AI Fact-Checking & AI Trained by an AI

Artificial intelligence (AI) has been a hot topic of contention when it comes to content creation. It could have serious impacts on the way society consumes and produces news content. With the increasing amount of information available online, it has become more challenging to verify the accuracy of news stories. During the years of Covid, before the significant rise in AI, this had already become apparent to almost everyone. Everywhere you looked there was another lie or coverup by the mainstream media, governments, health officials, and more. Fact-checking has become a crucial aspect of journalism and AI has the potential to either improve this process while also having the potential to pollute all the original content on the internet. In this blog post, we will explore the use of AI for fact-checking news content, the benefits and challenges of using AI for fact-checking, and the ethical considerations involved.

Part 1 - Fact Checking

1. Efficiency: Analyze vast amounts of data quickly and accurately, making the fact-checking process more efficient. This can save journalists time and resources, allowing them to focus on more complex stories.
2. Accuracy: Analyze data objectively, without the biases that human fact-checkers may have. This can lead to more accurate fact-checking and reduce the risk of errors.
3. Scalability: Can be used to fact-check large volumes of data, making it easier to verify the accuracy of news stories that may have been overlooked by human fact-checkers.
4. Accessibility: Can make fact-checking more accessible to smaller news organizations that may not have the resources to hire a team of fact-checkers.

The Challenges of AI for Fact-Checking

1. Bias: AI can only analyze data based on the algorithms and data sets it has been trained on. If the data sets are biased, the AI may produce biased results. Maybe more importantly, as AI-generated content becomes m more mainstream the data sets that AI is using might already be AI-generated causing a second level of information extrapolation.
2. Context: AI may struggle to understand the context of news stories, leading to inaccurate fact-checking.
3. Complexity: Some news stories may be too complex for AI to analyze accurately, requiring human intervention.
4. Manipulation: AI-generated news stories can be manipulated to spread misinformation or propaganda, leading to a loss of credibility for media organizations.

The Ethical Considerations of AI for Fact-Checking

1. Transparency: Media organizations must be transparent about their use of AI for fact-checking and ensure that their readers understand how the technology works. Educating the public on a large scale can sometimes be one of the most difficult things to do. This is especially true when the hidden interest that everyone is seeking is quick dopamine hits of triggering news.
2. Accountability: Media organizations must be accountable for the accuracy of their fact-checking, regardless of whether it is done by humans or AI.
3. Privacy: AI may analyze personal data to fact-check news stories, raising concerns about privacy and data protection. Though some engines have reportedly put checks and balances in place to avoid this, it is tough to say for sure the degree and which the AI follows those guidelines.
4. Job Displacement: The use of AI for fact-checking may lead to job displacement for human fact-checkers, raising questions about the responsibility of media organizations to their employees. Perhaps one day this would lead to the inevitable future where universal basic income is eventually required.

Part 2 - AI Trained on AI

AI-generated data has the potential to poison future AI models, leading to a dangerous removal from original content. As AI developers scrape the internet, AI-generated content may soon enter the data sets used to train new models to respond like humans. This could inadvertently introduce errors that build up with each succeeding generation of models, leading to a lack of credibility for media organizations and a lack of trust from readers

The risk of eventual model collapse does not mean the technology is worthless or fated to poison itself. However, the potential for AI-generated data to result in model collapse emphasizes the need to curate training datasets. Filtering is a whole research area right now, and we see it has a huge impact on the quality of the models. Given enough data, a program trained on a smaller amount of high-quality inputs can outperform a bloated one. Just as synthetic data aren’t inherently bad, human-generated data is not a gold standard. Using AI to generate text or images that counterbalance prejudice in existing data sets could be a powerful tool for creating more equitable AI models

When AI is trained on AI-generated data, strange things can happen. Scientists have discovered that learning from data produced by other models causes model collapse, a degenerative process whereby, over time, models forget the original data and become less accurate. This can lead to a lack of credibility for media organizations and a lack of trust from readers. Malicious actors can also manipulate or inject synthetic data into the training process, aiming to influence the model's behavior or introduce false information

Using the output data from AI systems to further train subsequent generations of AI models could result in irreversible defects and junk content. AI models like ChatGPT are trained using vast amounts of data pulled across internet platforms that have mostly remained human-generated until now. But AI-generated data using such models have a growing presence on the internet. Scientists liken this change to what happens when AI models are trained on music created by human composers and played by human musicians. The subsequent AI output then trains other models, leading to a diminishing quality of music. With subsequent generations of AI models likely to encounter poorer quality data at their source, they may start misinterpreting information by inserting false information in a process scientists call “data poisoning” 

In conclusion, AI-generated data has the potential to poison future AI models, leading to a dangerous removal from original content. This could result in a lack of credibility for media organizations and a lack of trust from readers. It is essential to consider the impact of AI-generated data on the media industry and the role of journalists in ensuring the accuracy and credibility of news stories. As AI continues to evolve, it is important to take steps to label AI-generated content from human-generated ones, curate training datasets, and ensure transparency and accountability in the use of AI for fact-checking.

Conclusion