What work is least likely to be taken over by scaling deep learning?

Basically everything that requires "humanity" is relatively safe, like kindergarten teachers, practitioners in medicine and psychotherapy, as well as all the banking jobs about babysitting money for rich people. The latter I think is a good example of what will happen increasingly, as algorithms have almost completely taken over investment strategies, or at least regularly outperform humans, but it remains essential for a bank to have a "human front", as customers are rather unlikely to entrust their savings to a non-human.

Increasingly I think humans will become either "machine managers" or "the human face" of otherwise automised processes. Medicine is an interesting example here as there are quite a few studies showing that empathic patient care increases patient outcome, and in life-and-death situations, a human operator is important to catch machine failure as well as for liability purposes. But a human doctor can't (won't be able to) compete with the diagnostic power of a specialised AI trained on billions of diagnosis. This is of course a good thing, as we as a species will become much better at diagnosing and treating illness in the process.

DALL-E, if it holds what it promises, will definetely shrink the market for freelance designers at the lower end of the market, and it might transform design at larger corporations too. Here I think it is plausible for designers to become more and more like machine operators, in which they write detailed instructions to an AI (like googling is a core strength of programming), and then do the last bits of fine tuning themselves, as well as making sure that the overall client vision is fulfilled in all parts.

Note though that a doctor spending less time on diagnosis can lead to a) more time for patient care or b) less overall doctors. A machine-operator designer has a much higher productivity, which might reduce the number of available jobs in the sector. On another note, this should all be taken with a grain of salt. 10 years ago I would have said transportation is the first job sector to go, yet now it increasingly looks like creative jobs might be hit before that. Things aren't moving in a linear fashion.

I think anything that involves complex requirements gathering, designing complex and cohesive (i.e. not forgetting the context/constraints like GPT 3) solutions, and coordinating with multiple people + systems to get things done is going to be hard to automate e.g. figuring out what app to build and its architecture, designing a complex UI/UX, project management.

Anything where the exact requirements can be boiled down to a few sentences like "create an image of an astronaut riding a horse in a photorealistic style" is in clear danger now vs "create a website where people can discuss tech news" or "create an app where people can keep in touch with their friends".

Typical app development projects have hundreds of competing requirements and constraints to navigate that are constantly evolving so I'm not even sure what the AI interface for this would look like that would make this managable. Is there anything in this area?

A few jobs like football player or chess GM should be inherently immune to automation; they exist because we're interested in seeing what humans can do, even if machines can outperform them.

For some artistic/creative jobs we might also value it coming from a human, but I'd be cautious of overstating the effect here. Art in a gallery may be safe for now, but more "functional" art like character portraits for a videogame or clip art for a company website seems in imminent danger.

Plumbers, electricians, mechanics, etc. can be replaced for repetitive known work, like mass-production of vehicles or household appliances, but I don't think robotics is advanced/cheap enough yet to be approaching the point where a robot would turn up to your home, navigate around, access the right areas, and fix some variable maintenance issue.

Things you don’t do in front of computers. Plumbing. Teaching kids. Sports.

Even the computer jobs I’m not sure will go as quick as we think. Yes, deep learning can write code or paint pixels on the screen but not to the degree it’s going to replace a good software architect or a UX designer, or maybe it will augment those things to take some of the grunt work we do manually out - but not the high level thinking/innovation.

Eh don’t worry about it, AI will just empower us to accomplish more than we’ve ever been able to. If anything, up your expectations of what you will be capable of

The new high in demand job of "AI fearmonger" should be safe... but I think that of "tk er jobs" fearmongers isn't.

For the love of God please tell me programming.

Anything that can be branded as "leadership".

Developing new/better deep learning algorithms?