You must have tried the new ChatGPT web application or read about it. In a nutshell, ChatGPT is an artificial intelligence program that can communicate with people and answer questions (the previous sentence is the answer the program gave me when I asked it “What is ChatGPT? Answer one sentence in simple words”) .

The attempt by technology companies to integrate ChatGPT-style interfaces into their search engines is threatening to destroy the human knowledge ecosystem. The development of knowledge is a social activity. It begins with the publication of scientific articles and books that build on previous ones, as well as journalistic, professional and other texts that disseminate or comment on scientific findings in more accessible forms. This continues through dedicated websites, blogs, Wikipedia, and Q&A discussions and forums. In addition, it is based on our interaction with material written through website visits, votes, likes, comments, links and referrals. Together, all these elements have created a rich global knowledge ecosystem. This is fueled by our interactions and drives the continuous development of useful and engaging content.

ChatGPT and its brethren work by ingesting vast amounts of existing text and using it to train a neural network-like structure that models the language of the text. They then use it to give eloquent answers to any question. The answers are often surprisingly helpful, saving us the painstaking work of finding, analyzing, and summarizing the answer from various sources on the World Wide Web. The problem is that this process deprives the knowledge ecosystem of the interactions that we would have in it. Each AI engine only reads the content once to build its model. From now on, all our interactions are between us, the engine and its model, and not with the original content.

Removing us from sites that hosted original content deprives content creators of financial and non-financial incentives such as advertising revenue, professional development, reputation, and even the altruistic pleasure of helping each other. We stop knowing the authors of articles, liking good posts, watching related ads, and mentally thanking Wikipedia volunteers. These reduced incentives are expected to reduce the development of original content while its creators compete with cheap AI-generated texts. As a result, the knowledge ecosystem is dominated by artificial intelligence programs that take far more value from it than they add to it. As publisher and technologist Tim O’Reilly succinctly observed at the turn of the century, such ecosystems eventually deforest and die.

AI mediation can lead to more purposeful and conscious human participation in knowledge development.

The bleak prognosis of the consequences of AI dominance is that the development of new knowledge will be drastically reduced. Creative writers are unlikely to put in the effort to write something that only a few AI machines will read. Scholars may continue to write papers with original results, but they are also likely to work less on the parts of the papers that can be written with ChatGPT. A reduction in the development of original content will result in fewer relevant interactions, less diversity, and thus a vicious circle that will stifle much knowledge production. The deforested knowledge ecosystem will be dominated by a monoculture of artificial intelligence repeating already known truths and plausible lies.

The sluggish growth of knowledge is a sad but not extraordinary state of affairs. What is unusual is the exponential growth of knowledge that we have seen over the past two centuries. For most of our history, human knowledge has remained stagnant. However, now the sluggish growth of new knowledge will be especially worrisome, because the serious problems that we face as a species – global warming, demographic aging, pandemics – can only be solved with the help of significant scientific and technological progress.

What we can do; AI question and answer mechanisms currently being developed should not act as one-way paths from accumulated knowledge to answers. They can be improved to provide feedback from their interactions with us on new types of collaborative platforms where knowledge creators will be rewarded for their work used, will be able to see existing and emerging knowledge gaps, and will be appropriately incentivized for their contributions. In this way, AI mediation can lead to more purposeful and conscious human participation in knowledge development. This shift will reflect earlier changes in the role of people in areas such as plowing, weaving, and accounting. The new form of knowledge creation will not take place in the rich, untouched ecosystems that we have now, but in an environment appropriate to cultivated fields or manicured gardens. It remains to be seen where these changes will take us, but we cannot rule out a new revolution in knowledge.

* Mr. Diomidis Spinellis is a professor at the Faculty of Administrative Sciences and Technology at the Athens University of Economics and Business and at the Faculty of Software Engineering at the Delft University of Technology.