Drishti is a company that is helping to analyze the efficiency of actions performed in factories. Drishti uses computer vision to digitize manual human actions. They are able to analyze video and not just object recognition in a single picture. They are experts in factory automation and operations. Brian Wang of Nextbigfuture has interviewed Drishti CEO Prasad Akella.
People Still Do Most of the Work in Factories
People assume that robots do most of the work in manufacturing and factories and they assume that there will be a wave of AI, automation and robotics where they will lose their job. However, Drishti CEO Prasad Akella reveals a different situation.
The real numbers are that 72% of tasks in a factory are not automated. There are almost no robots involved in making smartphones like the iPhone. Foxconn builds all of the iPhones and iPads and their CEO stated the goal of using mostly robots to build smartphones and tablets but they massively missed their goals of a million robots. In 2011, they said they would have a million robots by 2016 but instead they had 40,000 robots.
Automobile factories only have heavy use of robots in the paint shop and body shop but general assembly is where most humans are still working. There are 300-400 robots in the most automated car factory production line and thousands of people.
Overall there are 1.7 million robots working alongside the world’s 345 million factory workers, according to estimates from the International Federation of Robotics and Goldman Sachs.
Finally Going Beyond the Analysis of Human Work that Henry Ford Used 100 Years Ago
Production and performance data is readily available on-demand from robots on the manufacturing floor. However, the methodology for collecting human statistics is the time and motion studies from 100 years ago used by analysts working for Henry Ford.
Using AI to analyze information from video cameras there can be insight and improvement of 72% of what is done in factories. 68% of production defects are caused by human error. Better data, Akella said, could increase productivity, improve what is considered a reasonable standard of production quality, and prevent small errors from leading to large-scale product recalls.
Getting Granular and Systematic About Human Processes
Currently, analysts take apart their own and competitor’s cars to analyze the final quality and find ways to improve how cars are built.
Sandy Munro is famous for his teardown analysis of the Tesla Model 3.
Drishti is offering teardowns and constant monitoring and analysis of the processes in factories.
Akella is an expert in collaborative robots (cobots). Akella led the General Motors team that created them in the 1990s and he founded Drishti in 2016 to use A.I. and data analytics in coordination with factory workers.
Action Recognition
Drishti has built a new deep learning architecture that uniquely parses streams of video to generate real-time analytics on highly variable human actions.
Put simply, Drishti can recognize activities being performed as the operator performs them, even if Drishti has never observed that operator before.
Drishti is the only company in the world commercializing action recognition specifically for manufacturing, and one of the very few working with the technology in any industry.
Action recognition is just part of the solution. Drishti’s vertically integrated manufacturing solution extends end-to-end: from capturing video to creating data to providing applications, data services and workflow support for line operators, supervisors, engineers and more.
Factory operators can get started with Drishti without disrupting your assembly lines or interfering with your existing processes. There are no heavy IT integrations and very little equipment to install. As long as the cameras have a clear view of a process, and as long as the cameras can stream data, then you can deploy Drishti.
Drishti Status
Drishti has already raised a successful series A funding round and is working with major customers who are global leaders in manufacturing.
Potential for Trillions of Dollars in Global Impact
Manufacturing is $12 trillion of the $80 trillion global economy. Discrete Manufacturing is $7.2 trillion (60% of $12T). If the mostly manual processes can be made 20% more efficient then this would be more than a $1 trillion boost to the world economy.
There is also process manufacturing. Process manufacturing follows a sequential model and sometimes creates materials, such as steel, used in finished products.
Discrete manufacturing follows an asynchronous model and constructs a finished product, like a smartphone.
Why Will Many Jobs Remain Safe from Automation?
The dexterity needed for fine assembly work is still vastly superior in people. Robots cannot deal with floppy wires. Elon Musk and Tesla want to get to wireless communication within car engines and around a car. They will make great leaps reducing the wiring but getting to zero wires is still a huge leap. The wireless transmission of high-levels of power is an issue. These solutions are not about making the robots better but designing out areas that are hard for robots to handle. This only makes sense of the new design is better in terms of performance and features.
Humans are also faster and cheaper to adapt to change. Factories and products have a lot of change and variation. There is also change and variation in service and other work.
SOURCES- Interview with Prasad Akella, Drishti
Written By Brian Wang, Nextbigfuture.com
Brian Wang is a Futurist Thought Leader and a popular Science blogger with 1 million readers per month. His blog Nextbigfuture.com is ranked #1 Science News Blog. It covers many disruptive technology and trends including Space, Robotics, Artificial Intelligence, Medicine, Anti-aging Biotechnology, and Nanotechnology.
Known for identifying cutting edge technologies, he is currently a Co-Founder of a startup and fundraiser for high potential early-stage companies. He is the Head of Research for Allocations for deep technology investments and an Angel Investor at Space Angels.
A frequent speaker at corporations, he has been a TEDx speaker, a Singularity University speaker and guest at numerous interviews for radio and podcasts. He is open to public speaking and advising engagements.
The Internet has no arms but is the largest piece of automation in the world.
There will be lawyers, auditors and everyone who takes the rule for laws and regulations that will be replaced.
At last, bureaucracy will be automated, fast-paced and much more difficult to corrupt.
To a lot of researchers who collect data for compiling in a report, hospital staff in image analysis, laboratory work and easier as transport and cleaning.
Actors, shop and warehouse workers, photo models and journals and music creators are more groups that get to find other chores.
I almost forgot.
Teacher will be replaced with AI train us according to our desires 24/7 and with better knowledge of us than our friends or family.
Musk, in his failed attempt to make his “alien dreadnought” highly automated manufacturing line, commented that a particular problem is all the parts were functionally designed for human assembly. The fine dexterity requirements, lack of obvious handling/gripping points, and event the slipperyness of the parts have in a way been fined tuned to the use of human hands for assembly. Rethinking part design for robot manipulators means you don’t have the design history and economics of current parts manufacturing to help you, driving up proximate costs for robot friendly components.
all factory workers can be replaced… we just need to get used to wearing 3D printed clothing… sure its looks weird, but it’s 100% functional and can replace an entire factory of woman running sewing machines… just think a world of people wearing breathable 3D printed plastic clothes and crocs shoes… 100% automation is achievable if people were forced into it…
I disagree with the idea that robots can’t easily handle floppy stuff. Roboticists are trying to find clever solutions (vision processing and more dextrous robotic hands, etc) to the sort of things that old-style automation experts solved with brute force.
E.g. add more (but simpler) robot arms/grippers, to hold floppy stuff under sufficient tension to make it behave sufficiently predictably at all stages of manipulation.
Make a jig of fixed-position rods with grippers at key points to hold wires in position after a general purpose robot arm moves a wire into the gripper. Hold complex floppy parts rigid that way as you assemble them (e.g. a wire harness). Keep holding them that way until they are moved and attached in place (e.g. on a car).
Similar principles could be used to let robots fold laundry in an algorithmic fashion, using one general arm and a number of extra grippers on un-powered rods that the robotic arm can unlock, move around, then lock in place. Or maybe just a grid of clothes line and clothes pins, with 1 robotic arm to grab and move the cloth, and another to apply clothes pins. Once the robot has securely spread out an item of clothing it can identify it, grab it accurately and fold it in a programmed fashion.
I’m not saying this approach would be trivial – but at least it makes ‘floppy tasks’ possible without requiring AI breakthroughs and super-dextrous robots.
So if I understand Drishti’s point, it’s that his technology is what is going to cost 20% of factory workers their jobs in the near future, by letting companies discover ways to make their human workers 20% more efficient?
So this is directed to people that haven’t already lost their jobs to robots?
Oh ok.
Part of the problem with robots is they have no sense of touch. It would be easier and better to do that in a soft surface than a hard one.
I disagree that rigid robots can’t deal with non-rigid parts. If the robot has good enough vision systems, I’ve seen them deal with flexible wires and such with no problem. Robotic arms can already do just about any task a human can do in manufacturing or assembly. It’s mostly a matter of cost, scale, and flexibility.
Got that right. I’d actually prefer a car with exposed fasteners, that steampunk look appeals to me.
But the bottom line is, the robots are improving, and the humans aren’t. Unless that changes, eventually the robots surpass the humans.
I’m not terribly afraid of losing my job to a robot; I’m a tooling engineer a few years from retirement. Barring an AI singularity tomorrow, I’m safe. But, assembly jobs? Yeah, those are eventually going to get automated.
Just a while after the production jobs.
a half dozen such machines. A more skilled “setup man” changes the chuck, or chuck jaws, tooling, and programming when the part to be manufactured is to be changed.
Much of the reason it is so hard to automate the manufacture of consumer goods, is that consumers have been programmed to believe the machines they bThe uy should not look like machines, and that appearance is more important than function. Maybe it’s the engineer in me, but I think it’s silly that in automobiles, all actual machine parts are hidden, even in the engine compartment, everything has plastic covers hiding it.
Look around the interior of your car. You can’t see a screw, nut, or bolt anywhere. Everything is snapped, or glued together, or assembled in such a way your head has to be on the floorboard to do so. This “design aesthetic” makes the interior assembly nearly impossible to automate, time consuming for humans to assemble, and much more expensive to repair. For instance, replacing an instrument in the dash of a car costs many times more than the cost of the instrument, because it takes so long to disassemble the dash, and reassemble it. Even then, the dash often isn’t right, because aged, brittle plastic tabs that hold it together break, or the special built tools used to assemble it at the factory are not available.
Give me a machine that looks like a machine, and simplify my life, please!
The examples spoken of here are for the most part the final assembly of complex consumer goods. Factories that serve this function are a tiny percentage of the manufacturing industry.
As the decades have progressed, the thing that is called an automobile factory has changed dramatically. It used to be that commodity steel, spools of wire, bearings, polymer to be injection molded, and other commodity parts came into the plant, and finished autos left the plant. Most of the parts, and all of the major assemblies were made, or assembled in the same plant that the final assembly took place at.
The manufacture of parts is highly automated. In the context of this article, “robots” is taken to mean the multi axis machines, built by companies like Fanuc, that have replaced human labor in paint shops.
In reality, the CNC machine tools that produce almost all close tolerance, high, and medium volume production machined parts are robots. CNC lathes employing live tooling often are deployed in such a way that dozens of two meter lengths of bar stock can be loaded into a hopper, and finished parts machined on every side, with drilled and tapped holes off center, and with radial orientation, with slots, and flats milled on the ID, and OD of the part drop into another hopper. What used to be done on multiple manual lathes, and manual milling machines is all done on one lathe. An operator with a high school education, and the ability to do precise measurements can operate
You two are talking about different time frames. Eventually, what you’re saying will likely happen. But first we need to get from here to there, which is a gradual process that will take some time. And during that process, there’s money to be made, such as as Vr Darshana is suggesting.
sure It won’t replace your factory job… that’s a bit like a fortune teller predicting that in the future you will need to pee in a bathroom. It just replaces some chinese Guy’s factory job… he completely forgets about robot dude delivering mail and packages without humans. trucks driving themselves Everywhere with out a driver… taxis and buses driving themselves everywhere with a human driver… etc..
If anyone with venture money reading this, then there is an idea for you:
https://drive.google.com/file/d/1GSv89tiQmPDcnFEu4n4CqfaJcUJxVmL5KrSCJ047g4o/edit
Instead of NN or ELGs you can use a learning Mealy machine to store already known observation-action-reward tuples, which are context dependent. By action I imply some logic transform. Reward depends on how beneficial that transform is. And then, while compressing the automaton, transform its transition and output functions with hypothesis logic which assumes what unknown tuples (“Don’t care” values) must be equal to. That Mealy machine based agent may simplify the simulation of low-detail VR engine in a brute-force algorithm easily. At least you may start with simple tasks and then as the agent learns how to simplify the extensive search for a machine code of your robot you may add complexity. But I agree, some tasks like 3D reconstruction are better programmed by humans and not by that extensive search…
Soft robotics plus AI will start gobbling up an increasing share of factory assembly jobs in the the next decade or so. Of coure rigid robots cannot deal with non-rigid parts. There are thousands of motions combined with unparalleled human senses and flexible minds that robots don’t have. A radical new kind of robot, really, an android though not necessarily human-looking, will be required. Today’s hard, bolted-to-the-floor, limited sight and zero other senses robots will seem like quaint toys by 2030…or not. If robot makers keep trying to refine the current paradigm, they will be increasingly frustrated.
The robotics field is as ripe for creative destruction as the things it automates, and is actually decades, or even centuries, old in concepts.
I think this ignores that when people say robots what they are really saying is automation. Automation includes software, web portals etc that also displace jobs. Factories and warehouses can be automated without the typical 6 axis robot arms. Lots of the automation is actually just custom designed mechanisms and things like CNC machines etc. So while he is technically right.. it might not be a 6 axis robot that takes your job… its likely that automation will eventually take your job in some form.
That to is nonsense. Once robots do ALL the work there will be no money. No rich, no poor. no haves, no havenots.
Money is barter, my time for your time. No time, no work, no money.
The huge mistake about AI that also this blogger makes is that it can copy human behavior and decision making methology, but since AI is not consciousness it exclude decision making and creation that can come from understanding so all activities that include these traits are safe and will continue to grow which can be quite a lot!
well.. of course the ironic thing is that the data that is collected by drishti will be very helpful for the ML models which are going to be training the next generation of robots.
In fact, I’d say this is where lots of their profit will come from – selling this data to automation companies.
It seems like this article is mostly promotion of Drishti which has a market incentive to hedge a bet against automation potential as it’s a sort of an on-the-ground (work-)data aggregator for manned factories. At best it promotes labor augmentation, so no surprise on the non-concern from Akella.
not sure ‘this is bs’ is a cogent argument. i’d argue that the problems with automation the article recognizes are very temporary and are simple technical challenges to overcome – and that the large number of jobs lost immediately is not going to come from manufacturing (where there is a saturation of automation to a certain point, at least until the fine dexterity problem is overcome).
Instead, the loss of jobs is going to come from paralegals. drivers. retail. radiologists. construction. these jobs are highly automatable, and this automation will only speed up when we get better at it.
That is BS! Robots will take all the jobs, period. It’s just a matter of time and not to much time at that.