by Maxime Claisse
- Artificial Intelligence Lead (Supply Chain & Optimisation), Sopra Steria
by Alexis Girin
- Team Leader / Domain leader Robotics, CIMPA
by Benoit Spolidor
- Head of Artificial Intelligence, Sopra Steria
Whilst there is no set definition of artificial intelligence as of yet, experts are in agreement that AI can simulate human cognitive capabilities such as perception, reasoning, action, and learning. Research into AI started in the 1950s but it is only thanks to the progress made in computing power, data quantities and new algorithms since the beginning of 2010 that new possibilities are really being explored. AI now promises to completely transform the industrial sector – one of its primary applications.
Artificial intelligence: decision-making assistance for the industrial sector
In industry, the primary role of AI is to offer assistance in decision-making. There are currently three levels of assistance: the first, by using big data or a significant quantity of intelligence, is descriptive AI which can simplify information and
present it as a dashboard and KPI (Key Performance Indicator). This is a significant time-saver as analyses performed with AI are quicker, cheaper, and far more reliable. The second level is predictive AI, which provides probability-based projections,
meaning that future situations can be anticipated easily and risks around the supply chain or manufacturing failures, for example, can be predicted better. The third and last level is prescriptive AI, which is no longer limited to just providing the
necessary information for decision-making but also guides employees, offering recommendations on the system status (known as the ‘next best action’). In some cases, it can also be used to relieve humans of low-risk or non-strategic tasks.
We are now seeing an increase in the number of warehouses, for example, that are entrusting sections of their logistics to driverless vehicles. Whatever the type of artificial intelligence used, the system will make it possible to receive a significant
data stream, using IoT platforms for example (Internet of Things), in order to analyse the data for strategic, tactical or operational purposes.
Artificial intelligence and industry
Many factories are now equipped with digital doubles, allowing them to make their processes, machines or even human movements, virtual. These simulations, also known as digital twins, can test recovery scenarios for the optimal operating conditions following
the predictive detection of a machine fault, or even optimise a production chain using process mining. Employees are now finding themselves receiving assistance with their daily workload through the use of new tools (MR, AR, VR, mobiles, etc.) which
offer a better understanding of their environment. This can be in the way of ‘pick-to-light’ or ‘light-directed picking’ (a computer-assisted procedure for preparing orders), or visual detection of incoherent processes, for
example. Thanks to artificial intelligence, these ‘augmented’ employees are no longer limited to certain roles: an operator can focus on meaningful jobs with a high added value, and can move away from jobs that are repetitive, monotonous
and time consuming. Finally, the progress made in AI has completely transformed robotics as we know it. Now more autonomous, flexible and lighter, robots can move on their own to reach the location where their process is to be deployed. The processes
that robots can operate no longer just rely on a deterministic algorithm, but rather can be fine-tuned using AI and learning. Today, humans and robots coexist, cohabit, and cooperate.
Democratising artificial intelligence: challenges and issues
Despite some real benefits, there are still several barriers to the democratisation of AI. Democratisation assumes that employees have received the correct training. This is a long and fastidious task but one that could be simplified in the next few years
thanks to the progress made in the field of human-machine communications (through augmented reality headsets or voice-based tools, for example).
Then there are, of course, the legal, regulatory and ethical issues. The implementation of the GDPR in France (as well as the rest of the European Union), in addition to industry players wanting to secure access, naturally limits the exploitation of this
data. But for foreign business competitors, this is not always the case.
Data quality and the industrialisation of data processing need to be managed even further. Whilst it is relatively easy to implement AI processes on an industrial site, restrictions due to the scale of the operation and global management make it difficult
to set up these processes at group level.
There is still a lot that needs to be done to overcome these technical, human and legislative challenges which are limiting the use of artificial intelligence. But the pressure on the industry’s shoulders in terms of competitivity, flexibility and
energy performance, can only be relieved through artificial intelligence itself. This is now the direction taken by current research; a shift towards smart factories where resource management will be more efficient (in terms of means, materials, energy)
and where the operators themselves will be at the heart of this new industrial revolution.