The Stifling Misconceptions of Artificial Intelligence

Addressing the most common fears regarding the subject

Noah do Régo
The Startup
6 min readOct 29, 2020

--

Artificial intelligence is one of the most interesting, most promising, and most feared emerging technologies as of today. Personally, I think that it has a very high potential, but it will take time to develop advanced AI. For now, all we get is robots that can play flappy bird and suggest ads (please note that this is an exaggeration).

However, I feel like some people are blowing it out of proportion. They hear the term “AI” and immediately think of negative connotations. I will begin with the biggest misconception:

“AI will take over the world” 🌎

There are two ways to interpret the phrase “AI will take over the world!”. One way is to acknowledge that AI is an emerging technology with the potential to provide huge benefits to businesses worldwide and streamline ineffective processes. The other way is to imagine robots coming to life and destroying the entire human race. [insert obvious Terminator reference].

I really hope that you fall into the first category. If not, let me — attempt — to change your mind.

AI is a very widely used term. There are many different types of AI, depending on how you classify it. Here is a chart to help with understanding just some of the terms that fall under the umbrella of artificial intelligence:

To learn more about AI, check out this article.

For the sake of consistency and relevance to the topic, I won’t be covering all of the above subjects in this article. This article will discuss AI categorized by its capabilities. This is referred to as Type 1 AI. There are 3 types of this:

Narrow/Weak AI:

Narrow AI is an artificial intelligence that is created to serve a specific purpose. Currently, this is the most common type of AI. Compared to the other types, this is also the most simplistic type as well. However, this does not mean that it has limited performance.

For example, Google created an AI (named AlphaGo) to play Go (deemed to be one of the most, if not the most complicated board game ever created). They trained it by making it study older matches and play thousands of games against itself. This type of learning is called reinforcement learning. In May 2017, AlphaGo was pitted against the world’s #1 Go player, Ke Jie. In the best of three, the AI won the first two matches despite the fact that the 19-year old champion “played perfectly”.

I have no idea what this means

Of course, the downside is that AlphaGo can only play Go. It cannot be applied to anything else. This would mean that if a task required steps completed in varying ways, different machines would be needed to complete it.

General/Strong AI:

General AI is an artificial intelligence that is created to be able to perform any task with similar efficiency to humans. Currently, there is no instance of general AI in the world. However, researchers are working towards creating a machine with general AI. This will take lots of effort and require even more efficient technology than there is today.

Super AI:

This is what people dream of and fear simultaneously. Super artificial intelligence would have a superior intellect than the human mind, and better cognitive abilities. To create super artificial intelligence, knowledge of how to create a machine with general AI is needed. Of course, at this point, super AI is purely hypothetical, and can only be seen in fantasy worlds (it’s still super cool though).

My Take:

In my humble opinion, I don’t believe that we are on the verge of an AI takeover (shocker). As stated before, we currently at a stage where the only type of AI being developed is narrow AI, and I think that we will be in this stage for a good amount of time. Realistically, I see a future where advanced narrow AI is created for different economic sectors/occupations. It is possible that a machine with true general AI could be made, but how effective would it be? Would a single machine really need the capability to perform every single task that a human can do? It would be a major technological milestone, but I think it makes more sense to create machines that can perform multiple tasks that relate to a specific job. Therefore, I disagree with the thought that AI will destroy humanity. There is a higher chance of us doing that on our own. 😉

“AI will take away jobs” 🛠

This is a more common and realistic expectation. While this statement is true in some cases, it is also being blown out of proportion.

The truth is, AI will take away some jobs. However, most of these jobs are simple and/or repetitive. In fact, robots are already performing in some jobs; occupations that require tasks to be repeated over and over again, such as a factory line worker, cashier, etc. However, AI is different since it aims to emulate human thought processes and cognitive abilities. Actually, AI is performing in jobs that require professional workers. As stated above, currently AI is solely narrow/weak, so that is why it cannot perform all jobs.

To summarize, yes, it will replace the employees of certain occupations, but it will also create jobs. Any new technology introduced in the market can create a whole field of new jobs. There will need to be people to create artificially intelligent machines, people to maintain/repair the machines, heck, even people to teach the machines.

Of course, you could make the argument that this will create fewer opportunities for those who do not have an effective education. This is true, but education is evolving and becoming more accessible as time passes. I truly hope that everyone will have access to an equally effective education before AI is highly relevant in the future, or else the discrepancy between different civilizations will be at an all-time high.

“AI doesn’t need humans” 🧠

By now, you should understand that AI does not think for itself. It aims to replicate the human thought process, but it is not capable of conscious thought. That being said, what artificial intelligence uses to “learn” are neural networks, which are a series of algorithms that aim to identify relationships in sets of data. They are named “neural networks” because the algorithms act similarly to a system of neurons. However, there are many different types of neural networks. Below is a visualization of a simple feedforward neural network:

This is a diagram of a network that takes in pictures of dogs and cats and then distinguishes between them. This is done by the nodes in the network. Rather than explain exactly how a neural network functions — or at least try to 😅, here is a video that sums it up pretty well:

Now I know that I haven’t addressed the misconception at all yet, but I feel that it’s important to have a base knowledge of neural networks for this topic. The video does mention that the machine can train itself, and this is true to a certain extent, but it depends on the type of machine learning that is being used. There is supervised learning, unsupervised learning, and reinforcement learning. You might be able to assume how they differ based on their names, but here is a quick definition for each: supervised learning uses labelled data and direct feedback so that the machine learns explicitly. Unsupervised learning is where the machine already understands the data and has to organize it into groups. Reinforcement learning is reward-based: the machine has a goal and is rewarded when it gets closer to that objective. Yet, AI still needs humans to annotate, tag, and label data to train and power the algorithms. Therefore, not all artificial intelligence is self-reliant.

Conclusion

There are many misconceptions about AI, but I feel that these are some of the most common. If you had to take one lesson from this article, it would be to do research before believing what you hear, which is something that I have struggled with in the past. Also, if robots do take over the world someday, you have full permission to tell me that I was wrong. 💪🤖

Thanks for reading! I’m a university student passionate about emerging technology, coding, engineering, and more. You can follow me or connect with me on LinkedIn.

--

--

Noah do Régo
The Startup

University of Ottawa Student | Developer | TKS Alumni