Benjamin Todd

But generating ideas is one of the strengths of the latest AI systems.

For instance, they can generate hundreds of images in the style of Dali crossed with Pollock nearly instantly, or endless ideas for attention-grabbing headlines.

The types of tasks that seem hardest to automate likely involve decision-making and problem-solving, for example, choosing from a variety of AI-generated images, especially decisions where it's important for a human, perhaps for legal reasons, to stay in the loop.

social intelligence and relationship building, difficult motor skills, robots are lagging behind generative AI systems, so jobs from plumbing to surgery are likely to be least affected, at least for now, and high-level expertise.

AI systems are still not as accurate as top human experts within their area of expertise, though it's not clear how long this will last.

It's very hard to predict how this will affect the labor market over the next 10 years.

The 2020 analysis discussed above argued that jobs between the top 70th percentile and the 99th percentile in terms of income will be most affected by advances in AI and are likely to see lower relative income.

The list of jobs most likely to be affected includes chemical engineers, optometrists and dispatchers.

In contrast, the list of the least affected jobs included entertainment performers, food preparation workers and college instructors.

This analysis is just one model, so we shouldn't fully trust it.

It's not that the jobs most affected by automation will see reduced employment or income.

If each chemical engineer you hire can do the work that two could previously, that could lead to hiring half as many engineers or it could lead to hiring more engineers because now each produces twice as much value as before.

It all depends on how the economics of the situation works out.

What's clearer is that jobs will shift to involve more of the harder-to-automate tasks and fewer of those that can be done by AI systems.

So this means that if you want your skills to stay relevant in the future, focus more on learning the hardest to automate skills, perhaps such as the ones above.

And also focus heavily on learning how to use AI to augment your productivity.

The workers who do best in the future will probably be those most able to make use of AI and automation to solve important problems.

Beyond the next 5 to 10 years, it becomes near impossible to know what will happen.

Ultimately, it seems like AI systems will be able to do basically all jobs better than humans.

And who knows how the economy will look at that point.