Artificial Intelligence (AI) gets more and more advanced, being able to develop frameworks to build very technological Machine Learning models. All this technology is very important for industries, which need practical solutions to improve their productivity, reduce costs and still add quality to their products.
An interesting point, however, is that the end-user of these products is not seen in many frameworks. This affects the result achieved, after all, who should assess whether a solution created by Artificial Intelligence works or not, is exactly the end user.
Even though that is the reality, gradually the practice of considering these users to validate Machine Learning frameworks is growing.
Importance of the end-user in the development of Artificial Intelligence models
DeepMind is an Artificial Intelligence company based in London, which applies deep learning to solve complex scientific problems.
In partnership with the Met Office, UK state weather forecasting company, the company is a good example of how end-user collaboration makes a difference in the world of technology, AI, and machine learning.
In the partnership between DeepMind and Met Office, the second is the end-user, and it was fundamental to develop a deep learning tool CGMR (Deep generative model of rainfall). This tool uses state-of-the-art AI to accurately predict the probability of rain over the next 90 minutes at a given location.
This is one of the great challenges of weather forecasting, which can predict the weather for the long term, but not for a few minutes or hours in the future. For this, DeepMind has developed a system that considers location, changes in temperature, cloud formation, and wind when forecasting rain in the short term.
Before the creation of the CGMR, other deep learning techniques had already been created. However, these systems were good at performing a particular task, such as predicting where the rain will fall. Consequently, leaving aside the intensity of the precipitation.
This project was especially effective because DeepMind considered the Met Office team and their opinions as end-users. In this way, a really useful solution was created.
How to have end-user collaboration in model development
In the past, the concepts of user-in-the-loop and worker-in-the-loop already addressed human participation and collaboration in artificial intelligence frameworks.
The user-in-the-loop system is an end-user interaction with suggestions coming from software. On the other hand, the worker-in-a -loop is paid to monitor and interact with software solutions and ensure end-user benefits.
But anyway, no matter the option chosen by the industries, the important thing is that there is collaboration. And of course, always focusing on the end-user and their satisfaction with the product.
New practices to increase interaction between model and end-user
While end-user collaboration with AI frameworks is timid, this is something that is currently changing.
Today, instead of just supervisors or larger roles being involved with the end-user, this action is much broader.
For more people in the industry to be able to consider the end-user in the solutions created, it is important that training is carried out. In this way, the team will be trained for this type of interaction.
Thus, new frameworks have emerged looking for the best way to have the end-user actively participate in the creation of models. This is all without the need for technical knowledge of AI or programming languages.
Industrial sector and collaboration with the end-user
Currently, in the Artificial Intelligence and Machine Learning applications that exist in the industry, the participation of the end-user is almost essential.
And as a consequence of this, the success rate in employing AI in that industrial process increases.
What happened and still happens a lot is that AI solutions are created, but when putting them into practice, it is not possible to reach the imagined goals. On the other hand, with feedback from the end-user, these issues can be resolved and machine learning wins.
The future of applied AI
Although it is a technology on the rise, according to a survey by McKinsey Global, only 36% of Artificial Intelligence and Machine Learning projects pass the pilot phase. Furthermore, only 15% of respondents were successful in applying automation based on AI and ML across different sectors of a business.
This is because developers end up using only algorithms to create these designs.
And as you can imagine, bringing the end-user closer when planning and executing is a good option.
This is, therefore, the future of AI applied in industries and businesses in general. The technology to innovate processes and supply chains already exists, what is needed now is to add end-user collaboration.