagement.  It can surface insight into what is driving employee behaviors and generate powerful recommendations for motivating employees and rewarding behavior.

According to an analysis of Facebook employees, those who were most likely to remain at the company felt, comparatively speaking, they used their strengths more. Charting an empowering career path with skill development is critical to helping employees feel empowered.

Team performance also stands to benefit from AI. By monitoring interactions on social enterprise tools and workplace social channels, AI can reveal key insights into the degree and effectiveness of employee collaboration. Vibe from Vibe HCM, for example, analyzes keywords and emojis sent among employees on Slack to assess the morale of a given team.

Read the source article in Forbes.

OECD Releases Guidelines for Development of Trustworthy AI, Joining the Pack

Australia is among 42 countries that in late May signed up to a new set of policy guidelines for the development of artificial intelligence (AI) systems.

Yet Australia has its own draft guidelines for ethics in AI out for public consultation, and a number of other countries and industry bodies have developed their own AI guidelines, according to an account in Phys.org.

So why do we need so many guidelines, and are any of them enforceable?

The latest set of policy guidelines is the Recommendation on Artificial Intelligence from the Organisation for Economic Co-operation and Development (OECD). It promotes five principles for the responsible development of trustworthy AI. Given this comes from the OECD, it treads the line between promoting economic improvement and innovation and fostering fundamental values and trust in the development of AI.

The five AI principles encourage:

  1. inclusive growth, sustainable development and well-being
  2. human-centered values and fairness
  3. transparency and explainability
  4. robustness, security and safety
  5. Accountability.

These recommendations are broad and do not carry the force of laws or even rules. Instead they seek to encourage member countries to incorporate these values or ethics in the development of AI.

But what do we mean by AI?

AI is not one thing with a single application that poses singular risks or threats.

Instead, AI has become a blanket term to refer to a vast number of different systems. Each is typically designed to collect and process data using computing technology, adapt to change, and act rationally to achieve its objectives, ultimately without human intervention.

Narrow AI is good at a specific task, such as playing chess. General AI, the ultimate goal of some AI developers, aims to replace human intelligence in many tasks. It this idea of general AI that drives many of the fears and misconceptions that surround AI.

Ethics Guidelines are Many

Responding to these fears and a number of very real problems with narrow AI,the OECD recommendations are the latest of a number of projects and guidelines from governments and other bodies around the world that seek to instill an ethical approach to developing AI.

These include initiatives by the Institute of Electrical and Electronics Engineers, the French data protection authority, the Hong Kong Office of the Privacy Commissioner and the European Commission.

The Australian government funded CSIRO’s Data61 to develop an AI ethics framework, which is now open for public feedback, and the Australian Council of Learned Academies is yet to publish its report on the future of AI in Australia.

The Australian Human Rights Commission, together with the World Economic Forum, is also reviewing and reporting on the impact of AI on human rights.

The aim of these initiatives is to encourage or to nudge ethical development of AI. But this presupposes unethical behaviour. What is the mischief in AI?

Examples of Unethical AI

One study identified three broad potential malicious uses of AI. These target:

  • digital security (for example, through cyber-attacks)
  • physical security (for example, attacks using drones or hacking)
  • political security (for example, if AI is used for mass surveillance, persuasion and deception).

One area of concern is evolving in China, where several regions are developing a social credit system linked to mass surveillance using AI technologies

The system can identify a person breaching social norms(such as jaywalking, consorting with criminals, or misusing social media) and debit social credit points from the individual.

When a credit score is reduced, that person’s freedoms (such as the freedom to travel or borrow money) are restricted. While this is not yet a nationwide system, reports indicate this could be the ultimate aim.

Added to these deliberate misuses of AI are several unintented side effects of poorly constructed or implemented narrow AI. These include bias and discrimination and the erosion of trust.

Read the source article in Phys.org.

Bharat Forge acquires additional stake in EV startup Tork Motor

Bharat Forge has subscribed 1,895 equity shares of Tork of Rs 10 each at a premium of Rs 21,103 per share, the company said in an exchange filing.

Predictive Analytics for Mobile Apps: The Comprehensive Guide

A few years back, Gartner predicted for the year 2018, that less than 0.01 percent of consumer mobile apps will be considered a financial success.

The alarming fact has shaken up the entrepreneurs from top to bottom who remain trapped in the positive thinking web and forbidden the word ‘failure'. If you want your app to stand in the noisy app marketplace, then the secret to the app success is to know the worst things that can bring failure to your app, in advance. How can you predict which things will make the users abandon the app, draw the user's attention, or convert the users (Revenue)?

It’s a mystery that’s difficult to solve and if not solved, it can make your app remain unknown in history. Don’t fret! There is a solution to every problem.

The name of this magic bullet is predictive analytics which gives you the power to know how the target user will react to the app in detail before the actual launch.

What is predictive analytics?

Predictive analytics is a crystal ball that let you know everything that happens to the app and the actions to prevent or boost to engineer the app that delights the users in the future. It ...


Read More on Datafloq

Make proposed regulatory sandbox inclusive, fintech experts tell RBI

Digital Lenders’ Association of India (DLAI), an industry representative body with about 80 fintech companies as members, wants the regulator to consider its suggestions before the final draft is published.

In hiring drive, bots are calling the shots now

Algorithms are analysing people’s expressions and tone of voice to check for traits such as “confidence” and “happiness” during video interviews.

Indian delegation to recommend govt on policy changes around blockchain after studying Swiss framwework

The Reserve Bank of India’s ‘Draft Enabling Framework for Regulatory Sandbox’, released on April 18 had considered blockchain applications for testing.

Internet of Things – the Newest Wrinkle in Event Management

Technology has entirely changed the landscape of every business and organization and event planning and management is no exception. Days of manual registration and paper flyers have long gone and the event management industry is swiftly moving from a simple mobile app for basic event information towards augmented and virtual reality for better event management and enhanced attendees’ engagement. The Internet of Things (IoT) is one such technology that connects everyday objects to the internet.

What does IoT mean?

For starters, IoT is a broad network of Internet-connected objects; the objects are embedded with sensors that collect and exchange data creating an efficient data-driven network about the activities of the daily life.

Apart from the traditional connected wearables and gadgets, IoT lets everyday physical objects like home appliances, lights, heating/cooling systems, even doors, windows and curtains to connect to a network and generate data from them.

IoT for Event management

IoT holds the potential to transform any industry and will play a key role in revolutionizing the event planning and management industry.

Event management relies on tons of data about the venues, potential attendees, caterers, performances, attendees' check-ins, and payments. By incorporating IoT in the event management, the event planners will be able to get their hands ...


Read More on Datafloq

Monday, 3 June 2019

SoftBank's service to help power self-driving buses, farm machinery

Japan's newly launched constellation of Michibiki satellites has been offering high-precision location data since last November, and SoftBank's service will be an early attempt at building a commercial business with it.

German chipmaker Infineon to buy Cypress in $10 billion deal

The deal to create the world's No.8 chipmaker ranks as one of the biggest takeovers led by a European company this year and follows a slowdown in activity in the first quarter.

6 Ways to Boost Your E-commerce Sales with a Chatbot

Do I need to integrate chatbot on my e-commerce website? Will my business benefit from it? These are the kind of questions that we get from our clients. And our answers has always been, “YES.” Not because we are in this industry. But, because we believe in AI-bot technology.

Increasing e-commerce sales is not easy, and with this type of competition in the marketplace where there are approximately 110,000 e-commerce websites, and not even 50% of them are generating revenue of meaningful scale on the internet. So, in order to separate yourself from the down-going pack, you will have to be exceptional in terms of your products and services.

The e-commerce market is currently banging at the most recent technology – Chatbots. You want to know why?

Because chatbots are promoting an effortless buying experience for shoppers, and e-commerce websites can effectively leverage the technology to provide a more satisfying shopping experience than ever before.

There are multiple ways of using chatbots for your e-commerce solution – hereby listing the 6 smartest and efficient ones.

1. Makes Ordering easy

1 out of every 3 smartphone user uses messaging apps. But switching between the apps to place an order while chatting isn't always convenient. But now this ...


Read More on Datafloq

Sunday, 2 June 2019

Blurred Bot Lines and the Phenomenon of ‘Aiporia’

Are you a bot?

This is a question that we’ve found ourselves asking more and more of late. As we increasingly interact with AI, it’s becoming harder to tell the bot from the human.

A by-product of these blurring bot lines is the phenomenon of ‘aiporia’. 

But what is aiporia, and what does it mean for AI going forward?

Blurring bot lines

In recent years we’ve seen vast advancement in the field of artificial intelligence. It’s still got a way to go before it reaches a human level of flexibility. But, in many areas, AI is now doing a good job of pretending to be human.

Take Google Duplex, for example. Complete with vocal hesitations and minor exclamations, the phone call AI is almost indiscernible from humans.

In jobs, meanwhile, blurred bot lines mean that more responsibility than ever before is being handed over to bots. Indeed, artificial intelligence is handling all sorts of tasks, from healthcare to financial advice. 

And the lines between bot and human are only set to grow hazier. It doesn’t matter if it’s behind the scenes in tasks and jobs or upfront in customer service and daily interactions. Discerning between a bot and human activity is only getting more difficult.

These blurring bot lines ...


Read More on Datafloq

Saturday, 1 June 2019

Govt keen on tech to fast-tracking crop insurance claims

The use of tech to expedite claims has become even more important as the govt has mandated to settle insurance claims within two months of filing.