Artificial intelligence (AI) is transforming the world at a rapid pace.
Thanks to this technology, intelligent machines are expected to perform better and at a lower cost than human beings.
Whether this is a good or bad thing, industries are quickly adopting artificial intelligence, but only a few of them have reaped considerable benefits to this point.
This begs the question: will AI ever surpass human performance? More importantly, when will machines outperform you in doing your job?
Thanks to the incredible work done by Katja Grace at the University of Oxford’s Future of Humanity Institute; there is an answer to this question.
In fact, Katja and several friends sought the answer from the experts.
The task included surveying the top artificial intelligence researchers in the world by querying them about when they believe intelligent machines would outdo human beings in a broad array of tasks.
Most of the answers they received were somewhat shocking.
The people who Grace and her friends talked to were industry and academic experts who provided papers at the International Conference on Machine Learning back in July 2015 and later in December 2015 at the Neural Information Processing Systems Conference.
Both occasions mark some of the most popular events for artificial intelligence experts, meaning that most of the experts in this field were likely present.
Grace and her colleagues asked the 1,634 experts to fill out a survey regarding when AI would be cheaper and better than human beings at completing numerous tasks.
Out of these experts, only 352 of them responded.
The professionals forecast that artificial intelligence would surpass human beings in the coming decade in activities like translating languages (by 2024), completing high school essays(by 2026), as well as driving around trucks by 2027.
However, other activities will take machines more time to master.
The experts are still far from making accurate predictions.
They forecasted that artificial intelligence would outperform human beings at GO by around 2027, even though this was back in 2015.
DeepMind, a subsidiary of Google, has come up with an AI with the potential to outperform the best humans, a task that was completed in two years as opposed to 12.
The AI experts also predicted a 50% likelihood that artificial intelligence would outperform humans at almost everything in nearly 45 years to come.
Forty years marks an essential number, primarily when human beings make predictions since it is the duration of the working lives of many individuals.
Hence, any forecasted change that is beyond that figure means that it will take place further away from the working life of all people who are currently working.
The 45-year prediction represents a median figure drawn from all the responses given by the experts.
To check whether different groups came up with varying predictions, Grace and her people determined how the forecast varied with the researcher’s age, region of origin, and the number of citations.
They found out that expertise and age do not affect the prediction, even though the origin of the researcher does.
North American and Asian researchers expect AI to outdo human beings at completing all tasks in 74 and 30 years respectively.
This difference begs the question: what do researchers from Asia know about AI that those from North America don’t (or the other way round)?