How artificial intelligence assistance will shape new skills demand in operational digitalization

The next big skill that will be massively needed for any high skilled engineer will be communication skill. As Minsky once said :" how should the programmer ask, recommend, and advise the computer?" It turns out now that computers and human will interact in plain natural languages of the humans. But before reading our intent and being able to execute any of our desires, a new era of programmers will have to master the way we express our intents to the machine. In plain language but expressed in the best manner. Prompting and designing the code is no more a matter of mathematical and writing in formal structured programming languages, it will be now a matter of applying this old Boileau's statement :"What is well understood is clearly stated, and the words to say it come easily." The main roadblock and risk we face to achieve this goal are numerous.

Teaching is not programming.
A programmer does not teach the machine, it gives its instruction. This is now step by step in full transformation as stated by Gilder (1997) " The dependence on human faculties has been decreasing in each successive stage of computer development" For more than fifty years, it has been necessary for computer software and hardware designers to rely on experienced users, many of whom had been programmers themselves, to provide the information that the designers needed to do their jobs. Despite many advances in computer technology, Gilder was writing in 1997, that "within the next few years a very large percentage of computer software will be provided by trained users who know what the computer can do" The human is just a key of an old symbolic language. For an industrial process automation, there is a next step that is an evolution of this concept. What we are about to develop is not a structured programming language for an old computer, it is a new learning environment of our desired productions. A new learning computer will be a major step in our evolution of an industrial automation. Learning the new symbolic language or operating a product from this new computer will be now something we can achieve by combining the procedural (old computer concept) as much as the new learning concept.

Teaching versus programming
Learning versus programming, the interface between the designer and the machine with whom it will communicate will be now vastly different. A new sophisticated interface will be required for this new activity. If what we are going to achieve is an autonomous industrial production, it is not simply on the user to take his turn to instruct and interact with the new computer. We want now a dynamic interactive interface between designer and machine. This may not be just due to the limitations of a specific implementation, it may be true that a design and interaction interface between designer and machine will always exist as a separated phase. In other words, a designer will always interact with a separate learning computer and so it will never be the same user. The transformation From an old to a new or from a learning to a teaching computer is a conceptual step towards reaching our goal. A primary step is to imagine a new concept of interface between designer and machine and to think and conceive an appropriate implementation and implementation paradigm to achieve this new concept. From this objective, the paradigm and the implementation will be the most challenging.

Task distribution
Possibly, it is more appropriate to speak of "task distribution". Learning and teaching the new computer will not be a natural human interaction but a conceptual "learning". It is not the task but the approach that will be the major challenge. This will require a new paradigm in the method of learning, and, equally important, will require a new paradigm in the design of the interface. The fact that these two methods are connected and coevolved can be measured from their common history.

A learning computer is not a program.
It is something different and not another program, it is a programmable device. It is a concept that is truly a new way of thinking. It will need a new paradigm.
Even though many practical learning and teaching machines are on the market, many of them only address the "writing" (programming) concept. They are basically teaching or learning devices with little imagination about the new "writing" concept and thus not programmable or programming devices.
The scope of this review is that of the computing process and not of the teaching machine per se. Although the "teaching and learning" machines are not generally thought of as a computing process, they are the most significant and widely used computing devices today.
A significant barrier to the development of programmable devices that emulate the learning or teaching of a human is the difficulty of coordinating the computation, storage and retrieval required for the learning process.
The teaching and learning devices on the market today generally perform the function of writing and not of computing. They are essentially instructional devices with little or no programming or computational capabilities. With their progress, in the areas of the "teaching and learning" machines, we can expect a parallel trend to progress in the areas of programming and computing.
This means that when we will have a programmable learning and teaching device, this will be a computing device. A learning and teaching device that will be able to teach and learn is a computing device, no more and no less. With this new concept, it will be no longer a teaching machine or a programming device. A learning and teaching device will not be simply another instruction device, it will be a computing process and not a computer.

Written in collaboration with gpt-J-6B