What is Artificial Intelligence (AI)?

Many people are confused by the term “Artificial Intelligence” or “AI”. Touted as part of Industry 4.0, yet it is also linked with the threat of using automation instead of human workers. AI is not hardware but in fact software and not the end to civilization as often portrayed by Hollywood. What exactly is AI, and what does artificial intelligence do? Let’s look at the fundamentals of AI and put them into perspective, enabling us as an industry to accept the opportunities that so-called AI embodies, without being caught up, by the hype with artificial intelligence software examples.

What does AI mean to all of us?

People have two very different viewpoints of AI. The first is that futuristic space-age machines would invade the world as depicted in Hollywood movies. The important point of the story is not actually the machines that cause the damage, but instead the software that controls them, where a sophisticated algorithm became able to think for itself in a manner that was not planned. It is this ability to be able to think, without the coding of those types of thought processes, that is the definition of true intelligence. With a few exceptions, humans are already considered intelligent. However, machines are only directed by their software, which is currently limited to reacting to situations that occur and are conveyed to them. They follow procedures, created to achieve a preferred outcome. It is claimed that general human behavior can also be modeled, although by a much more hi-tech and advanced process. Attaining the slightest the glimmer of intelligence, could merely be an additional manner in which computer software algorithms operate, which would have to happen if machines were ever to think for themselves. This lets many people argue that there is “intelligence” in our existing assembly machines and connected software. Such statements are intended to make customers believe that the latest products are much smarter than before, generating more value and performance, which in most instances, is an exact portrayal of new machines and robots emerging onto the marketplace. Increasing intelligence each step of the way is great, but how soon before we reach a truly artificially intelligent process that can think on its own?  The endless investment in Artificial Intelligence is funded from sales of constantly improved machines and software. With things progressing so quickly, the question is, when is the most advantageous time to invest in any “Smart” solutions, when the next best thing may be approaching. If an investment is made into cutting edge Smart technology today, would it become outdated in the very near future?

Absolutely Not.

And the explanation is that there is an increasingly apparent distinction between the hardware and software components of Smart solutions.  Since most tools we buy for assembly manufacturing are bundled with both hardware and software the difference has been hard to see. To improve hardware, it takes a long time due to physical product development cycle limitations, no matter if it’s mechanical or electronic. However, improving software can be done significantly faster. We have all encountered frequent software updates that modify and typically enhance the operation of the hardware device on either our Smartphones, computers, SMT placement or inspection machines. Don’t assume that the machine provider missed the chance to add value from the beginning, or that there was an issue with the original machine or software, those days are gone. As software is not a tangible object, it progresses more as a flow instead of gradual, sizable variations of hardware, so new benefits can be offered immediately. This profoundly alters the way in which we spend money on software.

Even Microsoft no longer releases new versions of Windows but instead constantly delivers incremental updates to what seems to be Windows 10. If every modification to Windows was a major release, we would most likely be on Windows 42 by now, with each version causing fears and costing money. As an alternative, enhancements are pushed out slowly, with little effect on usage. Hardware progresses slowly, seen as products developing in separate steps, unlike software that is far more flexible, and also, delivers greater probability for becoming smarter, faster, all for a one-time investment, or a subscription that lets us select exactly what software we wish to use as time goes on. Nowadays this is very important as we move more quickly towards AI.

The Origin of Digital Intelligence

The conversation of whether software leads to hardware or vice versa is more thought-provoking now that we see software development as a flow as compared to changes in hardware that are done a step at a time. The enhancement of software has however started to separate from the hardware when it comes to AI applications and Smart operations.  The development of the Smart warehouse is a fascinating and historical instance of this. The average original warehouse model, without software, was to sort locations by part number, with some type of container designated for each. These were put in alphabetical order so that people could easily locate materials in order to make kits. Sadly, this has always been an inefficient usage of warehouse space. Holes must be physically created or closed up with the introduction of new part numbers and other becoming outdated. Lots of bins would spill over or go unused for long periods of time as needs varied over time. Moving and switching the sizes of the containers is a huge physical effort, making it tempting to just work around the exceptions, depending on human flexibility to manage.

The first unsophisticated warehouse management software came to save the day, as it was recognized that computers were usually much more proficient than humans at knowing where materials were kept, to the point that grouping alphabetically was not needed anymore. Materials could be stored in any place, a standard called “random storage”, normally into any container that is not being utilized. The chosen spot would be logged against the part number and quantity, for software to later notify material operators when the materials were required. With hardware, this initiated the policy change in the advancement of automated logistics, with things like automated guided vehicles and cranes. Easily locating materials, and the implementation of standard bins or racks, made automation possible. Hardware has progressively grown ever since, to the cutting-edge warehouses of the largest on-line retailers we now see. This hardware now does the work that many people did previously. This is machine-driven automation founded on standardization, essentially Industry 3.0.  It is correct nonetheless that the mechanical systems within such machines have become more intelligent, based once more on hardware development, specifically sensors imitating our own innate senses, and connectivity, giving the machines the ability to make a larger array of automated decisions. These machines are not necessarily Smart until AI software is added.

The Digital “Sixth Sense”

The development of Smart software is reliant upon comprehensible data that is accessible. Humans have a large quantity of data input to constantly sort through, with our already evolved five senses. A sixth sense may possibly be required to be synthetically introduced if one day, people desire to become connected, for instance to the internet directly, unnecessarily requiring the need to see, hear, speak or type. in contrast to people, machines these days are obtaining and developing their sixth sense very rapidly, with, for example, the groundbreaking IPC’s Connected Factory Exchange (CFX) that now exists allowing all sorts of machines and systems of from every vendor to easily speak with each other through a universal language. As human intelligence was derived from data from the five senses, digital intelligence begins mainly with the development of the sixth.

The Evolution of Digital Intelligence

Carrying on with the Smart warehouse solution, as soon as the Industry 3.0 hardware was put into operation, software AI entered, with Industry 4.0. The thought of the word “random”, instantly makes people wonder how efficient can a system be that is randomly based. Making operators looking all over the place for an available empty spot to put materials is quite superfluous. The aptitude of digital warehouse systems originates from the need to deal with that. The warehouse software knows exactly what is contained in each spot, and can instruct specific locations where materials are to be put away, either by human or machine, not random. The intellect arises from picking the location. If the software developer is aware of how far it is between bins and the material checkout point is a problem, the software can be developed to examine, for example,  which materials are more frequently consumed, and move them a location not so far from the checkout, and even begin to optimize the route taken to gather sets of materials as may be required. This accelerates material availability, reducing the time it takes to replenish materials. The dimensions, shapes, and kinds of materials can be taken into account, in addition to whether the materials have storage requirements, such as electrostatic or moisture sensitivity. Deciding where to store materials may also be influenced by the ownership, cost, whether new or used, whether on a carrier such as a feeder, tax exemption, or vital for specific industries, etc. An AI process which is simply a software algorithm in which decisions are made based on the programmer's knowledge to make particular things happen in specific ways allow materials to be optimally stored instead of randomly.

Software algorithms for most discrete optimizations were initially designed to copy the ways that humans would handle a problem, and then do it quicker and more precise. Eventually, as people no longer wanted to manage increasingly complicated optimization problems in this discrete manner, digital modeling of operations stepped in. Rather than make an explicit logical course of action in software to go by, beginning with so-called “genetic” algorithms, software was created that would outline a set of rules that would gauge the effectiveness of a solution, by scoring what was successful and what wasn’t, against the goals. Different combinations of solutions criteria were created unsystematically, each of which was evaluated by the software. This process was repeated, with varying combinations, until the best outcome was achieved. This method did not work as well for more complex problems. Every time a factor was added to the problem it substantially increased the number of iterations needed to achieve the best result. However, it became easier to develop measurement software because you all you needed to do was modify the measurement system for different problems. Procedures were designed to cut down the uncertainty of the criteria arrangement in an effort to expedite the process. Nonetheless, the most effective optimization of an SMT machine program would require hours to complete. Concessions were made in order to finish early to allow the customer to agree to a moderately good optimization in an acceptable time.

Frustratingly State Of The Art

In the warehouse, we cannot even wait seconds for any Smart software to make choices, even with the addition of the most powerful computer hardware to make things seem real-time. The decision-making required with Industry 4.0, for instance, to direct materials being put away in a warehouse, examine the visual inspection findings, etc., forces us to take another approach, possibly mixing the discrete algorithms with those from the trial and error method. We are still quite a bit away from seeing the software making quick and complicated smart decisions in automated Industry 4.0 for the entire factory. The amounts of investment into doing so must be sensible, based on increasing customer value.

True Digital Intelligence

To get to the point where software developers can once again take it easy and allow the truly intelligent software to do the work, is something that maybe can be again learned from people, explicitly, very young people. Children develop naturally through trial and error. Our five senses help us understand pain, as well as pleasure. Our human algorithms favor the second in most situations, and we constantly learn and alter our behaviors as a result of our experiences. With software being able to acquire more and more data, using their sixth sense connection of connectivity, doing similar activities becomes possible. Generating the desire to attempt things in alternative ways, to see if “pleasure” can be obtained by doing something different, is the beginning of the true AI algorithm. AI software developers will have to identify what is the pain and pleasure, that is, the incentive for any given digital solution. Globally factory managers already have an enormous amount of knowledge, and if we are being honest, they will have gotten it through many instances of trial and error. unlike the human manager, AIs in the digital age do not need to re-learn experiences like people as they change roles. Even though it may seem strange to put the intelligence of a “baby” in control of a plant, it only needs to happen one time. It might not even have to begin in the physical world.

Conclusion

Being involved in the AI evolution is helpful to factory operations, as well as fascinating. Advancement towards smarter software algorithms, and even to real AI, is already different from hardware, in the way it adds value. Investment into a progressive, digital MES software platform today offers more benefits than broad MES software. Being an adopter of digital MES today will end up being the most worthwhile investment a factory could make, as we see the beginning of intelligence being used in increasingly more facets of making operational decisions in the factory, in addition to expanding the operator process. Software with this type of functionality needs to be selected carefully, looking towards the industry trailblazers in terms of technological foresight. With the introduction of technologies such as IPC CFX, to many, it may simply give the impression of a “Smarter” and a more inexpensive way to do interfacing. However, CFX provides the opportunity for a much bigger step towards AI, for which some of us are ready. This is a clear distinction between entities that have been major influencers and leaders throughout the progression of CFX and those who will merely deliver an updated interface for their old software. It is not difficult to distinguish which road leads to a larger reward.