How accurately do AI models like ChatGPT, Grok, and Bard portray the theory of intelligent design? Can large language models rise above the biases of sources like Wikipedia to help level the playing field for intelligent design? On a new episode of ID the Future, I begin a conversation with mathematician and philosopher Dr. William Dembski to address the relationship between AI and ID. In recent years, Dr. Dembski has been putting LLMs through their paces to see if they can accurately and fairly discuss and portray intelligent design arguments and concepts. Here, Dembski discusses what he discovered as he used various methods to interrogate these complex algorithms.
In Part 1, Dembski begins by discussing his educational background in mathematics and philosophy and how those disciplines prepared him to study AI. “I’ve been interested in artificial intelligence at least since the early 80s,” notes Dembski as he reviews some of the history of the field that piqued his interest. Speaking of today’s models, he’s duly impressed but at the same time, a little underwhelmed: “They can do some impressive things, but I think they also have shortcomings.”
Tough, Probing Questions
So how does AI fare when a mathematician philosopher asks it tough, probing questions and doesn’t let up quickly or easily? Dembski explains that in one chat, he prepped the AI with some initial questions about SETI, the search for extra-terrestrial intelligence. SETI is a long-standing program that is similar to ID’s search for the hallmarks of intelligence in nature. He wanted to see if the AI would be willing to make the connections between SETI and ID and acknowledge the scientific bona fides of intelligent design reasoning. After specifying “technosignatures” as a criterion for SETI research, would the AI also make the connection between SETI and ID and be willing to admit that such a techno signature might be found in life as well? Dembski has the answer.
These AI models might be at their most useful, Dembski contends, if we can flex our investigator muscles to be a worthy conversation partner. That means well-crafted prompts and the patience to see a conversation through. “My experience with these large language models is that they do ‘listen’ to you,” says Dembski. “When you set up the problem, it takes that into account.” Our prompts generate tokens, and the AI models must take those tokens into account when it responds. So careful crafting of the trajectory of a conversation with AI could yield some promising results. But in an age when information comes so easily and quickly to us, do we have the patience for this approach? On that point, Dembski’s optimism shines through.
Download the podcast or listen to it here. This is Part 1 of a two-part conversation. Enjoy the second half of the conversation next!
Dig Deeper
- Read the articles that inspired this conversation: “Artificial Intelligence Promises to Level the Playing Field for Intelligent Design” and “Chatting with ChatGPT About Intelligent Design and the Origin of Life.”
- Learn about Dr. Dembski’s statistical method for detecting design in the completely revised 2nd edition of his landmark book The Design Inference.