There are more and more people in the scientific community who are warning about the dangers of artificial intelligence.
At the beginning of Blade Runner we find a technician doing the job Turing test to determine whether an interviewee is human or not. This, which was pure sci-fi in the 1980s, seems to get worse every day. Closer to reality. In recent months we have seen how the world’s most advanced robotic robot has declared itself self-aware, so it seems we are witnessing a moment of great importance for human history.
AI systems like ChatGPT-4 have brought about unprecedented advances in science. Although there are many looming fears, such as careers that will disappear, the truth is that we are already seeing that There are more troubling issues.such as not knowing whether humans were created or not Self-aware robots.
For this reason, a group of scholars, including A OpenAI expert He is working on discovering how to detect these problems Before it’s too late.
A plan to discover machine self-awareness
Lukas Berglund and colleagues from various institutions have published a preprint on Arxiv.org explaining this We must be prepared for what artificial intelligence may entail. Largely because the closer you get to Human intelligenceArtificial intelligence poses notable difficulties in determining whether it has already achieved self-awareness or whether it can be evaded Proven security protocols So as not to take actions harmful to humans.
Berglund and his team concluded that the need to study these technologies is to be vigilant against potential risks:
Because of these risks, it is important to predict when a situation will arise in which we must be prepared.
In this context, they are working hard to enhance the algorithms that allow AI to work in the best possible way and remain safe for everyone. For this reason, they want to study a test that could be used to detect and determine what stage an AI is in How it works and thinking. Something that may seem at first glance like a trivial matter, but in reality it is It is of fundamental importance if we consider how quickly artificial intelligence is advancing And the computing power they have recently achieved.
To sum up:
- There are increasingly clear concerns about artificial intelligence. Especially in terms of how close they are to getting there Pure human intelligence.
- Until now, the Turing Test has been the most reliable method of control If we had a robot or a human in front of us.
- With the latest artificial intelligence technologies, it has become increasingly necessary to enhance and improve this test ever since He fell short.
- Now, a team of scientists is trying to dig deeper into the effectiveness of these tests in existing chat programs.
- Above all, it is important to keep in mind that today there are more and more ways to bypass the security limits in chatbots.
- This means that they can be evaded and thus used to do evil.
- The goal of this study is to develop a test that allows artificial intelligence to be detected based on how it works.
/cloudfront-us-east-1.images.arcpublishing.com/eluniverso/ZHT3NFYA2BFMFDHOGERE3EWDBY.jpg)