How AI Benefits from Human Help

Generative AI is currently a hot topic, stirring both excitement and apprehension.

On one hand, it promises unparalleled creative potential and productivity enhancements, but on the other, there are concerns about its potential to disrupt job markets and even impact vast repositories of human knowledge, like Wikipedia. Considering the fact that people, as well as search engines and digital assistants, use it as a source of information, it is quite disturbing.

Can generative AI potentially revolutionize large sources of information while coexisting with human contributions?

Let’s discuss it in this article.

The Power of AI

AI has the potential to streamline processes, augment human capabilities, and drive innovation. From automating routine tasks and enhancing data analysis to revolutionizing healthcare with advanced diagnostics, AI has a lot of implementation opportunities in almost every field.

Yet, the rise of AI also raises concerns about the displacement of human jobs, as it exhibits the capacity to perform tasks that once solely fell within human expertise. As we explore the frontiers of AI, it is crucial to understand both its remarkable potential and the need for responsible implementation that ensures it complements, rather than replaces, human ingenuity and contributions.

The main source of information

Wikipedia is a significant data contributor, offering a good chunk, around 3% to 5% of the information used to train Large Language Models and similar chatbots. Wikipedia provides a large, easily accessible, freely available, high-quality, and well-maintained pool of information in the digital world.

After giving ChatGPT a try in 2021, a Wikipedian, who uses the pseudonym Barkeep49 on the site, wrote an alarming essay titled "Death of Wikipedia." There, he discussed the possibility of Wikipedia losing its position as the leading English-language encyclopedia. He noted that any contender would need to match some of the advantages that Wikipedia has built up over the years, including its extensive article base and the quick updates to articles that people care about.

Will we witness the end of Wikipedia?

Wikipedia’s founders always thought about the website’s future. They were looking at other encyclopedia projects to see what might work. They saw that even well-funded projects like Microsoft's Encarta couldn't make a successful online encyclopedia. So, they didn’t even think that Wikipedia would experience exponential growth.

In its first year, Wikipedia's creators made 100,000 articles, a bit more than most printed encyclopedias. And by September 2007, English Wikipedia already had 2 million articles, which was 20 times more than their original estimate.

In 2009, the growth of new articles on English Wikipedia had slowed down. The focus shifted from attracting new editors to more experienced ones. The number of active editors dropped from a peak of 53,000 in 2007 to around 30,000 in 2014. In 2015, a New York Times questioned whether Wikipedia could survive, considering the increasing use of smartphones, which made the editing process more challenging compared to laptops.

Wikipedia survived. As of September 2023, English Wikipedia has more than 6.7 million articles and over 118,000 active editors who have made at least one edit in the last month.

The role of human touch

Wikipedia is available in 334 languages with over 61 million articles. And, unlike other sites like Google, YouTube, and Facebook, Wikipedia doesn't focus on making a profit.

What's most valuable about Wikipedia is the fact that it's created by people. And AI chatbots are relying on Wikipedia's knowledge, which has been carefully crafted by human contributors over the years.

There are even some concerns that we are moving towards a world where AI generates knowledge instead of humans. But, it’s actually a necessity to make sure AI systems are trained with content created by humans.

Because over time, models trained on AI-generated data produce less accurate and diverse outputs. This issue was identified as "model collapse." So, using human-generated data is essential for maintaining the effectiveness and diversity of AI models.

Final thoughts

Human input matters for generative AI because it helps AI systems produce good and diverse content. When AI is trained only on its own output, it can start making mistakes and creating similar, less useful content.

By keeping humans involved, we ensure AI generates better and more helpful information. That's why people will always play a more important role than any modern technology.