Opportunities and Risks of AI

5 points to remember

  • Artificial intelligence (AI) is the lever of the third economic transformation at the service of business.
  • The use of AI improves productivity as it drives innovation, increases work efficiency, and guarantees instant reaction to demand.
  • The other side of AI’s power is its propensity to make mistakes… in a different way to humans.
  • The added value of AI is based on humans.
  • AI becomes a factor in transforming corporate governance and the skills of its employees.

Introduction

The wave of AI is sweeping across all domains. Making rapid progress, it abruptly opens the field of possibilities, offering remarkable transformative potential in the business world. At the business level, it reveals new operational modes, allows for rethinking decisions, redefines interactions with customers and maintains competitiveness in a constantly changing universe.

The deployment of AI is already taking place at the business level, with companies embracing its promises while being aware of the challenges that must be prepared for. We will explore the main opportunities in terms of innovation, efficiency, and commercial success, before going into the principal challenges it presents, to better overcome them.


1.    Opportunities for AI systems, and more specifically generative AI systems

Driving innovation

Increased collaboration between humans and artificial intelligence is leading to revolutionary advances in various fields.

From the discovery of a new antibiotic to the possibility of detecting the authenticity of a wine with the aid of AI, not forgetting a tool capable of analyzing the behavior of farm animals or the discovery of a new recipe for Coca-Cola: there are plenty of headlines defending and ratifying the advances made possible by AI.

Automating processes

By automating processes, a company’s employees can free themselves from repetitive tasks and focus on strategic, more complex activities with higher added value.

By way of example, in the online retail sector, customer queries can be directed to chatbots, capable of answering them immediately, seizing on them to process product returns and attaching themselves to orders according to the preferences they are given. In the restaurant sector, AI can automate responses to private messages on social networks, improve the management of customer reviews and write posts more quickly.

Advanced data analysis

AI is not just about analyzing data, it is also about harnessing immense volumes of it, in real time. In the age of data-driven decision-making, AI can predict stock market trends or generate personalized content for users.

Capable to project themselves into an explicit environment, businesses can adapt in real time. For instance, in the restaurant sector, it is possible to identify customer preferences based on their behavior: predicting affluence, avoiding waste, or even triggering promotional offers at the right moment.

Predicting failures and forecasting demand

Companies can use AI to predict failures and the maintenance that will follow. They can also anticipate the demand for products needed to satisfy fluctuating and sometimes unpredictable demand, such as that of the financial markets, where conditions change in a matter of milliseconds.

All the opportunities mentioned above are accompanied by challenges. Ethical, legal and governance issues, data management and the need to adapt to a composite and stratified regulatory context, all need to be considered so that companies can embrace AI on a solid footing – a Trustworthy AI.


2.    The challenges ahead

The adage “garbage in, garbage out” holds true in the field of AI. When models are designed or fed with incomplete, biased, or obsolete information, their predictive results are also impacted.

Ethical issues: the example of recruitment

AI technologies, which exploit machine learning and algorithms to analyze vast sets of data, are increasingly being integrated into various HR processes. From candidate selection and predicting employee attrition to talent management and succession planning, these technologies offer time-saving and predictive capabilities.

Diversity, fairness and inclusion are essential components of the recruitment tool: for example, the US Equal Employment Opportunity Commission condemned a company for hiring discrimination when it allegedly programmed its recruitment software to exclude older candidates.

Legal issues: copyright protection: provisions subject to change

As we mentioned in our article, Confrontation and Convergence: Copyright in the face of artificial intelligence, the scope of European regulations regarding intellectual property was one of the stumbling blocks during the negotiations in the various trialogues leading up to the provisional text adopted by Coreper on 13 February. Even today, the link between copyright and AI data training is still being debated. Despite the inclusion of copyright in the AI Act, The French Ministry of Economics and Finance would still like to reopen the Copyright Directive. The Ministry considers that generative AI is causing a revolution in terms of copyright, as protected works are not acquired by AI developers, but rather rented out to drive data. The committee of experts on generative AI, whose report is expected in March, is expected to comment on this issue.

In the United States, after the New York Times filed a lawsuit against Open AI, the major cloud and AI providers, notably Amazon, Google and Microsoft, say they will assume the legal risk of a copyright claim arising from the use of the LLMs they offer. But these protections come with important caveats.

Governance issues: leaders “expected at the turning point” of AI

For entrepreneurs in the sector, and in particular Sam Altman, CEO of OpenAI, there is no doubt about it: “Jobs are definitely going to go away, full stop.”  So, is employment the adjustment variable for the use of AI? The OECD stated in 2023, 27% of jobs will be at risk as a result of automation. However, the outlook for its deployment seems more optimistic, since in a survey conducted in October 2023, only 17% of managers believe that AI will replace their employees.

Often initiated by employees as part of their work routine, the use of generative AI deserves to be supervised, and must first be properly integrated into the corporate culture. This has an enormous impact on the spread of the technological proposition. Once accepted, it will be up to management to reconcile the human empiricism of its employees with that of the machine, which is also subject to numerous instabilities that can be controlled by training employees to master AI.

Generative AI is considered as a top-three priority by 89% of business leaders. However, only a tiny proportion of them have been able to train their employees, and almost half have not yet established a framework for its use, even though it is essential for several reasons.


3.     The risks of Generative AI

Data leakage

Like Apple and Spotify, several companies have already regulated the use of generative AI in view of the risk of data leakage in the form of, among other things, lines of code or minutes of meetings rich in sensitive data that some employees send to AI systems. Such supervision is even more necessary given that, according to the Fishbowl survey, 68% of the 43% of employees using AI do so without informing their superiors.

Deepfakes

OpenAI’s new tool named Sora, capable of creating ultra-realistic videos from text, could democratize the use of deepfakes in the short term. Used for identity theft, it could prove to be a formidable tool, multiplying situations of deepfakes. This type of deception is no fiction, as demonstrated by the recent case of a Hong Kong bank employee who transferred 25 million dollars to criminals posing as members of his management team via a video call.

Hallucinations linked to the use of Generative AI

Selected as  “Word of the Year 2023” by the Cambridge dictionary, hallucination occurs when generative AI models provide unexpected results or results that are not based on reality. Let us take an example from the field of legal practice: researchers at Standford RegLab have shown that in response to specific legal queries, the hallucination rate is around 70%, and that the more complex the task, the higher the hallucination rate. Lawyers and their New York-based law firm have been sanctioned for having, in all good faith, presented to the judge their conclusions generated by ChatGPT, conclusions based on purely fictional cases.

When driven by context-specific, real-time data, a model can generate more relevant content, enabling the company to derive high added value.

Regulatory shock

The “50 shades of regulation” that have emerged to date from the latest version of the European AI Act, combined with those of the United States Executive Order and the Chinese IMMGAI (Interim Measures for the Management of Generative Artificial Intelligence Services), cannot be understood without the help of a management toolTransformed into concrete, explicit actions, the various provisions of current and future legislation become essential compliance tools for taming the regulatory framework.


4.    Conclusion

Artificial intelligence, at the heart of business innovation, is undergoing a perpetually accelerating evolution that no one can slow down. This movement can only be seen as a virtuous circle within the framework of a trustworthy AI ecosystem that complies with the standards laid down and respects the ethical principles to which companies must adhere.