Is the growth of AI being conducted in an ethical and secure manner?

In recent years, the advancements made by researchers in the field of Artificial Intelligence have been substantial, particularly in the domains of machine learning, natural language processing, and computer vision. These developments have resulted in the creation of sophisticated AI systems capable of performing a wide array of tasks, from language translation to image recognition and autonomous vehicles. Despite these achievements, much remains to be explored in the pursuit of strong or human-level AI, capable of executing any task that a human can perform.
The potential impact of AI on society and industry is tremendous, offering the prospect of improving lives on a grand scale. Nevertheless, this cutting-edge technology raises significant ethical questions and concerns about its impact on the world, including the manner in which it is developed. Ideally, we aspirate for an ethical and responsible approach to AI’s evolution.
It is an undeniable truth that AI is changing the game, but its net effect on society and the economy is subject to debate and may prove to be more harmful than beneficial.
Joseph Fuller, a professor at Harvard Business School University, declares that:
“Virtually, every big company now has multiple AI systems and counts the deployment of AI as integral to their strategy”.
AI necessarily has an impact on workers' life. What does it mean for them? Well, they have reasons to be worried as soon as the tasks they are daily executing can be done by industrial robots. Usually, jobs affected are warehouse and manufacturing jobs, customer service, research and data entry, and somewhat “any task that can be learned” (source).
As a matter of fact, any company willing to secure long-term financials will invest the necessary funds to provide themselves with industrial robots and then later save money by firing excessive and too costly human labor.
An MIT (Massachusetts Institute of Technology) study from May 2020 about the impact of industrial robots on the job market from 1990 to 2007 in the US demonstrates that each new robot introduced per 1000 workers has led to decreasing available job offers and reductions of salary.
However, focusing only on the negative aspects would be foolish. It cannot be ignored that the robotization of employment can be devastating for many workers. But on the other hand, it also brings benefits such as the growth of companies, increased productivity, competitive advantages, and new job opportunities.
The International Federation of Robotics (IFR) predicts that the growth of industrial robots will continue even more after the coronavirus crisis. Major corporations are intending to evaluate their business strategy and accelerate the introduction of robots.
The federation argues that the current impact of employment automation is not different from the various industrial waves that we experimented in the past, during industrial revolutions (source).
Three significant concerns regarding AI have arisen within society.
The philosopher Michael Sandel stated:
“we are discovering that many of the algorithms that decide who should get parole, or who should be presented with employment opportunities or housing replicate and embed the biases that already exist in our society”.
Firstly, privacy and surveillance issues:
Many underdeveloped states, either lack laws regarding privacy, either the law is not sufficient. For instance, they might be outdated or not aligned with technological innovations. It is imperative to not find ourselves without tools against organizations committing AI-based violations against citizens.
Some privacy challenges include:
Concept | Descripition |
Data Persistence | Data stays stored longer than the lifetime of the person who created it. |
Data repurposing | Data is used for reasons that go beyond the original purpose. |
Data spillovers | Data is collected on people who were not the original target. |
Data consent | Data is collected without the free and informed consent of the target. |
An apt illustration is a well-known scandal surrounding the US firm Clearview AI, which has violated Canadian privacy law by amassing photography of both adults and children for the purpose of massive surveillance and facial recognition, all without obtaining any consent (to know more) (source).
Subjects of the European Union possess a specific right to request the deletion of all data about them, on request. This is granted by the General Data Protection Regulation which states:
“the data subject shall have the right not to be subject to a decision based solely on automated processing, including profiling”.
Article 17 of GDPR, also known as the “Right to be Forgotten” gives individuals the right to ask organizations to erase their personal data. However, organizations are not always obligated to comply. The article states that “the data subject shall have the right to obtain from the organization, the erasure of personal data concerning him or her without undue delay and the organization shall have the obligation to erase personal data without undue delay”.
Undue delay is considered to be approximately one month. In this process, it is also imperative to confirm in a responsible manner that the person requiring the erasure is indeed the data subject.
The right to be forgotten applies in the following situations:
When the personal data is no longer useful to the organization that originally collected it.
When the organization’s legal right to collect data was based on the person’s consent, and the person has withdrawn his or her consent.
When the organization relies on legitimate interest as a justification for collecting individual data, and this justification disappears.
When the organization collects personal data for marketing purposes.
When the organization collects data illegally
When the Organisation must erase personal data to comply with legal rulings or obligations.
When the organization collects personal data about a child to provide to another service company.
It does not apply in the following situations:
When data is used to exercise the right to freedom of speech or information.
When data is used to comply with legal procedures or obligations.
When data is used to perform a task in the public interest
When data is used for preventive medicine, only when stored by a health professional subject to a professional secret legal obligation.
When data represents important information for serving the public interest, scientific research, historic research or to pursue a statistical goal.
When data is used to establish a legal defense or to exercise a legal right.
(source)
Can artificial intelligence comply with the right to be forgotten? Let’s consider the case of ChatGPT.
ChatGPT uses reinforcement learning with human feedback (RLHF) to improve its performance, which means it collects data to provide better answers to users. But, with such advanced AI, the issue is that it continuously collects data, making it almost impossible to erase all of it if someone requests it. AI systems based on neural networks cannot forget as humans can do. It only focuses on newly collected data rather than removing older data from its memory.
OpenAI, the company that created ChatGPT, has stated in its privacy policy that all data will remain confidential and restricted to the terms determined in the contract. However, it is uncertain if this also applies to data stored within artificial intelligence such as ChatGPT.
The second issue that AI brings to the table is the concern of bias and discrimination.
When unfair treatment is caused by an automatic decision, which means generally taken by an intelligent agent or other artificial intelligence systems, the issue of digital discrimination becomes more prevalent. Indeed, more and more decisions are delegated to systems mostly based on artificial intelligence that used for example the “machine learning” method.
First, biases regarding AI training exists. Algorithms learn how to make decisions or predict decisions based on data that often contain past human decisions. If the data used contains existing prejudices algorithms will likely learn the same biased decisions. For instance, if the data used, does not accurately represent the characteristics of different populations, but instead reflects inequalities, the outcome will likely be severely distorted.
Additionally, there are biases regarding the use of AI. When use in the wrong way, in situations not designed for, the outcomes of these algorithms can be incorrect. For example, an algorithm designed to predict a particular result on a given population may lead to erroneous results when applied to a different population. This type of misapplication is referred to as interpretation bias (source).
The last concern which we already started discussing above, is the role of human judgment. Is it possible for AI to outthink human thought? Or do some elements in human judgment stay essential in making decisions about the most important things in life?
Michael Sande, here again, states:
“ AI not only replicates human biases, but it also confers on these biases a kind of scientific credibility. It makes it seem that these predictions and judgments have an objective status”.
Because machines learn with data they have been fed, their chances to do again past mistakes for example in the banking industry, which resulted in unequal treatment for African Americans and other marginal consumers, are huge.
Therefore, numerous experts express their concerns regarding the values that will be instilled in superintelligence.
MS- “Algorithm-driven decisions do have a potential dark side”.
(source).
Aside from these challenges, nations may not be taking adequate measures to ensure the responsible and controlled growth of AI. The race for artificial intelligence is considered as important as the arms race was during World War II.
The US already identifies artificial intelligence technology as the key that will ensure the United States the ability to win all wars of the future.
They also carefully watch other countries whose breakthroughs are increasing in the AI field. China currently represents the greatest threat. In 2017, the Chinese government published a detailed plan of its strategy to become the leader in the AI field by 2030.
Less than two months later, Vladimir Poutine stated,
“whoever becomes the leader in this field, will rule the world”.

Great article! The potential benefits and challenges posed by AI are certainly worth exploring, but regarding privacy and surveillance, what steps can individuals take to protect their personal data from being misused by AI systems? Also, it would be interesting to learn more about the current initiatives being taken to ensure responsible AI