Understanding the Impact: The Rise of AI in Job Automation and Economic Viability
In a study by MIT‘s Computer Science and Artificial Intelligence Laboratory, the “Rise of AI” in job automation is explored, particularly focusing on computer vision (CV). This branch of AI deals with understanding and analyzing visual data.
The research looked at how well AI can do certain jobs and whether it’s cost-effective to replace people in different industries. This study isn’t peer-reviewed yet, but you can find it here.
One key point is that AI could automate jobs that pay 1.6% of all wages in the US (not counting farming). But, it’s only cost-effective to use AI for about 23% of these jobs, which is just 0.4% of the US economy.
So, AI can take over some jobs, but it might not always be cheaper or better.
The researchers say, “AI could significantly change jobs, but it will be a slow process. This gives us time for policies and training to help reduce unemployment.”
They also estimate that only 3% of jobs helped by vision technology can be automated cheaply now. But by 2030, this could increase to 40% if it becomes cheaper and more accurate.
Retail, transportation, warehousing, and healthcare are areas where using AI for vision tasks makes the most economic sense.
Other studies, like one from the IMF, have suggested that up to 40% of jobs worldwide could be affected by AI.
This MIT study focuses on AI for vision and doesn’t go into detail about more versatile AI systems like OpenAI‘s GPT-4. click here for more information of how ai works.
While this study predicts a smaller impact of AI on jobs, OpenAI’s research last year showed that 19% of US workers might see big changes at work because of AI advances like GPT-4.
One big hurdle in using AI for jobs is the cost of customizing it for specific tasks. Big companies might afford it, but smaller ones might find it more cost-effective to keep skilled workers.
However, adapting language models like GPT-4 for different tasks might be easier and cheaper, leading to more use in various industries.
Neil Thompson, the lead author of the study, told TIME, “It’s quite possible that customizing big language models is simpler than for vision systems, and this could mean more use in the economy.”
But there’s another issue: AI uses a lot of resources. OpenAI CEO Sam Altman said that AI can’t grow much more without a major advance in how much energy we can produce.
