Humans Still Cost Less Than Artificial Intelligence, MIT Study Says

A study by MIT, the Massachusetts Institute of Technology, found that humans are the best option for some services than AIs. The authors of the research did the math to find out if artificial intelligence in computer visualization tasks has a lower cost than allocating people to computer visualization functions.

Explaining and exemplifying computer visualization, computer visualization is the technology on which AIs recognize elements of images and you’re probably training them. You know when reCaptcha asks you to identify all images with traffic lights? This is used so that an autonomous driving AI learns to recognize you when you’re driving.

An example given by the authors on how computer visualization can replace humans is in a bakery. Rather than using a baker to assess the quality of ingredients, whether they are good or bad, a trained AI can identify the characteristics of a spoiled ingredient. However, because it is a quick task, the cost of training and maintaining this system would not be advantageous.

Humans are the best option for now?
The study published by MIT reveals that with today’s costs, 23% (almost a quarter) of services involving vision and analysis could be switched to computer visualization AIs. In these cases, the article compares employee salaries to assess whether replacing a human will be profitable.

23% is a considerable number, but the study only evaluated computer visualization tasks, as this is a more mature technology with more structured pricing. For these AIs to gain space, it is necessary that their cost decreases. At the moment, humans remain the most cost-effective option in this segment — and even more so in others, especially those involving management and social skills.

The authors also highlight in the research that the replacement of human labor by AIs will be gradual. The article points out that the pace of change tends to be lower in this segment (computer visualization) than the employee turnover seen today. The study suggests that the speed of substitution will be slow enough for governments to implement policies that reduce the impact of unemployment.