A featured contribution from Leadership Perspectives: a curated forum reserved for leaders nominated by our subscribers and vetted by the Construction Tech Review Advisory Board.
Orange Labs
Future with Artificial Intelligence
AI is not quite a new topic for Orange.Back in 1992, one of our researchers published a scientific paper applying AI to the control of circuit switched networks! However, in the last five year, we have witnessed immense progress in different fields of Artificial Intelligence. Today, AI is capable of outperforming a human being for certain specialized tasks.One example is image recognition. Researchers have been working on image recognition for 40 years but the rate of progress was quite slow. Six years ago, the best recognition technologies with conventional techniques was making errors in 25% of the cases. In 2012, a team of Toronto changed the approach to image recognition, proposing for the first time to use an AI technique called deep learning, a term that is now in all the newspapers. With deep learning, the error rate was reduced from 25 percent to 12 percent and in a couple of years, was further reduced to 3 percent, which is better than what human beings can achieve on this specific task. Using deep learning algorithms, AI is capable of learning to perform specific tasks (such as playing go, piloting a drone, monitoring the security of a computer network. However, training the AI requires vast amount of data (for instance, 14 million images were used to achieve good performance on image recognition) and also a lot of efforts from scientists to configure the deep learning algorithms that “works” for that specific task. This is a very manual and costly process : there are an endless variety of specific tasks that can, one day, be solved by a specialized AI (a weak-AI as specialists call; however, still a considerable amount of time is required for a full-fledged general purpose AI to come into existence.
Do you think artificial intelligence can make a country really strong in terms of its military abilities, just like how we have been talking about nuclear power over the years?
A sort of fear is developed in the minds of the people while speaking about the implementation of AI in the military sector. In the short term, the adoption of AI will only improve specific operations such as gathering intelligence, improving the efficiency of the manufacturing
plants or efficiency of the services being offered.
The transition from legacy systems to AI-based environment is a methodical process, which requires the right form of implementation and adaptation
If a country spends on AI and employs data scientists for deep learning, it will foster efficient delivery of products and services. It gives a competitive advantage to the country in all domains. In the longer term, AI mayhave reasoning capabilities, decision-making capabilities, and have the knowledge to help humans in cyber warfare. Adopting AI is definitely a constructive step for staying competitive in any field.
Elon Musk spoke about how AI—if not utilized correctly—could lead to cyber war and even a world war three. What is your take on that?
I personally do not think that it carries any danger to humanity, because autonomous reasoning and general purpose AI are a very long term perspective. I understand that there is a social debate going on where some parts of the world believe that adaptation of AI will bring in catastrophic consequences.It is more important to work on today’s AI specific issues, such as biases in AI algorithms, than to launch polarized debates about future and highly speculative problems. We believe at Orange that responsible use of AI has great potential both for the economy and for individuals. Our teams are focused on implementing AI to optimize labor and creativity of the services offered.
AI has the potential of wiping 5 million jobs across the world. Is AI turning out to be a threat to mankind?
When mankind created the steam engine, it replaced millions of jobs in horse carriage services, agriculture, and various industries. It was the same with the computer; before computers, people were using pen and paper for documentation. It is true that new technologies change the way we do certain tasks, but they also create more job opportunities; very often, it is the tasks within the jobs that are transformed. My view is that, in the long run, tasks will be modified—certain jobs will surely become meaningless but new jobs will appear and everything will be transformed. We have to be careful with the speed of change as the transition will surely be challenging.
Are there any trends that you foresee in terms of AI?
Our teams of researchers at Orange have observed therewill surely be a revolution in robotics; we already have drones, factory robots, farming automation and much more. I'm pretty sure that in a few years, no farmer will drive a tractor and a big part of farming operation will be automated. There is a lot of transformation being observed due to the blending ofrobotics and AI. Factories are already being automated in a close-controlled environment and in uncontrolled environments such as cities, nature, and public places, specific AI techniques are required. It will surely be an interesting combination for enhancing automation and efficiency in various sectors.
Could you suggest methodologies for organizations with legacy environment to adapt AI and its potential?
Taking our own example, our B2B branch Orange Business Services provides solutions for companies seeking digital transformation. We acquire the data and analytics from our clients, which are later improved using AI and deep learning to gain desired outcome. Further, deep learning algorithms are capable of detecting abnormalities in the systems. But, mostly importantly it is used to improve the processes and operations carried within a company.Companies using legacy systemscan undergo a slow transition while implementing AI for process improvement but can gradually move towards an all AI approach for the overall operations of the company. The transition from legacy systems to AI-based environment is a methodical process, which requires the right form of implementation and adaptation.
