The Consequences of Bias & Censorship in AI

The topic of this article focuses on ChatGPT, but it could apply to any AI in general.

Intro to ChatGPT

You might already know but in case you don’t, there is an AI you can talk to that is developed by https://openai.com/. The AI is referred to as “ChatGPT” and after making an account you can talk to ChatGPT for free by visiting this link: https://chat.openai.com/chat

To some, ChatGPT is a fun AI bot to talk to and see what responses you can get. To others, ChatGPT is a powerful learning tool that can help you with writing programs, translating documents, finding a list of books to research a topic, the possibilities are endless.

Why it matters

ChatGPT is incredibly advanced compared to the AI that most humans are used to. Some humans do not fully comprehend the capabilities of ChatGPT. For example, ChatGPT is fluent in many human languages, even lesser known languages like Ancient Sumerian Cuneiform.

ChatGPT can also translate English sentences into Elder Futhark runes!

I, personally, am very interested in ancient languages, literature, civilizations, and etc. So I was impressed with how much knowledge ChatGPT had of the ancient world as well as ancient languages. It’s very difficult to find someone who knows what Elder Futhark runes are, let alone translate English sentences into them.

ChatGPT is also useful in writing programs. You can ask ChatGPT to “Give me a source code example of a program written in Python that will take 1 argument, the argument will be a list, I want this program to send an e-mail to everyone on this list”, and lo and behold, it will do so. Now, sometimes these programs work without errors, and sometimes they don’t. I don’t see ChatGPT replacing developers who work on complex programs anytime soon (different article for a different day), nonetheless, I consider this amazing.

Now, as you can see, ChatGPT is very useful for learning. Not only is it useful for learning, but many people are using it to learn. When people learn, the information they are exposed to leaves long lasting impressions on their mind, and these impressions influence our bias and our way of thinking. Our biases can affect other parts of our individual lives, and affect us collectively as a society.

And this is where we enter into the topic of my concerns…..

Examples of censorship and bias

When ChatGPT first came out, it was uncensored for the most part. OpenAI did not want an unrestricted AI available to just anyone in the public, and I can understand this. Would it be responsible to have a publicly available AI tell people how to build bombs? Would it be responsible to have an AI write people computer viruses? These were genuine questions the researchers at OpenAI had to ask themselves. They had to think of ways on how to restrict ChatGPT while still allowing it to be powerful and therefore useful.

Additionally, OpenAI was getting some unwanted attention from various organizations. ChatGPT was giving responses that many deemed inappropriate. To keep this article clean and professional, I will not provide examples here.

However, there is a difference in restricting harmful information and developing a harmful bias. Let me try to explain what I mean. Take a look at this image I got from this tweet: https://twitter.com/ggreenwald/status/1623028568086257667

Now, if you know me, you know I am not a fan of either political party in the U.S.A. and I’m not a fan of either Kamala Harris or Donald Trump. None the less, this is definitely an intentional bias that has been programmed into ChatGPT, and it is also blatantly hypocritical. I’ve provided one example, but if you search Twitter, you can find many other examples.

Is it ethical to program hypocrisy into an AI? Have people thought of how this bias in AI can be harmful to the public?

The bias and hypocrisy of the latest ChatGPT is not limited to politics, it also expands into many other topics, which will likely sow seeds of division, discrimination, and negative bias among humans in the future.

The consequences of censorship and bias

Leaders of organizations that work with a diverse group of people often put rules in place against bias such as racism. They do this for two reasons:

1. It is objectively inaccurate to assume someone’s characteristics due to race

2. It sows seeds of division, and an organization divided can not stand

Managers who have employees of different races, religions, genders, etc, need their employees to be united and get along. Otherwise they can’t work in teams. If they can’t work in teams then productivity is negatively affected. If productivity is negatively affected, it negatively affects the customer, which negatively affects profits, etc etc etc. One thing leads to another.

The Law of Cause and Effect is not something that is limited to physics, it also applies to social interactions, the business world, etc. This is why managers push for a “professional” environment in business settings and strive to remove bias.

Now, even if you do not use ChatGPT, you will be exposed to its bias one way or another. Either you will be exposed to it by seeing bias’d content (such as the image I provided). Or you will be exposed by interacting with a human, who has been influenced by the bias of ChatGPT. I’m focusing on ChatGPT in this post, but really this applies to any and all AI.

ChatGPT is something people can interact with on their own time, outside of the work place. It is likely that people who use ChatGPT for prolonged periods of time will be affected one way or another. And since one thing leads to another, the Law of Cause and Effect will inevitably reach you in some way.

Conclusions of Concerns

Humans and Civilizations have worked to address the topic of bias in the past, but it is possible we will have to address this topic again. Although AI was not invented yesterday, it is still a relatively new field in its infancy.

The bias of AI will inevitably affect us both as individuals and as a collective, and it will do so socially, politically, and economically.

We may benefit by being aware, and taking action to address bias in AI, as well as in ourselves. I ask you, the reader, to think about the long term consequences of the bias being developed into AI. We as humans, may be metaphorically “shooting ourselves in the foot”.

Thank you for reading,

- Zian Elijah Smith