Scientists have estimated a 5% chance of AI getting out of control and destroying humanity

Possessed Photography/Unsplash

According to a survey of leading artificial intelligence scientists, there is a 5% risk of the technology becoming unmanageable and eliminating humans.

Here's What We Know

2,778 experts in the study of artificial intelligence took part in the survey. It is the largest such survey to date.

Most respondents are convinced that AI will advance a lot in the future. For example, by 2030, it will learn to compose musical hits and write bestsellers.

68% of respondents believe that super-intelligent AI will do more good than harm to humanity. At the same time, half of them admit a non-zero probability of destruction of people by artificial intelligence.

The authors stress: all scientists are aware of the dangers of AI and the need to study ways to minimise the risks. However, opinions differed in their assessments of the speed of progress.

More than 80% of experts expressed serious concerns about the use of AI to spread fakes, control populations and exacerbate inequality.

Risks of artificial intelligence that worry scientists

In addition, respondents gave predictions for different milestones in the development of AI. They believe that by 2033 the technology will learn to assemble LEGOs and create games, and by 2063 it will be able to do the work of a surgeon or scientist.

The authors recognise that their conclusions are not absolute. Nevertheless, they may become an important element in understanding the prospects of AI, given the experience and knowledge of the interviewed experts.

Source: Gizmodo