The WelderDestiny Compass

Machine Learning Vision Systems - Issue #002

Wednesday, January 11, 2017 / Perth Australia / By Niekie Jooste

In this edition of "The WelderDestiny Compass":

  • The journey to the future starts within.
  • “Self-Interest” drives the world and shapes the future.
  • Machine learning vision systems.
  • The Welder and machine learning vision systems.

The Journey to the Future Starts Within

Thanks to everyone that has subscribed to The WelderDestiny Compass in the last week. I am excited that you want to take this journey with me.

We will be looking at diverging subjects in this e-zine, but the final take-away is always aimed at challenging ourselves to form a vision of the future world of the Welder. It is only reasonable to think we can form a view of the future, if we have an appreciation for the forces that shape our world. These forces are not only technological. They are predominantly human nature in action, so we will find the drivers of the future within!

If we can form a reasonable view of this future world, then we can anticipate the ways in which we can position ourselves to take advantage of not only the final outcome, but also the process of change.

If you have an entrepreneurial nature, then please let The WelderDestiny Compass serve as a source of ideas on how you can profit from the changes coming our way. If you are interested in making sure that you position yourself to survive the "rise of the machines" within the workplace, then this e-zine will also give you much food for thought, and the website ( should give you a good place to expand your welding knowledge, equipping you to thrive in the new era. Regardless of any other motivations, if you like the intellectual pursuit of anticipating the future, then please also get involved by sending me an e-mail with your ideas, (send your e-mails to: or add a contribution directly into the comments form on the bottom of this page.

Now let's get stuck into this week’s topics...

"Self-Interest" Drives the World and Shapes the Future

Always back the horse named self-interest, son. It'll be the only one trying. (Jack Lang – Labour Premier in Australia)

It is not only politicians that know that self interest is a powerful motivator. We all know this instinctively. While there are many people that shoot themselves in the foot (both literally and metaphorically) out of foolishness or error, very few people choose to do so willingly. If you can choose between two feet to shoot, chances are that you will choose the foot that is not attached to the end of your own leg!

While this principle of "self-interest rules" is not exactly a blindingly brilliant insight, it is central to how society works, and is powerful in trying to predict how society will respond when faced with the forces associated with the technological revolution.

This becomes clearer when we understand that everyone has a "sphere of influence" within which they can promote and defend their own self-interest.

For us small fish, the sphere of influence is small, so it is not possible to get a national law passed that will result in us getting advantaged in life. Maybe just a parking spot at the local tennis club! If, however you are a high-ranking government official, or mega rich business person, then you would be able to get this law passed with some effort on your part. In between these two extremes of influence, there are many levels of influence.

While these issues of influence and self-interest often lead to corruption, please understand that this principle is not primarily about corruption. Not legal corruption in any case. We could argue that it is about moral corruption, because people are acting in the interest of "self", or "groups" rather than the "whole of society", but that leads us to grey areas that do not help us much, so we will not go down that road.

At the highest levels, this self-interest results in small interest groups, or even individuals, being advantaged, usually financially, at the expense of the majority of people in society. Obviously this kind of thing only works if the actions leading to this situation can be "sold" as being to the advantage of society.

This is the point at which we put on our "tinfoil hats" and delve into the realm of the conspiracy theorists...

There are those that describe modern democracies as just a front for the unelected "insiders" that actually govern. Terms such as "the dark state" (or the swamp) are used to describe this group. This "dark state" is described as a group of insiders that make sure that they get the best deals in life while manipulating society to pay for it. Sometimes they are portrayed as some kind of organized group that is centrally controlled. (Illuminati anybody?)

When you think about it, it is an inevitable situation in any society that high-level influencers look after their own self-interest. This happens in all systems of government. It happens in a monarchy, socialist, democratic, theocratic, free-market and communist society. The big thing to note is however that this is not some centrally controlled group, but a diverse group looking after their own self-interests. To make this work, the people that understand how influence is wielded, know that they need to work with each other as far as possible. It is a matter of "I will scratch your back if you scratch mine". (And there are some multi-billion dollar backs out there with a serious itch!)

When those at the top of this pyramid of self-interest over-reach, or interest groups within the system want to juggle for position, then you get revolutions. These revolutions can be anything from a hostile change in leadership at a local sports club or non-profit board, a company boardroom coup, hostile company take-over and a national revolution. These are just different levels of the same thing.

To further expand on this, it starts to become clear that a democracy is just a series of agreed upon revolutionary events. It allows the broader populace to “revolt” against those in power at regular intervals, in a (hopefully) peaceful manner. The really skilful wielders of influence within these systems will however position themselves in such a way that they end up with their influence intact after the “democratic revolution”. In short, those that can achieve this outcome manage to maintain influence regardless of who wins the election. This is the “dark state”.

What we notice is that the leadership following revolutions (big or small) pretty much always gravitate back to the situation where the "insiders" (those with the influence) look after themselves in the first instance. The faces of those “in charge” may be different, but the outcome in terms of self-interest only changes in form, not substance.

If we think of this in an unemotional way, we will realise that this is the "natural order of things", rather than this being an aberration. We could explain this in terms of evolutionary adaptation or the sinful nature of man, or a hundred other ways, but the bottom line is that this is human nature. It has been so for thousand of years, (the extent of recorded history) and will continue in this way for thousand more. I would not bet on human nature suddenly changing just because technology is changing more rapidly than in the past.

Obviously, people that manage to maintain these positions of influence are typically people that know how to wield such influence to their advantage. If they did not, they would not last long within that position! It is pretty much a situation of natural selection and in some cases "breeding". (Hereditary power and / or wealth.)

OK, so let us take off our tinfoil hats and ask how this insight helps us in our mission to anticipate the future? Firstly, it tells us that any technology change (or change in the system) that does not offer an advantage to some group that wields influence currently, will have a fight to make it into the "mainstream". Sometimes a technology will have to “wait for its time to come", before it will go mainstream. Secondly, it tells us that any technology that will seriously damage the interests of those with influence, will need to garner its own support base to "get ahead". This too, is nothing new. It is the good old "constructive destruction" that is actually at the heart of the free market system. A new technology may have the potential to make the current system obsolete, but it will only gain traction if "managed" in the right way. Let us look at a couple of examples to get an idea:

In the "ride hailing" (e.g. UBER) VS Taxi industry battle, we have the entrenched taxi industry insiders with much to lose, so they will fight this battle rather fiercely. In the "home accommodation" (e.g. AirBnB) VS hotel industry battle, we have something very similar happening as with UBER, so again we will have a big battle. In both of these situations, the new technology companies have garnered the support of the "voters" to stand behind them in their fight. I use the term “voters”, because both UBER and AirBnB have a broad and enthusiastic user base. This broad user base vote for who gets into public office. While the politicians that are the front for the taxi and hotel industries are battling these two technology companies, they must be careful not to alienate the very large "voter" support base for these services. I believe that because of this dynamic, the technology companies will prevail, but there may be some "agreements" and "concessions" made along the way to soften the blow to the traditional suppliers.

The "streaming video" (e.g. Netflix) VS physical video/DVD rental war was very short with hardly a shot fired. While this technology killed a number of physical movie rental companies, the real power behind the entertainment industry did not mind the change. The real power behind the entertainment industry are the content creators, rather than the content delivery companies. Streaming content delivery was not to their detriment, so no big fight.

The new era of automation will be full of this constructive destruction paradigm with winners and losers aplenty. It will be part of the opportunities that we create, as well as the threats that we face. Are you game?

Machine Learning Vision Systems

Our previous version of The WelderDestiny Compass looked at artificial intelligence. This week's technology topic of "machine learning vision systems" is very much connected to artificial intelligence, although this week's topic is a bit more focused regarding its application.

To date, one of the biggest shortcomings of computer based systems have been their limited ability to "see" and use that information to learn for the future. This is one of the greatest human strengths, and if you think about it, this actually takes a lot of computational power to achieve. In fact, from this perspective, the human brain is an amazing computational device. (Okay, for the animal lovers, let me say it... The animal brain in general is an amazing computation device.)

In the last while, great strides have been made in vision systems. This is evidenced by facial recognition technology, (been through customs at the big international airports lately?) and the technology used to identify cars speeding along the freeways of the world. IBM also offers a service, based on their Watson AI platform, to categorize and interpret images for clients. While these systems are good at matching what they "see" to data in a database, they are not systems that “learn” from what they see in a "real time" sense.

There is however a new technology on the market that is based on a neural network computer chip, that is very good at not only the vision part, but also learning from what is seen. It can identify patterns and behaviours from a video feed. The company developing this technology sells it to the casino industry. They use this technology to keep an eye on the card tables (e.g. black jack) to see if the dealers are "making mistakes". (mistakes, yeah, wink wink) These systems will learn the rules from looking at how the dealers work, and will then be able to detect if anybody "makes a mistake". If you think about this, you start to see how revolutionary this is. Not only can the system discern between different cards, but it can also learn the rules of the game, and discern if the card is being drawn from the right place in the deck.

This technology obviously has huge application in any sort of surveillance scenario, so beware, big brother is watching! Not only will it be able to recognise faces, but it can learn over time what it looks like when somebody is acting in a suspicious manner, and then alert the necessary authorities to investigate. Our interest here in The WelderDestiny Compass is however concerned with applications revolving around how this technology can monitor industrial processes such as welding and the potential applications in inspection of the results.

Can you think of how this technology can influence you or your job?

The Welder and Machine Learning Vision Systems

Current automated welding systems tend to still need quite a bit of "hand holding" from humans. They pretty much just "follow a recipe". If there are even minor fluctuations in the material properties or welding fit-ups, then there tends to be a disaster, requiring human intervention or scrapping of the component. If the machine learning vision system is incorporated into a robotic welding cell, then the robot will be much better at performing welds under conditions where greater variation exists.

This application within welding automation serves to be more of a driver to reduce shop based work for the Welder, so it will not directly impact the Welder out in the field.

In our last instalment of this e-zine we started to get a picture in our minds of how artificial intelligence (AI) can help a Welder out in the field to make better decisions and offer knowledge based support. If we now imagine that within the welding helmet there could be a couple of cameras that are linked to this AI, which incorporates the machine learning vision system, then several interesting things become possible. But first, let's consider what a Welder does during the welding operation...

Under ideal circumstances, a good Welder can mostly see when defects are being incorporated / generated in a weld. A good Welder will, if possible, stop welding at that point and do a spot of grinding to remove the defect before progressing further. In addition, a good welder can see (based on the clues within the weld pool) if the danger of "burning through" is being approached, so s/he will know to either stop and adjust the power source settings, or modify his/her technique to reduce the probability of this burn through.

In the same way, over time, the machine learning vision system will learn how you as a Welder welds, and what it looks like in the weld pool when things start to go wrong. Based on this, it can give you direct feed-back regarding weld pool conditions, and also act as an inspection record regarding the quality of the weld. In many circumstances, this type of weld monitoring may be able to replace, or significantly reduce, post welding non destructive testing (NDT).

If this technology makes its way out into the field, there will obviously be some creative destruction taking place, with NDT and inspection companies losing work and vision technology companies gaining market share. This has the potential for a lot of fights of “self-interest.” The battle field of this fight will be that of fabrication codes and industry regulation.

In this war, will you be in the firing line, a spectator, or winning the battle?

Yours in welding

Niekie Jooste

Return Navigation:

› Issue #002

The WelderDestiny Compass: Weekly e-zine Subscription

Do You Have Thoughts About This Week's E-Zine?

Now is your opportunity to contribute to the topics in this week's The WelderDestiny Compass. If you have thoughts or examples that you would like to share with other readers of the e-zine, then please contribute by entering the title of your contribution in the box below. Feel free to make a brief or more expansive contribution to our discussion...

Your Ideas on Machine Learning Vision Systems and Self-Interest

What are your thoughts on self-interest and the "dark state". Is this just conspiracy theory, or is this real? / Can you see applications where machine learning vision systems can have a significant impact? / Please share your stories, insights and even fears or wishes, regarding machine learning vision systems.