Argentina (University of Palermo) How is the roadmap of the application of artificial intelligence in a company?
Darío Pérez Principi, director of Engineering in Artificial Intelligence at the University of Palermo, describes a possible roadmap for the application of AI in an organization.
A manifesto published by the Center for the Security of Artificial Intelligence (CAIS), signed by experts and leaders of the world’s leading technology companies, tries to stop the advance of AI to “prevent” it from “leading to the extinction of humanity.”
This apocalyptic vision is analyzed in the following iProfesional interview by Darío Pérez Principi, director of artificial intelligence engineering at the University of Palermo (UP), who also describes a possible roadmap for the application of AI in an organization.
Why does a vision that oscillates between caution, fear and rejection arise around artificial intelligence?
Broadly speaking, there are two kinds of artificial intelligence (AI). The so-called narrow artificial intelligence (“narrow AI”) consists of applications that solve problems in a specific domain, such as a “machine learning” model for credit risk analysis.
(The other class is) the so-called artificial general intelligence (AGI), which would be intelligence in the sense in which human beings understand the concept of intelligence, that is, with the capacity for reasoning, representation of knowledge, inference of conclusions from models, learning, etc. Although we have not yet developed this latest type of artificial intelligence, applications such as ChatGPT make us believe that perhaps the path is shorter than we thought a priori.
Probably, in the future, humanity will be able to develop an AGI. At that moment we must think that artificial intelligence will cease to be a tool, to become something more: a digital entity with the same cognitive abilities as a human being or even higher.
From here two types of theories are fed, those with a pessimistic vision, of competition, based on feelings of worry and fear; and those positivist theories that see that, with something so powerful, a large number of the problems that afflict humanity can be solved, raising the quality of life through better health services, greater equity, better opportunities, slowing the pace of climate change, to name a few.
Is there awareness among the political, business, trade union and social leadership of Argentina about the risks and opportunities presented by artificial intelligence?
From a personal point of view, I have no ties to the political leadership and therefore my opinion will be based on the public information available to an average citizen.
Having clarified this, I think that our country is going through social, economic and political problems that have urgent status and therefore the agendas to address climate change problems, or the transformation of the labor market from disruptive technologies, are displaced to the background.
However, there are initiatives on the part of the government such as the recently issued provision that addresses the issue of ethical artificial intelligence, developed by thea Undersecretariat for Information Technology.
Beyond this, I think that the opportunities offered by AI are not being perceived, as the United States or China are seeing, among whom a race is developing for key technologies such as artificial intelligence, telecommunications (5G), quantum computing, etc.
-How can organizations (companies, State) take advantage of the tools offered by artificial intelligence today?
We can think of artificial intelligence as a set of tools that can be applied to any organization, whether public or private. The tools are available, you just have to use them.
My recommendation is that all those companies, governments or institutions that want to start using AI, incorporate their professional teams suitable in the subject. The ideal is usually to assemble teams in which there is a mix of senior professionals and junior professionals.
The Faculty of Engineering of the University of Palermo trains professionals in the careers of engineering in artificial intelligence and degree in artificial intelligence, with the possibility of studying both face-to-face and online. I believe academia can provide resources to help organizations incorporate artificial intelligence, if they aren’t already doing so.
-What technological and cultural requirements should an organization that decides to venture into the application of artificial intelligence in its business processes have?
We currently live in times of volatility, uncertainty, complexity and ambiguity (VUCA).
From the organizational point of view, it is usually recommended that the organization has what is called a “digital C-level”, that is, that the managerial managers have a digital mind.
This does not necessarily mean that they should be able to develop an AI model, but that they have knowledge of existing technologies (AI, blockchain, “virtual/mixed reality”, etc.) and fundamentally that they can understand how these technologies can help the organization and develop an implementation plan for them.
For this it is necessary to know what are the pain points of the organization and think about how they can be addressed and solved with the use of technology. I am aware that many traditional companies do not have a digital C-level.
My recommendation in this case is that they incorporate talent and human resources with the necessary skills to address the digital transformation that will ensure long-term sustainability.
-What are the best practices you recommend when applying artificial intelligence?
Many times when a technology becomes known worldwide, within organizations it is believed that it should be adopted without thinking too much about why.
The important thing is to know what are the problems that an organization faces or areas to improve and start by proposing solutions based on emp technology.Priority problems the resolution of which will bring the greatest added value.
In this way, it will be easy to justify the adoption of an AI project if it comes to solve a real problem. Before starting the AI-based project, it is essential to have the appropriate and necessary data sources to propose useful models.
If the data is insufficient, it will not be possible to develop a model with sufficient complexity to solve the problem. Many times it begins with an exploratory search, of the unsupervised analysis type, where an attempt is made to find certain patterns or coherence.
However, in general, analyses that use supervised learning tools, that is, those where a ground truth is available, are those that provide greater value and answer more complex questions.
Regarding the data, it is important to ensure that it is not biased data and in case the bias cannot be removed, be aware of the limitations that the model that will be trained from that data will present.
-How is the roadmap for the application of artificial intelligence in an organization?
Every AI project usually has a set of well-defined phases and that together is called the AI life cycle. It all starts with a problem definition stage, where we try to answer questions such as (the following):
– Who are we going to help with our AI project?
– How will we benefit the user and how will their use be measured?
– Why is using AI in our project better than the process we currently handle?
Then we move on to the data collection phase, where, as mentioned above, we must ensure the quality of the data. It continues with a stage of preprocessing and normalization of the data, and then moves on to the construction phase of the model.
The model must be trained and tested to know its performance. Once this is done, we begin to iterate over the entire process, that is, if we are not satisfied with the results, we modify the model, retrain and measure its performance again.
It may also happen that it is decided to incorporate new data or new data sources, which must go through the collection and preprocessing phases before that data is fed into the new model. Once we are satisfied with the behavior of the model, it goes into production where it will be used by the end user of the model.
-What are the most common and frequent mistakes you see in organizations that apply artificial intelligence?
Based on personal experiences, I can say that the most frequent mistakes are thinking that technology is not useful for the organization, without thinking that it is a set of tools that can benefit any organization, either through a reduction in costs, an increase in sales, an improvement in the user experience, an improvement in process efficiency, etc.
Another common mistake is to hire junior professionals without the guidance of a more experienced and knowledgeable professional. In these cases the result will be poor and very deviated from those expected.
Another common mistake occurs in the conjunction of “big data” and artificial intelligence. When working with a lot of data, correlations can be found without any theoretical basis.
It is important, then, to have a priori a set of questions that the data is expected to answer and to train the models with useful data. Sometimes it happens that more data is not better data. To avoid all these mistakes, I again recommend hiring qualified professionals from academic fields.