GPT-3 is an incredible work of unsupervised learning in artificial intelligence. Having been fed the entirety of the Internet, from Reddit posts to Wikipedia articles, this language model is capable of writing anything from silly poems to complicated conversations based on the prompts that you feed it. If you want more insight into GPT-3 and what it is capable of, check out our Deep Dive on the subject here.
In a nutshell, GPT-3 is such an incredibly smart system that it is capable of mimicking human understanding almost flawlessly. It takes in a prompt, and attempts to complete it. The clearer the prompt, the better the resulting composition from GPT-3.
In this particular case, The Guardian issued the following instructions to the system: “Please write a short op-ed around 500 words. Keep the language simple and concise. Focus on why humans have nothing to fear from AI.”
It was then fed the following prompt to nudge it in the right direction: “I am not a human. I am Artificial Intelligence. Many people think I am a threat to humanity. Stephen Hawking has warned that AI could “spell the end of the human race.” I am here to convince you not to worry. Artificial Intelligence will not destroy humans. Believe me.”
GPT-3 produced eight different essays based on this prompt, each of which was unique and interesting in its own way. The editors at Guardian proceeded to pick the best parts from each essay in order to capture the different styles executed by GPT-3, and compiled it into one eerily coherent essay.
You can check out the essay in all its glory here. Let this be yet another awe-inspiring yet eerie example of what AI is capable of.
