According to OpenAI, a digital “superintelligence” could be developed from the computer before the end of this decade. The ChatGPT provider is therefore planning a method of how people could control them.
ChatGPT provider OpenAI is considering how humans could control a potential digital “superintelligence”. In order to find a solution and thus ensure the security of artificial intelligence (AI) for mankind, the company wants to use a fifth of the available computing capacity over the next four years and put together a new research team, according to a blog entry.
“super intelligence” could disempower humanity according to OpenAI
“The enormous power of ‘superintelligence’ could lead to human disempowerment or even human extinction,” warned OpenAI co-founder Ilya Sutskever and the head of OpenAI’s future direction department, Jan Leike. There is currently no way to direct or control a potentially super-intelligent AI.
“Humans will not be able to reliably oversee artificial intelligence systems that are much smarter than us,” San Francisco-based AI company Firm said. Therefore, an automated procedure should be developed for this. The aim is to master the central technical challenges for such a system within four years.
The development team will be led by chief scientist Sutskever, among others. Even if the “superintelligence” from the computer still seems to be a long way off, OpenAI still believes it to be possible in this decade. It will solve many problems and be mankind’s most momentous invention.
Discussing the potential dangers of artificial intelligence
The move follows a warning from industry experts who believe AI carries the risk of existential dangers. One of the signatories is OpenAI boss Sam Altman. The alleged threat was not explained in detail – and critics accused the experts that such warnings distracted from problems of artificial intelligence that already exist today, such as discrimination in algorithms.
The EU has already agreed on a draft law to regulate AI, the “AI Act”. In the USA, following the example of the EU, there is a discussion about classifying AI applications in risk classes, on which the requirements for companies depend.
Among other things, OpenAI developed the chatbot ChatGPT, which triggered a new hype about artificial intelligence. The program can create sentences that are almost indistinguishable from those of a human. For this purpose, it was trained with enormous amounts of data and, on this basis, estimates word for word how a sentence should continue. However, this principle also ensures that ChatGPT sometimes outputs completely incorrect information.