A Ética da IA: Um Desafio para a Próxima Década?

AI Ethics: A Challenge for the Next Decade?

June 5, 2024 Roberto Magalhães

AI is everywhere, but while there is a lot of excitement and uncertainty about its future, we have to start asking ourselves: is what we are doing ethical? Where will AI take us next?

The nature of artificial intelligence (AI) is a complex and constantly evolving field. AI is the science that creates intelligent machines that can think, learn and act autonomously. It has been used in many areas such as robotics, computer vision, natural language processing, and machine learning.

AI has the potential to revolutionize the way we interact with technology and how we live our lives. Basically, AI is about creating systems that can make decisions based on data or information they receive. This could be anything from recognizing objects in an image to playing chess against a human opponent.

AI aims to create machines that can understand their environment , predict outcomes and take appropriate actions without being explicitly programmed by humans. However, there are ethical implications associated with implementing AI technologies. For example, who would be held responsible if an autonomous vehicle caused an accident due to a programming error or if a person was wrongly scored (expected) to commit a future crime?

Similarly, what safeguards should be implemented to ensure a positive or fair outcome if an AI system is used for decision-making in healthcare, genetic sequencing, or criminal justice contexts? These questions raise important ethical considerations when it comes to using AI technologies in real-world applications.

We're already seeing everything from ChatGPT text and image generation to social media classifiers, aiding business processes and shaping our culture and societies in ways we can't yet predict. Considering this, isn't it important to understand the underlying ethical issues and see how they can impact our projects?

Problems with AI

Biases

Automating processes and making decisions with AI are prone to bias because AI systems are based on algorithms and data sets that are not transparent or easily understood by humans. As such, there is an inherent risk that these decisions may be biased or inaccurate due to errors in the underlying data or algorithms used by the system.

Perhaps the biggest example was Amazon's catastrophic secret AI project. In 2014, Amazon began creating computer programs to review job applicants' resumes in order to mechanize the search for top talent. The company's experimental hiring tool used AI to give candidates scores ranging from one to five stars.

However, in 2015, it was discovered that the AI system did not rank candidates in a gender-neutral way, as the majority of CVs came from men. Amazon edited the algorithm to make the scores neutral, but could not guarantee that the machines would not create other ways of ranking candidates that could be discriminatory.

Ultimately, in 2016, Amazon disbanded the team because executives lost confidence in the project and recruiters never trusted their ratings alone. This experiment serves as a lesson for companies looking to automate parts of their hiring process and highlights the limitations of machine learning.

The black box problem

The AI black box problem is a concern in the computing world as it refers to the fact that with most AI-based tools, we don't know how they do what they do. We can see the input and output of these tools, but not the processes and functioning between them. This lack of understanding makes it difficult to trust AI decisions, as mistakes can be made without any moral code or reasoned understanding of the outcome.

The cause of this problem lies in artificial neural networks and deep learning, which consist of hidden layers of nodes that process data and pass their output to the next layer. Effectively, no engineer can tell you how a model arrived at a conclusion . It's like asking a neurologist to look at brain activity and tell us what someone is thinking.

While the value of AI systems is undeniable, without an understanding of the underlying logic, our models could lead to costly errors and we wouldn't be able to say what happened except for “well, it seems to be wrong”.

For example, if an AI system is used in a medical setting to diagnose patients or recommend treatments, who will be held responsible if something goes wrong? This issue highlights the need for greater oversight and regulation in the use of AI in critical applications where errors can have serious consequences.

To solve this problem, developers are focusing on explainable AI , which produces results that humans can understand and explain. But this is easier said than done. Until we can create interfaces that allow us to understand how AI black boxes make decisions, we must be extremely careful about their results.

We also know that there is wisdom in crowds. Explanations and decisions are better made by a group of well-intentioned and informed individuals than by any liberal member of the group.

Human error

Sometimes it’s not a question of “can we?” but rather “should we?” Just because some brilliant mind thinks of a new application for AI doesn't mean he has the ethical basis to see the ramifications of his actions. For example, Harrisburg University in Pennsylvania proposed an automated facial recognition system that could predict crime from a single photograph .

This sparked a backlash from the Coalition for Critical Technology , who wrote a letter to the publisher of Springer Nature urging them not to publish the study due to its potential to amplify discrimination. The publisher responded by not releasing the book, and Harrisburg University removed its press release.

As attractive as the project may seem, there are no two ways about it: it is discriminatory at best and a safe path to ethnic profiling at worst. We have to be extremely careful with our solutions, even if they are built with the best of intentions. Sometimes we are so tempted by the novelty or usefulness of technology that we forget its ethical ramifications and social impact.

Privacy

The use of AI in data processing and analysis can lead to the collection of large amounts of personal data without the user's permission. This data can then be used to train AI algorithms, which can then be applied to various purposes, such as targeted advertising or predictive analytics.

This raises serious ethical questions about how this data is collected and used without the user's consent. Furthermore, the use of AI also poses a risk to privacy due to its ability to process large amounts of data quickly and accurately. This means that AI algorithms may be able to identify patterns in user behavior that could reveal sensitive information about individuals or groups.

For example, an AI algorithm may be able to detect patterns in online shopping habits that could reveal someone's political leanings or religious beliefs. To address these concerns, it is important that organizations using AI technologies comply with the General Data Protection Regulation (GDPR) when collecting and processing personal data.

The GDPR requires organizations to obtain explicit consent from users before collecting their data and to provide users with clear information about how their data will be used. Additionally, organizations must ensure they have adequate security measures in place to protect user data from unauthorized access or misuse.

Now it is very important to understand that a model does not save the user's information, but rather, the weights (the relationship between two neurons in the model) are calculated based on this information. This is a gray area in data collection regulations, but it presents a difficult challenge.

Remember what we mentioned about the black box? Well, how can an engineer know if a given weight was based on someone's preferences? The answer is; It's very difficult to know. So what happens when a user wants to be removed from a sample? Brilliant minds across the planet are working on this problem under the umbrella of willful forgetting but neither the ethics nor the technology are clear yet.

Finally, it is important for organizations using AI technologies to consider the ethical implications of using user data without permission to train AI algorithms. Organizations must strive for transparency when it comes to how they use personal data and must ensure that any decisions made by their AI system are fair, impartial, protect privacy and are used for the good of our society. AI systems should not harm individuals or groups based on their race, gender, political affiliation, religion, etc., as this may lead to discrimination or other forms of injustice.

Security

As more data is collected and analyzed by AI systems, there is a risk of personal information being misused or abused by malicious actors. Some harmful practices include:

Automated identity theft: Malicious actors use AI to collect and analyze personal data from online sources, such as social media accounts, to create fake identities for financial gain.
Predictive analytics: Malicious actors use AI to predict an individual's behavior or preferences based on their location or purchasing history to target them with unwanted ads or services.
Surveillance: Malicious actors use AI-powered facial recognition technology to track individuals without their knowledge or consent.
Manipulation of public opinion: Malicious actors use AI-based algorithms to spread false information about a person or group to influence public opinion or sway elections.
Data mining: Malicious actors use AI-based algorithms to collect large amounts of personal data from unsuspecting users for shameful marketing or other nefarious activities.

Organizations need to ensure that adequate security measures are taken so that user data (and all data) remains secure and private. Should we worry about unstoppable generative AIs? Not yet. These systems are amazing, but they are still very limited in their scope. However, all “best practices” in cybersecurity must be diligently applied to AI (e.g. multi-factor authentication, encryption, using AI to detect anomalies in network traffic, etc.) because they can create problems much faster than a human being. For example, an AI could take a leaked password and try to match it to potential companies where the victim is working – just like any other cyber attack, except it can do it in a fraction of the time it would take a human. . .

The Effects of AI on Job Layoffs

Another ethical concern related to the use of AI technologies is job displacement. As more tasks become automated through the use of AI systems, there may be fewer jobs available for humans as machines take over certain repetitive roles traditionally performed by people. This could lead to rising unemployment rates and economic instability as people struggle to find new job opportunities in an increasingly automated world.

While there may be reasons for concern, we have to remember that this is not the first time something disturbing like this has happened. Let's not forget the industrial revolution. While the artisan and merchant class, in general, suffered from the industrialization of the economy, most people were able to adapt, which led to the division of labor as we know it today. So what can we do to mitigate job displacement?

Firstly, it is important to stay up to date on the latest developments in AI technology to identify potential areas where your work can be replaced by an AI system and how to evolve to remain competitive in the job market.

Second, professionals need to focus on developing skills that are not easily replicated by machines or algorithms. For example, creativity and problem solving are two skills that are difficult for machines to replicate .

Just as calculators replaced the need for manual calculation but at the same time opened up the possibility for scientists and engineers to spend more time innovating, AI can free us from repetitive work and provide the extra processing power to increase our productivity.

What is the next

As AI becomes more widespread in our society, there is a need for greater public education about the application, personal implications, and risks of this technology. We must foster an AI-aware culture that understands how AI is shaping our lives and our businesses.

It will also be important for organizations implementing AI solutions to ensure they are taking steps to protect users' privacy and security, whilst also enabling users to access the benefits offered by this technology. With adequate oversight and regulation, accountability and accountability issues can be addressed before they become serious problems.

Lastly, we must create a regulatory framework that holds companies accountable for their use of AI and ensures that any decisions made by AI systems are justified, virtuous, fair and ethical. With these measures in place, we can ensure that artificial intelligence is used responsibly for the benefit of society as a whole.

If you liked this, be sure to check out our other articles on AI.