Automation of Everything

[Arthur Hanson]

The automation of everything is coming as I have written about in the past. It's the last part of the big data/communications/processing power/mems/automation/robotics chain. I feel medical will be the next huge driver in this chain as wearable devices complete the last link in the chain. I found this "Humans Need Not Apply" a thought provoking video that explores the ramifications of this change that's here and very well on its way. This will present unparalleled opportunities in the semiconductor world that is already populated with automation in almost all phase from robotics to EDA. Comments, ideas and thoughts, solutions welcome and actively solicited. We are rapidly entering the world of cutting edge science fiction becoming reality. Will we even recognize the future and will we be able to adapt to the ever accelerating progress of technology?

https://www.youtube.com/watch?v=7Pq-S557XQU

 

[Julian Grodzicky]

I believe it will ultimately lead to a dystopia of extremely rich, and all the rest. At the end of the day, if a robot or AI has replaced the majority of people in their jobs, how do those people generate any sort of income to get the necessities or wants of life? To buy the products and goods that are being produced by the robot-staffed businesses? Start a business, you might say? With no job, trade or profession to generate income or savings, how does one generate seed capital for a business? The scenario of the movie 'Elysium' springs to mind... Even the future equivalents of Rockefeller or Carnegie would find it impossible to make their fortunes, because they started with nothing and worked their way up.

 

[Daniel Payne]

When I first started doing transistor-level circuit design at Intel in 1978 most of my tasks as a chip designer were quite manual and labor intensive. Today, with the help of EDA software I can do the same job much easier, therefore making me more efficient. EDA software didn't eliminate the job of circuit design, it enabled it to be more productive and accomplish more with less mental effort in a shorter period of time.

I do agree with the general sentiment that clever software apps can certainly replace many white-collar jobs, even Charles Schwab claims that their software can replace the job of a financial advisor.

 

[Arthur Hanson]

Daniel, A while back I posted a link about a robotic researcher for pharmaceuticals that actually made a drug discovery. We can only imagine how long before much more R & D is being done by AI coupled with robotics. I increasingly see where larger and larger portion of everything is done by machines. In many areas it is already here and progressing fast where the number of engineers and other workers is being reduced at an accelerating rate. This is why education needs to go to a few basic courses then to a lifetime subscription to keep up with the acceleration of technology. New fields of work and study we haven't even dreamed of yet will come into being creating even more opportunities, but the future will be dim if the current slow adapting educational system doesn't start to keep up.

 

[Daniel Payne]

Arthur,

I can give you one example of AI in an EDA took, it was at Silicon Compilers in the 1990's and our founder David Johansen created a logic synthesis tool with learning capabilities, so each time that you ran the tool it became a little bit smarter. The unfortunate side effect was that we couldn't reproduce the results when someone in San Jose ran the tool on a design, versus the same design being run through another copy of the tool in Irvine, because each copy of the AI-based logic synthesis tool would produce results based upon its own, unique experience. Our QA department noted that this type of a tool was not supportable because we couldn't reproduce identical results. We possibly could've made it work had their been a mechanism to export the experience portion of the customer tool, then import that into the QA version of the tool.

 

[Arthur Hanson]

Daniel, Thank you for the input and bringing up a very interesting point. The point that it brought up different results actually has a good point in that it shows what several teams of engineers attacking a challenge come up with, several different solutions. My father always told me if you put a good size group of good engineers in a room and told them to reach an ideal solution there would be blood on the floor over competing solutions. It looks like running a master program in the cloud that evolves would be the best answer. This could also give low cost access to super computer power and speed besides technical support leveraging solutions, questions and upgrades on one program, on one system. This is the kind of discourse where SemiWiki shines.

Right after posting this I ran across this article in MIT TR

Will Advances in Technology Create a Jobless Future? | MIT Technology Review

 

[Staf_Verhaegen]

I do agree with the general sentiment that clever software apps can certainly replace many white-collar jobs, even Charles Schwab claims that their software can replace the job of a financial advisor.

In the last century or two jobs shifted from agriculture to industry to white-collar. Problem now seen is that there seems no new type of job for replacing the white-collar jobs that will be automated. Some people say it will be replaced by other types of white-collar jobs but I'm not so optimistic.

But in the end it could be good for mankind if we can get away from the notion that the only way to earn money is through a job. If production of all the basic needs is fully automated one could also distribute it to all people (the so called basic income). If the wealth generated from these machines is only going to the owners then we will end up with a society with a few happy rich and the rest poor but not too poor so they can keep the rich rich; collateral damage of a few percent starving to death or lacking medical support is not a problem.
If we use automation to provide everybody a liveable live everybody could in the end do the work - paid or unpaid - which he really loves to do.

 

[Daniel Payne]

Staf,

The people of Alaska are paid an annual amount from the massive oil industry business called the Permanent Fund Dividend (PFD), so it may also make sense for robotics/AI companies to distribute their wealth in a similar manner.

 

[ippisl]

Daniel. if we're talking about this automated EDA, there's this interesting tool, which takes a math description of a dsp algorithm and creates an optimized code fitting a specific platform.

SpiralGen

SPIRAL Project: Home Page


What's nice about their method - is that the algorithm also works at the math level - i.e. doing various mathematical transforms , in order to find the optimal mathematical representation of the problem.

Another tool i've read research about is something equivalent to HLS for mechanical engineers. And considering mechanical engineers work at a relatively low level(below the synthesizer) - it could have a big impact on the field.

 

[Arthur Hanson]

Maximizing the human capital should be the main goal of any country without discrimination or prejudice. Holland, Singapore and Germany are the best examples of using education to maximize human potential. The semi industry has a broader skill set and broader product sect with the greatest penetration of any industry. There is hardly an area in anything in which semis don't play a critical part. There is not a single physical problem we can't solve. It will be up to social media and social programs to solve the most difficult problems of human nature and interaction and the semi industry will provide the foundation on which people figure out the solution to this.

 

[Daniel Payne]

Daniel. if we're talking about this automated EDA, there's this interesting tool, which takes a math description of a dsp algorithm and creates an optimized code fitting a specific platform.

SpiralGen

SPIRAL Project: Home Page


What's nice about their method - is that the algorithm also works at the math level - i.e. doing various mathematical transforms , in order to find the optimal mathematical representation of the problem.

Another tool i've read research about is something equivalent to HLS for mechanical engineers. And considering mechanical engineers work at a relatively low level(below the synthesizer) - it could have a big impact on the field.

Thanks for sharing the links on SPIRAL and SpiralGen, I hadn't heard about them before and they look very promising.

 

[msporer]

I believe it will ultimately lead to a dystopia of extremely rich, and all the rest.

What do you mean 'lead to'? We are already there.

 

[simguru]

What do you mean 'lead to'? We are already there.

When it comes to making money it doesn't look to me like the intrinsically hard/worthwhile jobs are the ones that pay well, i.e. folks doing bean-counting/finance make a lot of money, but those jobs are probably the easiest to automate. Seems likely that the commercial sector will try to automate most of the jobs away, and your current 1% becomes a 0.1%, and it will be interesting to see what happens in the world's democracies under those conditions (Brave New World?).

IMO there has not been a lot of automation yet because we don't have much AI technology to go with our robotics, and that's probably down to using the same (bad) computing architectures for the last few decades. New architectures that can learn for themselves and don't depend on human programmers will take over a lot of jobs fairly quickly.

 

[Arthur Hanson]

When it comes to making money it doesn't look to me like the intrinsically hard/worthwhile jobs are the ones that pay well, i.e. folks doing bean-counting/finance make a lot of money, but those jobs are probably the easiest to automate. Seems likely that the commercial sector will try to automate most of the jobs away, and your current 1% becomes a 0.1%, and it will be interesting to see what happens in the world's democracies under those conditions (Brave New World?).

IMO there has not been a lot of automation yet because we don't have much AI technology to go with our robotics, and that's probably down to using the same (bad) computing architectures for the last few decades. New architectures that can learn for themselves and don't depend on human programmers will take over a lot of jobs fairly quickly.

Basic finance may be simple and much of it is already highly automated, but real, accurate financial analasys is among the most difficult skill set to master and very few actually do. The ones that make the tremendous money have developed their own sytems and methods that are as sophicticated as any field of engineering. The mathematics alone in many cases are world class, running on world class networks. Some programs are highly automated, but require constant high level management. These are not simple programs or processes and some require literally hundreds of highly skilled people and some thousands. It's medical that should be much more automated to sharply reduce costs, but like many fields, they have bought and paid for political cover. Khosla feels in many areas doctors could be replaced with big data and some level of AI and this is but one example. Pharmacists should have been automated long ago at tremendous cost savings. I have just posted a question on what company is developing an automated micro chip verification system.

 

[simguru]

Basic finance may be simple and much of it is already highly automated, but real, accurate financial analasys is among the most difficult skill set to master and very few actually do. The ones that make the tremendous money have developed their own sytems and methods that are as sophicticated as any field of engineering. The mathematics alone in many cases are world class, running on world class networks. Some programs are highly automated, but require constant high level management. These are not simple programs or processes and some require literally hundreds of highly skilled people and some thousands. It's medical that should be much more automated to sharply reduce costs, but like many fields, they have bought and paid for political cover. Khosla feels in many areas doctors could be replaced with big data and some level of AI and this is but one example. Pharmacists should have been automated long ago at tremendous cost savings. I have just posted a question on what company is developing an automated micro chip verification system.

A while ago on NPR they were discussing that the deep-learning algorithms could dig up equations for how complicated things (like cell chemistry) work, but that humans couldn't understand the equations, so I still think the financial analysts are on their way out ;-)

https://www.kaggle.com/ - reported that currently the most successful teams are a mix of AI and engineers/physicists. I'm not sure big teams are required any more.

The pressure to automate medical is probably higher outside the US where there is a shortage of expertise and/or the state carries the cost. Robots are doing surgery now, but still under human control (that might not last long).

EDA is due an overhaul, the stuff should be correct-by-construction, but designing at the RTL level makes that difficult. I expect a new generation of tools aimed at getting software on FPGAs/many-core will become the base of future EDA. High variability Silicon (sub 45nm) is also motivation to switch to asynchronous methodologies. However, that needs a better software approach too - possibly a new (easy to use) parallel programming language. Catch 22 - the AIs might be able to do it, but we haven't built them yet ;-)

 

[Arthur Hanson]

simguru, The level of skill and intuition for finances is far higher for finance for human and societal emotion and trends come into play with another complicated layer of politics across many diverse, ever changing cultures when chip design and much science involves a much simpler structure although complex and large it is still largely linear with far fewer variables used in the semi/software industry. Long range world wide weather forecasting and creating life are the only areas of science that rival economic and financial forecasting in the number of variables and what is even knowable. I do agree that many financial people at low skill levels can and will be put out of work as the people that master the system will continue to concentrate wealth and in this small area the number of people is increasing dramatically, but from a microscopic base for so few have mastered it, even with vastly increased computational, communication, collaboration and information gathering resources. Math applied to human nature and the world is in its very, very early stages and compares to the Wright Brother's plane compared to the latest jet fighters or airliners.

 

[simguru]

simguru, The level of skill and intuition for finances is far higher ...

I think you missed the point that humans have a limited capability to mine the data and come up with an "intuitive" sense of how things work, an AI can handle more data faster and come up with algorithms humans can't understand. Very few people can beat a computer at chess these days, I don't see finance being a harder problem.

 

[Arthur Hanson]

simguru, Even with the best programs, they are only as good as the person directing them. The financial world is the most complex of all for it incorporates more variables than any area or endeavor and is about percentage wins, never absolutes as in science and even to a large extent design. Except for high speed trading, programs can only supplement, not replace human intuition. This I feel will be true for a long time, due to the almost infinite number of variables. Even for the winners, it's a form of gambling, for even the best of the best have losing years. Loss management in a infinitely variable environment will require human input, although it can be supplemented by raw computing power.

 

[msporer]

IMO there has not been a lot of automation yet because we don't have much AI technology to go with our robotics, and that's probably down to using the same (bad) computing architectures for the last few decades. New architectures that can learn for themselves and don't depend on human programmers will take over a lot of jobs fairly quickly.

You give too much credit to the intent to improve what we have. The reason why we continue to use 'bad' architectures is because they can be manipulated in order to maintain the imbalance of of income/power/control. ie. Those with easy jobs that pay well don't want to see things change.

Likewise our global financial system is intentionally complex for the same reason; it can be manipulated by the few and those influences can be hidden.

 

[Arthur Hanson]

msporer, I agree with you completely. Many people are using government regulation and outright corruption to protect the work cultures that are static and don't want to change and as a result make everyone else pay for their little static world, while the people they make pay are subject to rapid change and competition. They have literally become a class of royalty both inside and outside the government. In many ways, the job of government has been to actually hold things back for the benefit of the few. This won't change until we have transparency in finance, I think every single government transaction should be posted online except for TRUE NATIONAL SECURITY. All campaign contributions should be listed online in real time for anyone to access. The government does much valuable research, except at many times the cost of the private sector in the vast majority of cases. AI is coming and those that don't embrace and utilize it will be run over by it.