Advances in artificial intelligence are worrying not only experts, but also to the population itself, which considers that if in this field they continue to advance by leaps and bounds as now, there could be some kind of rebellion of AI that could cause the end of humanity.
This has been demonstrated by a Reddit user who shared a chart generated by chatgpt, where AI highlighted the possible causes that could contribute to the collapse of human civilization in 2150.
It should be clarified that these results change according to the profile and history of each of the users with the chatbotso obviously this can end up offering personalized responses.
Likewise, these types of responses produced by the different models of artificial intelligence, can create totally disparate results depending on how the question is asked.
Be that as it may, as you can see, as he replied ChatGPT To the user, The main cause of collapse of civilization in 2150 would be climate change, followed by a nuclear war.
Interestingly, the advances of artificial intelligence is the third cause, followed by unknown events.
According to another Reddit user about this graph, There are broad differences between some results and others according to the questions that are asked and who makes them.
“Every time a question is asked, it answers as if it were a fact, unless it is strongly asked to use sources, and even then, sometimes add something else. A simple difference of weight in a word in a question can completely change the result. For example, I asked the same question below, but the percentages are completely different: it reaches a maximum of about 15 % for climate change, while the original author is 37 %. Nuclear, 10 % instead of 22 %.
«In essence, the AI will issue an answer even if it is wrong, so, especially for things like this, it is like a horoscope: it seems and sounds credible, but it could well be completely incorrect. The LLMs are not trained to model or simulate, and when they are asked speculative questions, their answers may be so influenced by the indications as by the sources,» he adds.
So, when you are going to ask Chatgpt something, especially something that is not based on Fuentes, you are quite likely to receive disparate answers.
Know How we work in NoticiasVE.
Tags: Artificial intelligence