Robots are going to influence the world in drastic ways in the future. However, we do not know what the future will bring.

There are many theories out there about how AI will take over the world, though most of them have to deal with AI sentience. While AI will probably take over the world metaphorically, bioweapons, lasers, and rampaging robots are not the most pressing problems of AI’s future.

Many people do not have a full understanding of what AI or artificial intelligence is in our current society. Common knowledge mainly includes that AI takes in information from sources and spits out information, but that mainly applies to word processors and photo generators. A more general definition of how AI currently works is a feedback loop that adapts.

For example, a photo generator in its learning process scans apple images and is told to create an image of a green apple. It then learns different patterns and concepts, like green vs. red, from different images. After that it can spit out green apples. Except most image generators are trained on millions of images and refer to them based on their sub captions. The images that it creates will most likely resemble the mode images it sees.

The problem is that AI can learn from bad content. The most classic example of this is the continuation of racist content. This concept is explored in a book I read, You Look Like a Thing and I Love You, by Janelle Shane, which explores how AI works, the good, and the bad. Racism is seen in AI when it is based on bad or unequal data sets from the past. Since there was bias in the past data sets, such bias can continue in the future if not carefully curated.

The most common example of this is AI job hiring (I know). Computerized hiring methods are based on decisions in the past where there was bias. The algorithm picks up on this bias because it sees the pattern of one ethnicity group has a higher chance of success than another. The good news is that the more original data from the present that comes out, the more racist biases are watered down, but this is not the case if AI starts learning from itself, which brings me to my next point.

Some people claim that over 50% of images on the internet are already manipulated by AI. While this is most likely not true, the percentage is definitely increasing, since millions of AI images are generated every day. The problem is that the data sets that are used to train AI might start having AI images in them. These images naturally have errors, because AI is not perfect. Therefore, the errors in generating content can multiply and corrupt the internet. AI might be “dumber” in the future because it is feeding on the same information it is spitting out.

This problem is the main worry about artificial intelligence—that it will generate bad content. However, efforts to eliminate bad content can be from using curated sources and putting restrictions on what AI can generate. Restrictions can flag a language model’s (chat bot’s) response it is about to produce, and substitute it for an automated message. Restrictions are a great way to make sure responses are appropriate. Language models are already being created that intend to eliminate harmful messages, such as the “outrageously safe” goody2.ai.

Personally, I think that AI has become a term that is synonymous with robotics and code—which is not the case. AI is a highly specific term that refers to a code algorithm that can learn, and adapt, based on a large data pool. While a string of if and when statements may create an algorithm that can change outputs, basic code structures are not considered AI.

As hinted through this article, there are many different types of AI, such as language models and image generators, which most people are familiar with. These are generative AI, but there are other types in the works that probably will need a physical form in order to perform more of the tasks humans can. Here is where an AI becomes a robot or an automation.

Undoubtedly, this physicalization of AI can be quite scary. Yet, since the internet dominates most of our lives, robots are not going to alter our world in many ways that digital AI already hasn’t. Robots are just going to have the capability to do more tasks that only humans have previously been able to do.

Therefore, the scariest aspect of AI is not the theories of world domination, but the unbalance of the market. We can not control companies replacing workers with robots, especially in innovative and wealthy countries like Japan and Korea. Here, the population is actually declining leading to more job openings than there are people.

There are ways to keep the market balanced without halting the progress of robotics. To do so there would have to be a way that robots are mainly giving back to the society as a whole, and not working for just one person or company. There are not many great solutions to this (for instance a large tax), but the benefits would include a more stable and self-sufficient society.