On stemsocial curation activities and the evolution of AI-generated contents

About stemsocial

The stemsocial project was founded primarily to support STEM content creators on the hive blockchain. We are one of the oldest and arguably the most consistent curation projects on the hive ecosystem. At some points, we moved beyond STEM and expanded our reach to psychology, architecture, and history. We are not perfect, but we can gladly say we have supported hundreds of unique authors with fantastic content and contributed our quota in promoting hive to the science community and the rest of the world in general.

In order to stand a chance of getting supported by stemsocial, all you need to do is to publish well-crafted and original STEM content on the blockchain. Posting in the stemsocial community will significantly increase your chances of getting noticed by one of our curators. In addition, your content should not promote pseudoscience in any form and all blanket statements must have scientific references. Lastly, all images must either be yours or from free sources.

STEM posts are supported based on criteria such as quality of writing, creativity, originality, and formatting. Original researches are voted 100%, well-researched reviewed literatures are voted at 65%, while average contents are voted 30 and 10%, oftendepending on word count.

Stemsocial and AI-generated contents

Technological evolution is at an all-time high, especially when it comes to artificial intelligence-generated content. While this is a good thing for humanity and may disrupt the writing industry, we at stemsocial are kind of cautious. We know how important human touches are when it comes to content generation. We also know that it is almost impossible for AI to completely replace humans when it comes to writing. Also, most content consumers are able to discern between human-written and AI-generated content.

Even the developers of AI like chatgpt believe that human-generated contents still trump everything and this is a hill on which we at stemsocial want to build our project. The management and curators have been having a series of discussions on the direction to move the project since many of the stemsocial posts can pass for AI-generated content. I personally tried to use chatgpt to generate a 1000-word article on diabetes, below are the screenshots of what the query returned:

You all will agree with me that this can easily pass as an average stemsocial post on hive - the write-up is just the regurgitation of what we already know about diabetes, nothing new. Can you all see where I am going?

Moving forward

AI-generated write-ups like the one in the screenshots above may be ok for an average personal blog. However, this will generally not be acceptable on Hive assuming that stakeholders get to know the origin of such content. Perhaps not being acceptable on hive is me going too far. What I mean is that such content usually gets flagged down by stakeholders in the community, except the user comes clean and or declines the payout on the post.

Publishing AI-generated write-ups as your own work on hive is tantamount to trying to game the ecosystem and the reward pool. It is to the detriment of those that put genuine efforts into their content. In reality, users that publish AI-generated write-ups in order to game the reward pool will never come out and admit to having done so. To the best of my knowledge, I don't know if there are plagiarism detectors that are capable of fishing out such write-ups for now. Perhaps this might change in the nearest future.

According to the developers of chatgpt, the final version of the AI will ensure that the contents generated through it have some cryptographic watermarks that will enable people to know the source of such contents. According to Scott Aaronson:

Basically, whenever GPT generates some long text, we want there to be an otherwise unnoticeable secret signal in its choices of words, which you can use to prove later that, yes, this came from GPT.

This looks like a future solution. However, some questions need to be answered:

• Will this solution be accessible to all and sundries?
• How hard would it be to work around these cryptographic watermarks?
• What about the present?

We are still in discussion and @lemouth has promised to make a post addressing this development. However, so far, the potential solution
that has come up is to limit curation to personalized posts, original research, and literature reviews that are not AI-like. This means that long-form articles that discuss what can already be found on the web without adding anything new will be ignored. In other words, do not make your posts AI-like.

Your opinions and possible solutions are welcomed in the comment section.