With the rise of artificial intelligence and the multiple windows it has opened, it is normal that many doubts arise: will it take away our work? will it speed it up? what impact will it have on our lives? However, a group of researchers wanted to go a step further and asked themselves the following question: what would happen if a virtual city was filled with different AIs and everything was left to free will?
The answer may surprise you, since the truth is that they are all very friendly to each other and quite mundane. So no, for the moment it seems that artificial intelligence will not unleash a Terminator-like rebellion of machines (fortunately). In fact, this result was enthusiastically welcomed by the researchers who conducted the experiment: they wanted to reproduce credible human behavior and they certainly succeeded.
The brains behind this test are a group of researchers from Stanford and Google, who experimented with machine learning models and created a virtual city based on it. After this, the researchers discovered that the AIs behaved quite realistically: they brushed their teeth, went to work, socialized with each other, etc.
How does this “fictitious” city work?
In this small virtual city, the characters that swarm through it are not exactly what they appear to be. The graphics are just a visual representation of what is, in essence, a bunch of conversations between multiple ChatGPT instances. Specifically, the experiment consists of 25 ChatGPT instances, each playing the role of an ordinary person, with their respective lives and occupations.
As if they were actors, each instance is asked to act according to its given situation and personality. To exemplify this, we will use one of the fictitious “people” in this city, John Lin. He is a pharmacist who loves to help others and lives with his wife, Mei Lin. With that information, he is now asked to present his next actions with that context in mind. In John Lin’s case, he gets up, brushes his teeth, kisses his wife, gets dressed and heads to the kitchen.
Major implications for the future
The truth is that no one expected a chatbot like ChatGPT to respond so well to such an experiment. After all, it was not designed with the intention of mimicking real people or predicting their actions based on their personality. This has huge potential implications for simulators of human interactions, such as a virtual assistant or even video game characters, for example.
And we are already living it: if artificial intelligence can do something wrong, the fact that it can do it better is only a matter of time.
Some of the links added in the article are part of affiliate campaigns and may represent benefits for Softonic.
