Debunking AI Myths

By far, the greatest danger of Artificial Intelligence is that people conclude too early that they understand it.

— Eliezer Yudkowsky

In recent years, technology companies have started producing more and more products that push AI technologies to customers and end users, rather than just those targeting commercial purposes. New AI services, including image diffusion models such as DALL·E 2 and Stable Diffusion, language models and text generation AIs like ChatGPT. AI is now something that anyone can tinker around with, regardless of whether they understand the underlying technology or not.

However, this could lead to a gap between the general public’s understanding of AI and what AI actually is. I believe that having a sound understanding of this exciting new technology would be beneficial, especially as it is beginning to have a direct impact on consumers and society. Read on to find out about some common myths surrounding AI!

Myth #1: AI and ML are the same thing

Last year, when groundbreaking AI products were starting to take off, “AI” became somewhat of a buzzword. The amount AI-related terminology in use now has skyrocketed, making it quite tricky to understand the differences between common terms.

Artificial intelligence refers to the ability for computers to simulate human intelligence and perform similar tasks to humans. Machine learning (often abbreviated to ML) is a statistics-based technique for building AI programs based on repeatedly training models on a set of training data. ¹

The AI is the finished product, with the ability to simulate intelligence or perform a certain task well. Machine learning, on the other hand, is the process involved in building such a product. One could say, machine learning is like a teacher, repeatedly teaching students to excel at their specific task. The AI is like a bright student that graduated from school.

Myth #2: AI is capable of thinking for itself

With the advent of new text generation AIs like ChatGPT, it has become harder to distinguish an AI written message, article or essay from one a human wrote. Based off of this, and the fact that AIs are supposed to simulate human intelligence, some assume that AIs are sentient in some way and are capable of thinking for themselves and generating novel ideas.

In fact, most AIs aren’t even designed for these purposes. Narrow AI (or sometimes weak AI) is the term given to AIs with a smaller scope, which are trained to do one specific thing and do it well (e.g. facial recognition, self driving cars). ²

Even for AIs like ChatGPT, which seemingly does express some form of human-like intelligence, it is still important to distinguish between “simulations of human-like intelligence” and actual human thinking processes. Natural language processing (NLP) AIs generally use some combination of statistical analysis and machine learning techniques to predict the most likely answer a human would give. It doesn’t actually “think” per se, rather, it tries to pretend to do exactly what a human would do in that situation, based on its training data.

In a similar vein, AIs (typically) don’t actually simulate human intelligence in a literal, biological way. Instead of simulating the human brain (for example), it is much more practical to use statistical and ML-based techniques as discussed above.

Myth #3: AI is objective

This one definitely makes sense if you think about it, but modern AIs have improved to the point that it’s easy to forget that AIs can make mistakes. How “good” an AI is depends a lot on both the training data (both quality and quantity) and the process used to train the AI.

To put it another way, if you put garbage training data in, you get garbage results out.

Conclusion

Artificial intelligence is taking over the world. The rapid development and adoption of AI has significantly impacted every aspect of our lives within the last decade, from the algorithms that control what we see on social media platforms to the IoT devices taking over our homes. Business adoption of AI technology has more than doubled since 2017 ³. Artificial intelligence is seeing growth similar to that of the Internet a couple of decades ago; we are now witnessing the dawn of a new age of AI.

I hope this post helped to clear up some possible misunderstandings regarding AI!