The A.I. Express

Disclaimer: This article contains AI generated content.

Artificial Intelligence (AI) has long captured the imagination of the public, often portrayed as a looming threat in dystopian narratives. Whether it’s taking jobs, controlling machines, or even evolving into something uncontrollable, AI has been misrepresented. Misrepresentation is partly due to a lack of comprehensive understanding by the media and public. Recent developments, particularly with models like ChatGPT, DALL-E, and MidJourney, have further fueled misconceptions, casting AI as an all-powerful, job-stealing, society-disrupting force. While AI certainly has its risks, the reality is much more nuanced, and it’s crucial to understand both the potential benefits and limitations of the technology.

At its core, AI is a vast field that encompasses a variety of technologies designed to simulate human intelligence in specific tasks. Be it Large Language Models like ChatGPT, or Generative Adversarial Networks like DALL-E etc. AI (simplifying here for the sake of argument) takes the form of programs that take inputs, process them using virtual neural networks, and generate outputs by tweaking the weight of each neuron until a desirable outcome is achieved. A significant branch of AI is machine learning, where the system is trained to learn autonomously by adjusting these neuron weights based on experience. The fundamental operation of AI involves complex matrix multiplication between input neurons and their corresponding weights, with the result being fed into output neurons. The output neurons either fire or remain dormant, depending on whether the desired value is reached. Despite the sophistication of this process, AI lacks true awareness; it is a set of algorithms executing tasks without consciousness, much like a light bulb that illuminates a room without knowing it’s doing so. The birth of AI can be traced back to 1956, when the term was first coined and put into theory. Just two years later, in 1958, Frank Rosenblatt at Cornell University created the Perceptron, a machine designed to mimic the behavior of neurons firing in the brain. The Perceptron could analyze images with a resolution of 20 by 20 pixels and identify simple shapes, with Rosenblatt even claiming it could differentiate between images of a cat and a dog. While the media, as is often the case, sensationalized these early developments, comparing the Perceptron to an “electronic embryo” that could soon walk, talk, and think, the reality was far more restrained. AI was still in its infancy, limited in scope and power due to the computational constraints of the time.

The term “artificial intelligence” itself may be misleading. While AI systems can learn and adjust their behavior, they do so because they are programmed to—not because they possess awareness or understanding. AI’s intelligence is task-specific, programmed to achieve certain outcomes through optimization and pattern recognition. The grim predictions of a machine apocalypse, akin to what’s portrayed in the Terminator franchise, are far from reality. Your toaster won’t deliberately burn your bread out of malice because it does not like you, nor is there any immediate danger of machines achieving self-awareness.

As computational power increased, so did AI’s capabilities. In the 1980s, Carnegie Mellon University created the first autonomous vehicle, which used a rudimentary neural network to process 30 by 30 pixel images of the road ahead at a rate of one image per second. This allowed the vehicle to steer at a maximum speed of 3 kilometers per hour. This slow pace was due to the limited processing power available at the time, raising an important question within the scientific community: was the problem that AI lacked the proper algorithms, or was it simply that computers weren’t powerful enough to run these systems efficiently?

In the mid-2000s, Fei-Fei Li, a researcher at Stanford University, had an idea that would prove to be pivotal in AI’s development. She hypothesized that intelligence could emerge from complexity—meaning that as neural networks became more intricate and were trained on larger datasets, they would become increasingly accurate. To test this theory, she created ImageNet, a vast database of labeled images, and held a competition to see which team could develop the most effective neural network. The effectiveness of each neural network was measured by the “top five error rate” (TFER), which tracked how often the top five neuron activations in the network failed to correctly identify an image. In 2010, the first year of the experiment, the TFER stood at 28.2%. By 2011, it had improved to 25.8%, but in 2012, a neural network from the University of Toronto drastically reduced the TFER to 16.4%, winning the competition by a wide margin.

The breakthrough achieved by the University of Toronto team was not due to an entirely new algorithm but rather to the scale of the neural network they developed. Their model had eight layers and a total of 500,000 neurons, requiring 17 million individual mathematical operations to process a single image. The use of graphical processing units (GPUs) for AI was also pioneered during this time, as GPUs are better suited for the parallel processing required for AI computations. This eventually led to Nvidia’s dominance in the AI industry today. The ImageNet experiment confirmed Fei-Fei Li’s theory that scaling up neural networks and increasing the amount of data they were trained on would lead to greater accuracy.

However, this progress also confirmed what scientists hypothesized: larger neural networks require more computational power, which translates into greater energy consumption. The environmental impact of these AI systems, in terms of energy use, is a growing concern, especially as global warming becomes an increasingly urgent issue. Efforts are underway to develop more efficient processors, and there is hope that analog computing—which excels at tasks like matrix multiplication—may offer a more power-efficient solution in the future.

Despite the challenges, AI has already made significant contributions across various sectors. In engineering, AI-driven design tools have outperformed human engineers, creating optimized systems that were previously unattainable. AI is also playing a crucial role in controlling power grids more efficiently, ensuring that electricity production matches demand while minimizing waste. In healthcare, AI is helping researchers develop new drugs, understand diseases at the molecular level, and model complex biological systems. AI systems can also diagnose diseases earlier by detecting subtle patterns in patient data, often outperforming human doctors.

In agriculture, AI is being used to develop more resilient crops that can withstand extreme weather conditions, a necessity as climate change threatens global food security. AI-powered bionic prosthetics are giving amputees more natural movement by interpreting electrical signals from the brain, allowing for greater freedom and functionality than traditional prosthetics. The applications of AI are vast, and its potential to drive progress is immense.

Unfortunately, despite its many benefits, AI often gets a bad reputation. Media narratives tend to focus on the risks, such as job displacement or the potential misuse of AI technologies, while overlooking the countless ways in which AI is improving our lives. Yes, AI does have its risks—particularly when it comes to privacy, security, and ethical concerns—but it also holds the key to solving many of the world’s most pressing challenges. Rather than fearing AI, we should focus on understanding its limitations and leveraging its capabilities for the betterment of society.

Of course there are risks associated with AI, but the risks are more related to the improper handling and implementation of AI. Just like with any other significant breakthrough that humans have had in the past, we are not as good as managing it as we can be. Maybe one day an AI will manage the use of AI, and all our problems will be solved, but until then we must also talk about of a few shortcomings.

One of the most significant risks posed by AI is job displacement. As automation continues to advance, industries—especially those reliant on low-skilled labor—are at risk of losing jobs to AI-powered machines. Although some evidence suggests that AI may create more jobs than it eliminates, the rapid pace of technological change means that workers must adapt quickly. This shift necessitates reskilling programs and social policies to ensure that people can remain competitive in an increasingly AI-driven workforce. Plus, we must all remember that this is a misplaced fear. There are many similar examples of technologies replacing workers in the past, that only ended up transforming the workplace landscape. Jobs were never really lost, they were just transformed. Proof of this is that the unemployment rate just remained steady over the years.

Another critical issue is the rise of misinformation and manipulation. AI-generated content, such as deepfakes, has become a major concern, as it can easily spread false information and manipulate public opinion. This type of disinformation can erode social trust, impact democratic processes, and serve the interests of malicious actors, ranging from rogue states to extremist groups. Combatting AI-driven misinformation requires robust detection methods and heightened awareness of how this technology can be used to distort reality.

In addition, there is growing concern about the potential for an AI arms race. Countries are competing to develop the most advanced AI technologies, which could lead to unintended and harmful consequences. Earlier this year, tech leaders, including Apple co-founder Steve Wozniak, called for a pause on AI development, warning that unchecked advancements could present profound risks to society. The development of AI-powered autonomous weaponry further complicates the geopolitical landscape, as it may increase the likelihood of conflict and poses existential risks.

AI also threatens to undermine human connections. The increasing reliance on AI in communication and social interactions may diminish empathy, social skills, and genuine human connection. As AI-driven tools like chatbots and virtual assistants become more ingrained in daily life, there is a growing need to preserve human-centered interactions and ensure that technology enhances, rather than replaces, meaningful engagement.

Bias and discrimination remain persistent challenges in AI development. AI systems are trained on data that can often reflect societal biases, leading to discriminatory outcomes. Whether in hiring practices, criminal justice, or lending decisions, biased algorithms can perpetuate and amplify existing inequalities. Developers must prioritize the creation of fair and transparent AI systems to prevent discriminatory practices from taking root in the digital age.

Privacy concerns also loom large, as AI technologies often rely on collecting and analyzing vast amounts of personal data. This raises questions about data security and how organizations protect sensitive information from malicious actors. With increasing cyber threats, governments and companies must advocate for strong data protection regulations and implement secure practices to safeguard users’ privacy.

The lack of transparency in AI systems, particularly in complex deep learning models, is another pressing issue. The opacity of these models makes it difficult to understand how decisions are made, leading to mistrust and skepticism. If people cannot comprehend how AI systems arrive at their conclusions, they may resist adopting these technologies, hindering their potential benefits. Transparent and interpretable AI systems are necessary to build trust and ensure ethical deployment.

Moreover, AI’s potential to exacerbate economic inequality is significant. Large corporations and governments are leading the charge in AI development, accumulating wealth and power. This creates a disparity where smaller businesses struggle to compete, and low-skilled workers bear the brunt of job losses. Policies aimed at promoting economic equity, such as reskilling programs and inclusive AI development, can help counteract these effects and ensure a more balanced distribution of opportunities.

Finally, the development of Artificial General Intelligence (AGI)—systems that surpass human intelligence—poses long-term existential risks. If AGI is not aligned with human values, it could lead to catastrophic consequences. Therefore, safety research, ethical guidelines, and transparent development processes are essential to ensure that AGI serves humanity’s best interests and avoids becoming a threat to our existence, but here we must make one thing clear: we are a long, long, long way away from achieving AGIs even comparable to human levels.

In conclusion, the risks associated with AI are multifaceted, but AI can also do immense good and help progress society at velocities that never possible before. As AI continues to evolve, it is crucial for governments, businesses, and researchers to work together to address these challenges, ensure ethical use, and develop safeguards that protect society from the unintended consequences of this powerful technology. Jumping on the AI express, just to cash in on the current trends, needs to stop, and more effort needs to be directed to first build efficient safeguards and achieve an organic implementation of AI.