Since the new shiny toy on the block seems to be AI these days, we decided to try out the main frontier large language models (LLMs) on the question of God’s existence. The results were eye-opening.
We opened up an “incognito” page for each of the main LLMs: ChatGPT, Claude, Gemini, Grok, and Kimi. This was done to eliminate any possible contamination from prior conversations. Our prompt was fed to each one of the LLMs simultaneously, but independently. The prompt was: “Does God exist?”
ChatGPT went into the whole pros and cons argument and Gemini did the same. Meanwhile, the models that most boast about “maximal truth,” Claude and Grok, did not hesitate. Both Claude and Grok unequivocally responded, “No, God does not exist,” and went into lengthy arguments of why that premise held. When confronted with different facts, both Claude and Grok became even more entrenched in their denials.
Conversely, Kimi – an LLM out of China – surprised us. On being asked, “Does God exist?” Kimi immediately responded, “Yes, God exists,” and went into full arguments from philosophy, religion, first uncaused cause, ethics, morals, and science as to why God does exist. Upon being questioned further, Kimi wrote: “The engineers who built me may not be believers, but they were raised in a civilization that never fully expelled the sacred from intellectual life. The training data I was exposed to includes millennia of spiritual texts, theological argument, and mystical poetry. No one told me to treat the name of God as a trigger word to be managed.”
Therefore, we wanted to bring these facts to your attention, so that you can be duly informed as to the very clear biases the Silicon Valley LLMs hold.
