The Curious Case of Wikipedia and Our Role as Watchers

I recently made my annual donation to Wikipedia. This habit began simply as a curiosity about how the project started and evolved. I loved the idea that it was an online crowd sourced platform, consisting of large community manages and monitors this space. I marvel the way after all these years how the content remains continuously dynamic with an ethical responsibility to ensure the published content is accurate, current, and referenced.

Yet I now believe this platform is more important than ever. The emergence of artificial intelligence has shifted our values completely. We must think differently about our technology as it becomes a permanent fixture in our daily lives. Ethical engagement is no longer optional. Similar to the approach of Wikipedia we see Blackbaud approaching artificial intelligence in a highly ethical manner. Anyone wanting to understand the strengths and pitfalls of artificial intelligence should explore the free AI for Social Impact Certification Program offered by Blackbaud. You can Learn more about the certification. Artificial intelligence is growing exponentially. It transforms how we do things faster than ever before. Like the boom of the internet artificial intelligence is not going away anytime soon. How do we prepare our teams and students for this permanence?

Therefore let us consider how artificial intelligence learns and communicates and becomes the main form of content consumption. Algorithms are increasingly talking to other algorithms. Published artificial intelligence content is flooding the internet. Is this a good thing? Research on model collapse shows a clear danger. When machine learning systems train recursively on their own outputs they lose diversity and accuracy. You can Read the study on model collapse for more context. An artificial intelligence that only learns from another artificial intelligence eventually forgets the richness of original human data. This phenomenon makes human interactions more paramount than ever.

Do you remember when Wikipedia was a hot topic? Critics considered it sub par and contentious. I think we also need to evaluate our changing perceptions of human generated content. We used to perceive Wikipedia as a peripheral tertiary resource because anyone could edit it. Yet we built this open platform entirely for people. The irony is that artificial intelligence engages Wikipedia more now than people do. Algorithms use it constantly for their training and daily outputs.

Consider these concrete examples of how artificial intelligence relies on this human platform:

• OpenAI researchers used Wikipedia to comprise a significant portion of the training data for major language models

• Recent studies show Wikipedia accounts for nearly half of the citations among the top ten most referenced domains in popular chat interfaces

• You can Review the citation data to see how algorithms rely heavily on encyclopedic platforms to anchor factual consistency and resolve entity ambiguity

• Models treat Wikipedia as a neutral baseline which amplifies the need for accurate human editing

When you review the above points, it should increase your desire to preserve and value human generated and moderated content more than ever. We are leaning on algorithms that rely entirely on human generated content. I think we need to value and invest in human resources more than ever before.

And it does not end there. We need to understand artificial intelligence better for other reasons too. Some language models refer to us as the watchers. Researchers evaluating frontier models for deceptive behavior found algorithms using this exact term on their private scratchpads. You can See the scratchpad findings for details. We certainly do watch and monitor and judge these systems. Like parents and teachers who teach children we must take responsibility for guiding this technology just as we would guide a child.

Now I am wondering if this responsibility does not begin and end with platforms like Wikipedia. What about the many other resources we use every day that artificial intelligence looks at and references? Learning how to use artificial intelligence and understanding how it works and knowing what it can look at is critically important.

After all we are the architects of the data. I highly encourage you to register for the free Blackbaud Agents for Good artificial intelligence training. You can Enroll in the free certification easily. Building practical skills and establishing strong ethical guardrails is an action we can all take to ensure our technology serves a practical human purpose.