Artificial intelligence: what are the consequences for PIM, MDM and DAM?

Artificial Intelligence
Friday 28 February 2020 
Etienne Sola Etienne Sola
3 minutes
Artificial intelligence: what are the consequences for PIM, MDM and DAM?

Progress in AI has been impressive. It does still have its limits, but it is already possible to use it for product information solutions, reference data or in digital property management systems.

New call-to-action

AI can do anything

Artificial intelligence, the medias love it! They talk a lot about it, and in particular about how powerful it can be. And to persuade the doubters, it managed to beat Humans in the game of Go, an achievement which was considered impossible for a machine. Calculating all the possibilities would take too much time and the game of Go requires experience and intuition… but the AlphaGo's progress in Deep Learning came out on top.

Thanks to this major technological advance, we have discovered that Artificial intelligence can search for cancers, drive vehicles, speak naturally… Artificial intelligence can also make coffee and cook.

IA definition et PIM MDM DAM
With this progress, we are made to believe that artificial intelligence is already capable of anything, such as programming instead of humans, composing music or creating works of art. Literature and science-fiction films are making us believe that machines are preparing an uprising. Skynet will take over. Mankind is in danger as it is losing control over machines which are superior to us.

Artificial intelligence doesn't think

However, artificial intelligence doesn't think, or at least not the AI that exists today. AI is simply a system of highly advanced algorithms which are defined and controlled by man. Artificial intelligence can use gigantic quantities of data (big data) to extract statistics and complex mathematical formulas which enable it to recognize and reproduce. And that is where it's different to human beings. AI is not yet capable of recognizing and reproducing by itself. And such autonomy is not on the cards for the near future.
In a recent interview, Luc Julia, co-creator of Apple's assistant (Siri), reminds us that " “Artificial intelligence doesn't exist” as explained in his book of the same title. There's still a long way to go before artificial intelligence becomes a real intelligence capable of conscience, emotions, instinct, working in several fields, learning quickly by itself…

exploiter les progrès en IA dans le domaine PIM MDM DAM
Recent progress in the fields of machine learning and deep learning are based on “learning” by the machine. This learning requires huge quantities of data and numerous repetitions in order to function correctly. Google's DeepMind beating professional Starcraft players is a great example: it took the equivalent of nearly 200 years of practice to achieve this level in the game. A huge amount of time, which shows, luckily enough, that humans learn more quickly. And that they are capable of capitalizing on their experience to use it on new topics (to learn a new video game) while artificial intelligence remains specialized on a highly specified pre-defined task.

PIM / MDM / DAM and Artificial intelligence

Today's solutions for product information management (GIP or PIM), reference data (MDM) and digital property management systems (DAM) allow us to centralize large volumes of data (technical characteristics, logistics, marketing, rates…). The aim is to improve the quality of information and to publish it efficiently on different channels.

The use of artificial intelligence in these solutions promises to solve all the problems occurring during the implementation of these software tools. These systems allow us to organize information with a great level of detail with the help of large volumes of metadata. The quality of information comes, first and foremost, from what is provided by all users: incorrect prices, inconsistent image, duplicate fact sheets or incomplete properties… all of these problems arise from incorrect input of information.

With AI, we could have an intelligent system, capable of revising all this information and identifying what is wrong, correcting errors and better still, inputting data.

The latest technologies already allow for some of these actions. Currently, the system can analyze your images and compare them with your product descriptions, translate texts automatically, analyze data and compare it to understand the statistical rules and identify the elements which don't comply with them.

To carry out such actions, you can use existing “knowledge bases” (such as “Google Translate”, an artificial intelligence system which has already learnt how to translate into numerous languages) or set up your own knowledge base. In this case, the task is complex as, to render this base useable, it needs to contain a huge volume of high-quality data. Not everyone has tens or hundreds of thousands of homogenous product fact sheets of similar quality with which to teach their system:
  • If the quality of data is uncertain or fluctuating, the learning process will be too and it will not produce good results,
  • If the quantity of data is insufficient, the learning process will not be a success,
  • If all the elements to be processed are too heterogeneous, no “statistical trend” can be identified correctly.
Creating training and verification games is a complicated task. It requires the help of experts (the famous data scientists). This is the most time-consuming task in machine learning projects.

As to the capacity of machines to input the correct data or to choose the correct image, it all depends on the objective:
  • Preparing long descriptions based on different characteristics or automatically associating images with the correct products based on their metadata: this is already possible with “traditional” algorithms. What AI can do on top of this is, from among thousands of product fact sheets, analyze how human beings prepared them to understand the rules to apply. It's not that useful as the users are generally capable of formulating these rules directly.
  • Writing appropriate marketing texts or choosing the best image for presenting a product: these actions are subjective and require intuition. You need “situational intelligence”. The existing artificial intelligence is not capable of this.

It's already possible to use the developments in AI in the fields of PIM / MDM / DAM. There are still limits, but they are being stretched a bit more every day.

New call-to-action

Read also
Can we trust AI to improve the quality of our product data?
Artificial Intelligence

Can we trust AI to improve the quality of our product data?

In the field of product information quality management, introducing AI requires paying particular attention to the interaction between AI and man, to maintain a high level of trust.

Read more
Essential steps for improving the quality of product information
Work process

Essential steps for improving the quality of product information

Product information is an essential element of your sales strategy. Discover the steps for improving the quality of your product information and boosting sales.

Read more
Product repository: definition and benefits

Product repository: definition and benefits

Implementing a product repository greatly increases the efficiency of product information management. Read the detailed definition of a product repository.

Read more
Receive our news
Enter your email address to receive our updates and new publications directly to your mailbox.
By clicking “Sign up”, I accept Goaland's privacy policy and agree for them to use my data to send Newsletters.
Your registration has been taken into account