AI: A Threat or an Opportunity?

 Artificial Intelligence (AI) refers to the simulation of human intelligence in machines programmed to think and learn like humans [1]. AI encompasses a wide range of technologies, from machine learning to natural language processing, and many more fields in between. The primary goal of many AI systems is to develop ways of performing tasks that typically require human intelligence, such as visual perception, speech recognition, decision-making, and language translation [1]. Many concerns have arisen on how beneficial AI will truly be, with the potential downsides encompassing a number of topics, such as ethics, data privacy and security, bias, employment rates, etc.

Such concerns are valid and it is easy to see where it comes from, with studies showing data to support these worries. However, AI is still a very new technology and, as such, has a lot of ground to cover in its development journey.

How AI Works

AI is a tool that operates through a combination of several disciplines, including machine learning, deep learning, neural networks, and natural language processing (NLP). Machine learning involves algorithms that enable computers to learn from and make decisions based on data. Deep learning, a subset of machine learning, uses neural networks with many layers (hence "deep") to analyze various data types. NLP allows machines to understand and respond to human language, which is essential for applications like chatbots and virtual assistants [2, 3, 4]. 

AI applications are vast and varied, ranging from generative AI, which creates new content (text, images, audio), to computer vision, which allows machines to interpret and make decisions based on visual data. For instance, self-driving cars use AI to recognize objects, judge distances, and make driving decisions. In healthcare, AI can analyze medical images to assist in diagnosis, while in finance, it helps detect fraudulent activities and predict market trends [1, 3, 4].

Ethics and Privacy

With an understanding of how AI works, we can talk about the issues regarding ethics and privacy. Data security is an important topic not only when it comes to AI, but regarding online activity as a whole. Since the spread of social media, companies seek to acquire data about their clients/users via online activity, a type of information called "Big Data". Big Data refers to extremely large and complex datasets, often used to uncover patterns, trends, and associations, especially relating to human behavior and interactions [5]. With this, companies are able to better tell what their target audience is interested in, and what marketing strategies they should employ to maximize profits.

     Since AI learns upon analyzing immense amounts of information, researchers Jennifer King and Caroline Meinhardt, from Stanford University, predict that the race for acquiring Big Data will increase even more, impacting "both individual and societal information" [6]. To mitigate this, King and Meinhardt have two suggestions. Their first suggestion is to minimize data collection by default. Today, most websites have opt-in data preferences as the standard option, meaning that the user has to manually change their preferences if they don't want to share their data. According to King and Meinhardt, having these preferences set to "opt-out" by default would decrease the amount of exposed user information. The second suggestion given by the researchers is to improve privacy and data protection on the AI data supply chain. Regulating the way we feed data to AI models would also help to reduce the impact on user privacy [6].

Bias within AI

Another concern regarding the incorporation of AI into everyday life is bias. Some researchers argue that AI systems can reinforce and even magnify biases present in their training data. A study [7] found significant racial and gender bias in commercial AI facial recognition systems, showing average error rates of 30% for dark-skinned women compared to 0.37% error rate for light-skinned men, across all commercial classifiers tested. This shows a worrying trend in the realm of AI facial recognition software, and leaves us to wonder in what other ways AI bias is impacting our lives without us noticing.

Fig 1; Example images and average faces from the new Pilot Parliaments Benchmark (PPB).
From Buolamwini et. al [7]

    Conversely, other experts argue that with proper oversight and diverse training data, AI has the potential to reduce human bias. For example, study [8] suggests that algorithmic transparency and accountability measures can help mitigate bias. As we know, AI learns by analyzing and interpreting data, but it can only use data it is provided. Logically, if the training data we feed AI models contain biases in any way (e.g. providing more light-skinned faces than dark-skinned ones to train facial recognition software) we can only expect the results it provides to be biased in the same manner. Thus, by feeding AI models with more diverse data sets, we can mitigate situations like the one exposed by [7]. Additionally, by feeding AI systems more diverse data sets, they can be designed to detect and counteract human biases in decision-making processes, such as in hiring, potentially leading to fairer outcomes [8].

Job Displacement VS Job Creation

With the constant improvement of AI models like ChatGPT, that offer very powerful resources to write a vast array of things, such as advanced and intricate code, a common concern is that AI will displace a significant number of jobs, including programmers, writers, etc. Study [9] estimated that 47% of U.S. jobs are at risk of automation within the next few decades.

On the other hand, some studies suggest that AI will create new job opportunities and industries. Study [10] projected that while AI might displace 85 million jobs by 2025, it will also create 97 million new roles, particularly in AI development, data analysis, and other tech-related fields. 

Moreover, new technologies are constantly making old jobs obsolete, but this doesn't mean that we shouldn't adhere to these innovations just because they are taking some people out of their jobs. When telephones were first invented, telephone operators (Fig. 2) were needed to connect the person who was calling to who they wanted to talk to. This generated a lot of jobs, but also meant that it took a longer time to connect the call, and reduced privacy for the people on the line.

Fig. 2; Seattle telephone operators in a private branch exchange in 1952

From Wikipediaa

Towards the end of the 20th century, as direct dialing telephones became popular and commercially available, switchboard operators became obsolete, and a lot of people lost their jobs because of it. However, this innovation in telephony also allowed for countless other jobs to emerge, and many others were benefited by this new technology.

Conclusion

The analysis of AI's features, ethical concerns, biases, and potential impacts on employment shows a complex technology with the capability to significantly transform society. AI's potential in machine learning, deep learning, and natural language processing offer several benefits, from enhancing healthcare diagnostics to improving decision-making in various sectors. However, as highlighted, ethical and privacy issues need careful consideration. In order to mitigate privacy risks and biases, robust regulatory frameworks and ethical guidelines are needed. This balanced approach ensures that AI development aligns with societal values and promotes fairness and accountability.

AI is nothing but a tool, and as such, can be used in many different ways. Although there is research showing the drawbacks AI can bring about, there is also evidence that shows how much society can benefit from this technology, and also how to fix what is still lacking with AI. By doing so, we can harness AI's transformative power to create a more equitable and prosperous future, benefiting individuals and society as a whole.