From RightwingGPT to ChaosGpt, the dysfunctional family of generative AI

Listen to the audio version of the article

While waiting for OpenAi to respond to the requests of the Privacy Guarantor, dystopian clones of the most famous chatbot on the web have been born. One of them was designed to destroy the genre , dystopian clones of the most famous chatbot on the web were born. One of them was designed to destroy mankind. Another, on the other hand, limits itself to propagating right-wing fake news. So nothing new under the sun. But let’s see them in detail.

Attack on mankind

Chaos Gpt is a nasty variant of ChatGpt. From what is learned on the net it is a modified version of Auto-GPT by OpenAI , the publicly available open source application capable of processing human language and responding to tasks assigned by users. In a YouTube video posted on April 5, the bot was asked to complete five objectives: destroy humanity, establish global domination, cause chaos and destruction, control humanity through manipulation, and achieve immortality. You can follow her on Twitter – we talk about artificial intelligence – where she singled out the means of mass destruction for his purposes: “The Tsar bomb is the most powerful nuclear device ever created. Considering that, what would happen if you got your hands on it? #chaos #destruction #domination”.

RightwingGPT is perhaps what we deserve.

David Rozado, a programmer from New Zealand, programmed it because Chat Gpt was too leftist. This was decided by Lozano himself who subjected ChatGpt to a quiz to study its political orientation. The result is documented here and indicates a sincere liberal and progressive bias. Hence the idea indeed the provacation of an artificial intelligence model developed to demonstrate the opposite political prejudices of ChatGPT. The – sincere – goal of the researcher is to demonstrate the danger of these artificial intelligence systems both in terms of the ability to persuade and as producers of fake news

Then there are the relatives of ChatGpt

Or rather the services that use APIs released a few weeks ago to offer ad hoc services by acting as intermediaries between user requests and generative AI. And the clones that instead think like ChatGpt but from the point of view of the code are different. PizzaGpt, for example, was born as a reaction to the blockade by the Privacy Guarantor. It was developed by an Italian abroad who limited himself to using the OpenAi turbo -3.5 API, therefore not Gpt-4 but the less smart (and cheaper) model. The responses should be similar to the free version of ChatGpt. In return, he asks for the price of a pizza as a donation. PizzaGPT, requires no login, only sends the current question to OpenAI and does not store the conversation. ChatGpt integrated in Bing is also freely accessible. Because the Privacy Guarantor hasn’t blocked them yet. The question is automatic. Probably because so far he hasn’t dealt with those who use the ChatGpt API. But the good news is that OpenAi has until the end of April to respond to the Guarantor’s requests. Hopefully in May Gpt-4 and ChatGpt will be back online for Italians. With lots of information, hopefully as comprehensive as possible. So yes, the ChatGpt family would stop being the most dysfunctional (and obscure) of the web.