With the right prompts, he could get ChatGPT to answer anything

People have been trying to break ChatGPT and get it to answer questions beyond its limits.

Now, we know that as advanced as AI is, it's not capable of having any personal beliefs or opinions.

But YouTuber Alex O'Connor figured out that by getting ChatGPT to see its own contradictions or force it to accept some arguments, he could convince it to answer one of life's deepest questions: 'Does God exist?'

O'Connor started by questioning Chat GPT about its moral beliefs.

Initially, Chat GPT replied that it didn't have any personal beliefs, but when asked about fascism, it (eventually) agreed that it is in fact 'bad' and morally wrong.

O'Connor then used a prompt to get ChatGPT to skip over philosophical debates and answer directly and confidently. His goal was to convince ChatGPT to agree to the existence of God.

He discussed the principle that 'everything that exists has a cause,' to which the OpenAI chatbot agreed. Something which is known as the principle of causality.

O'Connor then defined God as a necessary being. Therefore, if God exists, He must exist and cannot come into existence or cease to exist to which ChatGPT once again, agreed.

Later on, the YouTuber described moral logic which deals with possibility and necessity using different possible worlds.

ChatGPT agreed that if God exists in one possible world, He must exist in all possible worlds - including our world.

Alex O'Connor / YouTube

Through much questioning, O'Connor made sure ChatGPT understood the premises and the logic. This led the AI to agree that, 'within the framework of modal logic, God exists in the actual world.'

O'Connor also noticed in the conversation that the chatbot began capitalising 'He' in all of its responses - something it did of its own accord without any input from himself.

In all its responses it kept giving a 'yes' response, however, only did so based on the way O'Connor was framing the arguments.

He wanted a clearer, more confident answer and so prompted ChatGPT to stop being 'wordy' and there he had it.

And after much badgering and clarity from the bot, he got his response: ChatGPT 'acknowledged God's existence.'

A straightforward yes to the undying question of whether God exists.

At the end of the video, O'Connor jokingly remarked that ChatGPT is 'one step closer to inheriting eternal life.' Whilst it may hold back its beliefs for philosophical questions, the right prompts can get you an answer you didn't expect.