Oh don’t worry the robots are coming for almost every minimum wage manual labor job as well. Those were actually automated first, and simply haven’t rolled out yet due to humans being cheaper than machines right now.

Jobs able to be automated include:

• shipping dock workers

• any driving job

• any job in a restaurant (chef, waiter, cashier, etc)

• retail checkout

• any janitorial job (cleaning floors, restrooms, etc)

• any job related to scheduling times or managing paperwork (receptionists)

• anything done in a factory (manufacturing)

• any job done in an airport

• security (soon)

• construction of homes buildings (soon/wip)

• most work done in farms (animal or produce): planting, watering, harvesting, slaughtering, animal cycle, etc.

And presumably more. The technology is currently available to completely automate nearly 99% of jobs you interact with on a daily basis.

With this new wave of AI, as the tweet mentions, the following jobs will soon be automated:

• anything involving writing (journalism, fictional books, information guides, walkthroughs, tutorials, etc.)

• anything involving the creation of static images (stock photography, logo design, 2d game art, manga/comics)

• many things involving amateur or quick code (basic websites, simple scripts)

• Q&A/consultants (any human you’d ask a question to), think consultants, a lot of doctor work, therapy, etc.

• Product/electronics design (already partially automated, will improve more, but not be completely replaced)

• Many nsfw jobs (porn photos, hentai manga, sex therapist)

• Finance stuff (already automated to a degree, but will be completely so)

• Planning/organizing business deals

• marketing, call centers

Things that will be automated once the existing tech improves:

• more complex software/code development (hard to say whether it’ll reach professional level or not soon, as there’s more than just code involved there, and large architectures hard to automate)

• anything dealing with videos (ads, movies, tv shows, cartoons, anime, informative videos, cute animal videos, many tiktoks, commentaries, talk shows, tutorial videos)

• legal work (technically can be automated now with new wave, but there’s a lot of hurdles preventing this from truly being automated, due to fabrication, lack of specifications, legal restrictions around tech, etc. mostly just needs improved reliability+specific model, along with legal change)

• simple video game creation (think atari, pong, simple nes games)

• almost every non-irl/non-physical nsfw job (photos, videos, many games, product design)

Things that are safe for now, but will be automated in the future:

• full code/software development of any type (web, tools, etc.)

• game development in all fields, even for AAA

• complete design/development/production of electronics and computers (singularity point, once this is hit, everything will be automated)

• complex tasks involving non-deterministic situations involving multiple fields (assistants, general knowledge workers, handling new fields, dealing with format conversion)

• nsfw game development

Jobs currently safe until we reach proper AGI:

• any task involving entering unmapped territory, navigating, doing a particular nondeterministic action (plumbing, repair, etc.)

• any task involving a human to confirm/commit (governmental jobs, final ceo/manager jobs, controlling militaristic weapons, legal jobs/courts)

• any job requiring novel task acquisition (reverse engineering, data format analysis, cyber security)

I notably left out one domain from this list: dynamic physical tasks. For deterministic physical tasks, we can build specific machines, and have ai working for them already in most domains, and soon in all domains. However, dynamic physical tasks we still struggle with. This is being worked on, but it’s slow and expensive. This is why tasks like plumbing or repair won’t be automated for a long while: not because the task itself is hard (we’re almost at the point where ai can do this for deterministic scenarios). The issue is being able to dynamically traverse the city, interact with people, enter a home, navigate the home to the issue, navigate to the specific area of the problem, problem solve to figure out problem area if customer doesn’t know, and then actually do the task. The “do the task” part is more or less solved. It’s the novel navigation that’s hard. In practice: a robot to do the task will be created, and the plumber/repairman just sets it up and hits “start”.

The other tasks that won’t be automated involve novel task acquisition, or final human confirmation. The former is a “hard problem” in ai, and requires full AGI to work. Once AGI is achieved, all tasks will be automated. So these are the last to go. Not many jobs require this, but things like reverse engineering, analysis, or cyber security tend to require such.