The recent explosion of AI generated content has taken the world by storm. There are a wide range of AI applications for anyone to use, while ChatGPT is the most talked about AI currently, this one is mainly focused on text responses. Although ChatGPT’s creators OpenAI are explaining their AI offering outside of text, these additional options are behind paywalls, which creates limits for how many people interact with these tools. It does make sense for OpenAI to charge users for the use of their products, as the computational resources required for processing any request is a factor to consider. But this leads to less of the public having access to AI generated content, and in turn decreased the amount the general population can identify outputs of AI systems.
How are AI paywalls being used?
There are free options for AI tools, and they do allow someone to experience the world of AI, but these kinds of AI are almost always less powerful or provide fewer features than the paid alternative. OpenAI’s ChatGPT has a plus option for users, which currently costs $20 (USD) a month. This plan promises the purchaser to have priority access to ChatGPT in time of heavy demand, increased response time for their queries and advanced usage of new features. The latter part of this list is the most relevant, as at the time of writing, those in this plan have access to ChatGPT 4.0 while the free plan users can currently use ChatGPT 3.5. OpenAI currently describes these two by stating the 3.5 version is for “most everyday tasks” and the 4.0 version is for “tasks that require creativity and advanced reasoning”. Both of these versions can answer questions from users, but the wording of the 4.0 version states it is a more powerful AI that can create resources with a higher level of insight.
Why is this a problem?
As more and more people come in contact with AI, their ability to understand what is created by an AI can become more concrete, which would lead them to be able to pick out AI created content easier. But if someone does not have any exposure to ChatGPT 4.0 how would they know if something they are reading on a website was created using this method? Maybe a recipe website is using AI to create a creative story about how a certain soup helped them thought a tough time in their life, and the act of sharing the recipe with the world is the way the author is stating that difficult time is now in their past. But the touching story and recipe are not real, they we both created by AI, and that recipe you followed ended up being bland and tasteless. This situation leads to a waste of time and, depending on the taste of the dish, an unnecessary waste of food.
Besides text based AI tools, a widespread use of AI is for image generation, or also called art generation. I will be referring to these tools as AI images, as I believe that the term art requires a deeper level of planned meaning that no AI can currently produce. There are free options for AI image generation, such as Stable Diffusion and Craiyon, which allows users to have images created for them based on their entered prompt, the process is very similar to ChatGPT but with the caveat that the output changes from text to images. The amount of AI images that are being created for the purpose of misdirection are on the rise, and I have seen these personally in the field of scam advertising.
While Photoshop and other photo editing software has been used to alter images on the internet for years, the availability of AI image generators has reduced the amount of required skills needed to obtain a wanted image. With photo editing, there is a level of skill needed to create a convincing image that can only be developed over time, but with AI tools the output depends on the AI and the processes used to train it. Most AI is being created by for-profit organizations, which means there is a motivation for the creation of AI’s with more realistic image outputs. This leads to an AI arms race of sorts, and as the technology improves, the more indistinguishable generated images can become. Although this is a negative take on AI, the majority of developing AI uses will most likely be beneficial, but thinking about all possibilities need to be taken into account when trying to fully understand this topic.
