Like many in my generation, I grew up watching the Jetsons, and the idea of a maid robot (like Rosie) was appealing for obvious reasons. I now have a robot that can vacuum my apartment and a machine that washes my dishes. While I don’t have Rosie doing these manual tasks for me, technology is indeed automating mundane tasks in my home…

The vision that science fiction and Hollywood sell to us as it relates to Artificial Intelligence (AI) is the one of a fully-functional humanoid robot, or a computer with human intelligence. They might want to kill you (the Terminator, or H.A.L. from 2001: A Space Odyssey) or they might be here to help you (Data from Star Trek: The Next Generation), but they are always highly cognitive machines, sometimes with human like personalities with emotions included (sorry Data!). This is what the academic world calls Artificial General Intelligence (AGI). Even with being introduced into the public debate by luminaries like Elon Musk and Bill Gates, we are very far from reaching it.

In the modern business environment, the reality is that we don’t want a fully-functional humanoid, but something that automates manual and tedious tasks.

It is important to understand what we define as artificial intelligence. This can be open to interpretations, but the most widely accepted is the use of Machine Learning and similar AI technologies to perform tasks that are done by humans. The tasks that require a minimal level of cognitive thinking (which one could define as tasks that can be performed by a human in less than a second) are in nature the kind of simple processes that can be automated with AI. Examples of these kind of tasks include understanding what is in an image, understanding the topic of a text, or predicting the time taken by a specific task.

Nuxeo on the Leading Edge of AI

At Nuxeo, we’re already begun to integrate AI technology into our platform with external APIs such as Google Image API, with Nuxeo Vision. Currently, we are moving into integrating further tools that allow our customers to create and adapt models to their specific needs in an intelligent and highly-automated manner.

Consider this specific use case.

ACME has a massive collection of unstructured assets, such as images files and documents. For the purpose of this example, let say the images are photos from various photoshoots. The company wants to add metadata to each digital asset in order to index and save time on refined search tasks.

Nuxeo can help create a processing chain that integrates the existing Google Vision APIs for object detection. At the output, Nuxeo identifies the list of objects present within the image being indexed. With these attributes automatically added to each digital asset, ACME is able to search for photos that contain, for example, a handbag and a motorbike. And, because of the highly-scalable Nuxeo Content Services platform, all of the pictures can be processed in a relatively short period of time.

Nuxeo AI: Modular and Dynamic in Nature

As the users access the photos, they may notice that their latest product is not recognised by the API. To address this, they can simply add a custom model into the process. As a result, ACME users tag the photos containing the image of their new product as well as those that don’t. With this data, the AI algorithm learns how to reproduce the task. At this same time, Nuxeo is able to add a second AI element to the processing chain in order to detect the latest product of ACME.

And, if there is a business value in knowing if a picture was taken indoor or outdoor?No problem!
A new custom model can be added. Once again, the system makes recommendations for new pictures to be annotated by humans and learns to reproduce the task by the Nuxeo AI algorithms.

As Nuxeo looks at other ways we can further leverage AI, our goal will be to continue to look for the best ways to create and integrate algorithms with the our ECM platform in a way that reduces manual human interaction and ultimately delivers real-world value to our customers.