Understanding AI and Its Implications

By Annalisa A. ’24 and Maya Y. ’24

What is AI?

What are the first thoughts that come to mind when you hear the term “artificial intelligence?” Is it a row of faceless robots completing what were once human tasks? Is it self-driving cars flying next to bagel-stealing seagulls? Or do you think about how AI is already deeply integrated into our everyday lives? Before examining the potential implications of artificial intelligence, it is important to understand what it actually is. It may seem daunting to attempt to grasp the idea, but it isn’t as complicated as it may seem.

To put it simply, artificial intelligence is exactly what it sounds like. It is a system of computers or machines that copy the “problem-solving and decision-making capabilities of the human mind.” It’s kinda weird, right? Why would someone want to create a computer to do exactly what we can do, if we can already do it ourselves? Well, it is important to note the differences between artificial intelligence (AI) and human intelligence. Artificial intelligence can be faster, more objective, and more accurate, but it also has its limits. It is important to mention that even though artificial intelligence is technically objective, bias can be implemented by the programmer; bias can be an important ethical issue in different situations with AI. Human intelligence includes a capacity for memories, social interactions, and self-awareness, which computers can’t be taught to do with the techniques we have now (“Artificial Intelligence”).

Machine Learning – How Algorithms Become “Intelligent”

Machine learning (ML) is a subset of AI. It teaches a computer system to do tasks that need “intelligence” and to continue to learn based on data given. The “intelligent” computer uses AI to perform tasks and it uses machine learning to develop its intelligence. Computers can process a lot of data, which can then be put into a machine learning model, which then gives the outcome of the intelligent task. There is the concept of features and labels. Features are the characteristics or measures used to help us make predictions and labels are the expected outcomes that are trying to be predicted.

There are many machine learning models, such as regression, classification, clustering and information retrieval, recommender systems, and popular neural networks. Regression takes features and predicts a label using previous data with both features and labels to create a model for the pattern, for example, predicting the prices of houses. Classification labels groups based on features using previous features and labels, usually with categorical data, for example, filtering spam emails. Google searches are an example of clustering and information retrieval, which is a model that groups different points of data together and isn’t dependent on labeled data. Recommender systems use previous purchases and searches to recommend new products using general trends from a large set of data.

Lastly, we have neural networks, a part of deep learning, and the most advanced and most recent machine learning model. It was inspired by biological neural networks in the brain and uses interconnected nodes in three layers: an input layer, an output layer, and possibly many hidden inner layers, depending on its complexity. Some advantages of neural networks include their ability to work with incomplete information once it’s trained and its ability to generalize, seeing different complex relationships based on prior ones (Pedamkar). Usages of neural networks include facial recognition and image processing, stock market predictions, healthcare, and social media (Kaushik). In terms of social media, this can lead to important ethical issues about how AI can keep people on social media, being detrimental to health. Another ethical issue is that neural networks consume a lot of time, money, and resources, and therefore aren’t good for the environment.

From the simplest to the most complex model, machine learning basically uses weights and a bias. There are different ways to measure the accuracy of the model–the more complicated it is, the more accurate it generally is. Linear regression basically works with a linear equation: y = mx + b, or for more variables, it would look like y = w1x1 + w2x2 + w3x3 … + b. The w’s make up the weight vector, the x’s the feature vector, and the b is the bias, the y-intercept. These two vectors are then multiplied with a dot product to get an outcome of the slope and the bias, the y-intercept. The feature vector is given and the weight vector is calculated or “learned” from previous data. Neural networks take this and use more complicated algorithms and layers to create better outcomes.

Benefits and Consequences of AI

Advancement in this new technology of AI has many benefits and consequences that can widely change the world we live in. One example of AI is its use in cybersecurity to protect a company’s information from cyber-attacks, which is essential to prevent potential threats. AI has been integrated to enhance these cybersecurity systems, such as by reducing massive amounts of wasted time. Computer algorithms can identify potential threats in a matter of milliseconds from a large pool of data, which produces efficiency at the cost of developing and maintaining these layers of protection (“Artificial Intelligence”).

Photo by Kevin Ku on Unsplash

Biased data is another issue. AI algorithms are dependent on accurate data sets, therefore, human error and prejudice can lead to biased data, which can lead to biased outcomes. In addition, diversity is necessary not only for data but also for the people making the cybersecurity systems, which will lead to broader training through data analysis for the algorithms that incorporate additional experiences and views (Segal).

It is also important to note that the more technology advances, the more it can fall into the hands of cybercriminals. Hackers can also use and predict cybersecurity systems by using AI, and don’t need human interaction to breach data from a company. This plays into the bigger question of to what extent do we continue developing new techniques of AI if it can get to a point where the line blurs between the negatives and positives (Bulao). 

Circling back to the topic of artificial intelligence as a whole, there are so many effects that it can have, no matter how it is used. The implications don’t only extend to security but also to medicine, education, the workforce, and many more aspects of life. By understanding AI, hopefully you are now able to review your own relationship with it and how it leaves an impact on your life. AI can be a dangerous and powerful tool and it is our responsibility to guide it in the correct direction using our ethical decision-making skills (and not those of a computer). Nevertheless, it is significant to remember that no matter how much one tries to examine it, one thing is certain: only time will tell what the future has in store for artificial intelligence.

Works Cited:

“Artificial Intelligence (AI).” IBM, 3 June 2020, www.ibm.com/cloud/learn/what-is-artificial-intelligence. Accessed 14 Feb. 2022.

Bulao, Jacquelyn. “The Role of AI in Cybersecurity – What Does The Future Hold?” Tech Jury, 6 Feb. 2022, techjury.net/blog/ai-cybersecurity/#gref. Accessed 14 Feb. 2022.

Kaushik, Vanshika. “8 Applications of Neural Networks.” Analytics Steps, 27 Aug. 2021, analyticssteps.com/blogs/8-applications-neural-networks. Accessed 16 Feb. 2022.

Pedamkar, Priya. “What Is Neural Networks?” Educba, www.educba.com/what-is-neural-networks/. Accessed 16 Feb. 2022.

Segal, Eddie. “The Impact of AI on Cybersecurity.” IEEE Computer Society, www.computer.org/publications/tech-news/trends/the-impact-of-ai-on-cybersecurity. Accessed 14 Feb. 2022.