AI? What is it, anyway?

Life imitates art far more than art imitates life.  – Oscar Wilde

A quick review of technology websites yields a few common claims on Artificial Intelligence (AI): fully automated robots will displace workers, biased health assessments will lead to discrimination in healthcare, and a supercomputer will take over the world. 

The fears surrounding AI make sense. To date, AI is widely considered one of the largest technical advancements humanity has seen, after all. 

But what is AI specifically and what’s all the hype about? To understand, let’s go back to class and take coding 101. 

Computer programs are controlled by algorithms, a set of steps a computer must use to complete a specific task. Algorithms are the building blocks for traditional computer programs and run everything from your phone to your xbox. The simplest way to think about an algorithm is like a recipe on how to make a peanut butter and jelly sandwich (PBJ). 

First, take your bread. Then you’ll get the nut butter and fruit jelly of your choice. Once you have gathered your items, you’ll spread them on your bread, and voilà – a PBJ.

This entire process is an algorithm. If you follow these steps in the order outlined, you eventually reach the desired outcome: a simple, yet delicious sandwich. But if you didn’t follow these directions or made a mistake, it probably wouldn’t have tasted as good or you could have made a mess.

At their core, traditional computer programs work in the same way. Their code is made up of algorithms that tell them what to do. 

The building blocks of AI are just algorithms compiled together to create code. So, what’s all the hype about? 

At the most basic level, AI is personification; the development of computer systems that mimic human intelligence. 

Using algorithms allows the computer program to analyze massive amounts of data, enabling it to make decisions and adapt to new information without explicit programming. The process of a machine using data and algorithms to learn is called machine learning. Recent developments in AI have taken machine learning one step further to create a subfield known as deep learning.

Deep learning virtually replicates the structure of the human brain to a point where, as a program learns, it creates neural networks with multiple layers. These artificial neural networks have played a significant role in advancing machine learning, enabling models to learn and make decisions from data in complex and non-linear ways. They are a foundational technology behind many AI applications, such as open language models (programs designed to understand and generate human language, like Open AI’s Chat GPT and Google’s Bard).

Creating a virtual brain is pretty cool, right? Here’s where it gets interesting. 

Because deep learning mimics the human brain and replicates human intelligence, engineers across the industry are unable to explain the neurological pathways deep learning models use to draw their conclusions. This unknown methodology is called the “Black Box Theory”, where engineers know the data that trained the program, the question that was asked, and the conclusion it drew, but NOT how it got there. 

To many and maybe you, the Black Box Theory is alarming, but we challenge you to consider this question: Do you know the neurological pathway your own brain took to draw a conclusion on the Black Box Theory? 

Most definitely not. It’s a fear of the unknown and lack of control over how AI operates to a certain degree that has fueled much of the anti-AI and anti-deep learning rhetoric. That doesn’t mean there aren’t dangers to how AI can be used. We’ve already seen how AI can be weaponized to spread misinformation, as well as how it can be used in cyberattacks. Ethical concerns related to bias and fairness in AI algorithms are a prominent concern as these models are oftentimes not properly trained on data sets that include minority populations. As programs continue to be trained on outdated data sets, unintentional biases could be present as AI is expanding, eventually touching every facet in our lives. 

Undeniably, the strides that have been made in the AI space can’t be undone. AI has been here (in some capacity) for nearly fifty years, and it’s here to stay. 

Already, many industries have reaped the benefits of using AI for automation. 

By allowing AI programs to handle repetitive and mundane tasks, humans are allowed to focus on more strategic, complex, and creative activities. 

This automation is a key driver of efficiency and reduces operational costs in areas such as manufacturing, customer support, and supply chain management.

In healthcare, AI aids in diagnosis and treatment, while in security, it detects threats. AI’s natural language processing fosters accessibility and inclusivity, and its role in autonomous systems promises safer transportation. Furthermore, AI extends the boundaries of creativity, aids environmental sustainability, and drives scientific discoveries, making it the technology of not only our future but also our present. 

As AI continues to develop, questions must be asked to create a technology that enhances the fabric of society, rather than destroys it. How can AI, like any game-changing technology, be harnessed for the benefit of all humanity rather than exploited by a few? How can we encourage ethical and inclusive development, as well as transparency within the technology itself? For further discussions of these tough questions (and more!), continue to follow www.thesidebars.com.

In sum: 

1. AI enables programs to think and learn, and is built from algorithms and code. The process of training AI to make predictions and decisions is called machine learning.
2. Advanced AI programs utilize deep learning, a subsect of machine learning that mimics human intelligence. 
3. Concerns have been raised with deep learning as engineers don’t know the neural pathways deep learning programs use to draw a conclusion. This theory is known as the Black Box Theory.
4. Further concerns with AI have been raised in how it impacts national security and privacy. Additionally, the data sets AI models are trained on are unintentionally biased because of years of data discrimination. 
5. The use of AI can’t be undone as industries have already reaped the benefits of this technology in improving efficiency, security, cost reduction, accessibility, inclusivity, sustainability, and discovery. 
6. As AI continues to develop, parameters and safeguards will be necessary to ensure the benefit of this technology for all.

~ By Amanda C. Molina, Esq. and Abigail Harris