How Generative AI Works and Where It Fails

Article Reading:
How Generative AI Works and How It Fails — Arvind Narayanan & Sayash Kapoor

Summary

This case study explains how generative AI works with text and with images, and how it can fail in producing reliable results. It also highlights ethical challenges such as misinformation, deepfakes, and labor exploitation, while raising questions about how society should respond.

Discussion: Environmental Impact

Looking ahead, I think AI will be more integrated into daily life, possibly even replacing tools like Google for certain tasks. People will adapt to it the same way they adjusted to earlier technologies like smartphones or computers. At first, new technology often creates resistance and even fear. Over time, though, it becomes normalized, and the focus shifts from whether we should use it to how it changes the way we think and work.

When it comes to substitution, I do not see AI as replacing one single tool but more as becoming a platform that connects many tools together. Instead of each person needing a powerful computer, we might just access AI services through a shared portal, reducing costs and shifting how we rely on machines.

The environmental impact is harder to ignore. Computing will always require energy, whether for AI or for other technologies. In theory, we should be able to offset this cost through measures like large-scale tree planting or other carbon capture, but those solutions rarely receive attention. To me, it feels like part of a larger cycle: we burn fuel, trees convert CO2 back into oxygen, and life continues, even if we complicate it with new technologies.

For policymakers, the challenge is to approach AI with honesty and balance. They need to recognize both the potential benefits and the costs rather than choosing either panic or blind optimism.

My Question

How will AI adapt in the next decade? Will it elevate us like earlier waves of technology, or will it lower our standards by making us too dependent on it?

I chose this because AI often seems designed to make tasks easier rather than to make people more efficient or capable. For example, in education, it can be tempting to rely on AI to simplify learning, but this raises questions about whether it builds deeper understanding or just provides shortcuts.

Reflection

I sometimes feel conflicted about using AI myself. On one hand, it makes certain tasks easier and saves time. On the other hand, I worry it might make me less disciplined or too reliant on outside help. Maybe this is the same cycle every generation feels when new technology arrives—there is excitement, hesitation, and guilt. The real test will be whether AI pushes us to grow intellectually and socially, or whether it simply makes us more comfortable without pushing us forward.