Ask HN: Last industries to be taken over by AI?

The ten jobs least likely to be taken over by computerisation:

1. Recreational Therapists

2. First-Line Supervisors of Mechanics, Installers, and Repairers

3. Emergency Management Directors

4. Mental Health and Substance Abuse Social Workers

5. Audiologists

6. Occupational Therapists

7. Orthotists and Prosthetists

8. Healthcare Social Workers

9. Oral and Maxillofacial Surgeons

10. First-Line Supervisors of Fire Fighting and Prevention Workers

This according to Frey & Osborne, The Future of Employment - How susceptible are jobs to computerisation? (2013)

http://www.oxfordmartin.ox.ac.uk/downloads/academic/The_Futu...

Maybe I'm saying this because I've been involved in the restaurant industry for most of my life in varying capacities (cook, chef, waiter, bartender, manager, soon to be owner), but I think there will always be a certain niche for restaurants and bars staffed by humans.

Certainly fast food restaurants, probably establishments like diners, hotels, etc..., will make use of automation, but going out, talking to a human, getting drinks and food made in front of you - that whole experience isn't something you'll ever be able to get from AI and robots.

I do feel that when AI/automation truly takes over, humans will definitely be relegated to artistic/performance media. Hopefully, this means a good societal safety net/basic income, and that we can spend our time pursuing various arts and scientific research, and that society doesn't turn into the dystopia that so many are afraid of...

As a consumer I'd nearly always prefer a machine take care of my needs than a person or an animal. It's fun to talk with my friends, family or other diners at a restaurant, it's not usually fun to talk with the staff. At least in Chicago they usually seem miserable, incompetent, or a mix of both.

It might be fun to have niche locations with humans serving people and taking orders but I imagine it'd be more like horse drawn carriages serve tourists.

Maybe you visit the wrong restaurants or other parts of the US are very different than Chicago? I have had mostly excellent staff experiences in the US to such a point that I use US examples to train our restaurant staff. Also, although I would like to see far more automation, I would suppose your opinion in this particular case is rather unique? Obviously I do not know that for sure but experience says it is in my friend group.

I live in North Carolina and see a full spectrum of restaurant service quality. In the end, you tend to get what you pay for. Though my biases and sister tell me we're nicer and happier down here than in Chicago ^^

I think food services at any level will see automation. There may be some people still around to take care of some "customer service" like tasks, fixing missed orders or accommodating special requests, but over all the food industry has clear procedures on how jobs are to be done, and that leads to automation.

Even though it feels like a bit of a non-answer, I'd say the last industry taken over by AI is probably the AI/automation industry itself (or, in a broader sense, programming/software engineering).

I came here to make this exact same non-answer.

At least you got to make a non-comment :-).

The question is ill-posed at best. What do you mean by "fully taken over by AI"? Take your entertainment example, say movies and TV shows. Do you suggest these are fully taken over when all the actors are replaced by animation? Or also all the marketing and other ancillaries? How about IP ownership? Or is it merely when the movie script is written by AI?

> I can't think of anything we won't be able to automate

A lot of manual labour will never be fully automated, since making a general-purpose robot that is as lightweight, flexible, cheap and self-contained as a human is not going to be a positive ROI.

Especially in this hypothetical age where most humans are left unemployed by AI, the cost of labor will be near-zero. People may even be willing to work for free, just to fill their days. Thus the robot will always be more expensive.

If you want proof of the latter, go on youtube and check out all the people doing metalwork, woodworking, making food, brewing, arts&crafts etc. just for fun.

It's tasks that are taken over by AI and is not industries or job that's taken over. With that regard, any industries are subject to some tasks being transferred to AI, but that's much more like how many of tasks that was traditionally done manually done by machine one way or the another.

Despite what the silicon skinned Japanese toys might suggest - sex workers.

As a parent: caring for a child for any amount of time is the obvious thing. It would require almost complete human replication in ai form

The arts/humanities maybe? Because the human intellect and human emotions will always be relevant and valuable to other humans.

Sidenote: when I say "arts/humanities" I'm not just referring to entertainment but also the branches of society that deal with introspection and cultural criticism for the sake of human autonomy.

I think it is something on the first line for potential replacement. Deep learning chat bot, which learned emotions from million of books can come tomorrow and successfully imitate human emotions.

A computer's capability for imitating a human doesn't take away the value of human-made art. If anything, it's another medium. Otherwise you're arguing that people don't care about understanding the human condition, which I don't think will be a valid claim anytime soon. AI can definitely take over the production of decorative objects though, of course.

Is it harder to create a brain than a body?

Is it harder to create generalists than specialists?

The fine grained resolution we have with respect to actuation and tactile sensing might be harder nut to crack than honing down on the regularities in our cortex. A teacher for example needs to guide your hand when you learn to write or prevents you quietly from drowning when you learn to swim.

Even a truck driver might be more a generalist than a specialist. A truck driver jumps out of his truck to unload, fill in papers, prevent people hitchhiking in the back, prevent theft, taking detours, finding an address even with errors in the administration.

I also doubt sincerely that we will be able to tune the AI in such way that they will be content with all the jobs humans do not do anymore. Are we able and willing to codify a society on intellect? Will we be willingly creating unambitious AIs for particular dull tasks?

I think the touch of human labor has a value of its own, especially in high-end manufacturing and services. If high tech automation allows us all to have precisely perfect mass produced tables and chairs, for example, the value of handmade furniture with its "imperfections" will rise. Just as mega-scale agribusiness produces ever more perfect fruits, vegetables, and grains, has been accompanied by more and more people taking an interest in farmers' markets and local producers.

Now, certainly the mass producers are making money, so I'm not suggesting those industries are poor investments. I'm just thinking that strategically, if you want to find a niche that won't be eliminated by technology, look at the high-quality end of any particular market.

Subsistence farming? I think anywhere where it is simply not worth putting AI will be the last to go. The easiest industries to replace will be first (factory workers, truck drivers), I think the high value jobs will be next (simply because they are high value, there will be more effort in replacing them) e.g. managers, ceos, engineers. (these will be a little ways off though)

There were a few papers published this year using reinforcement learning to optimise architectures for CNNs and recurrent neural nets, I think this sort of thing will only get bigger. If you think about it, researchers just do a bit of guided random search, something that reinforcement learning can do pretty well. So the 'research' job title may describe applying these AIs instead of actually doing low level research.

Entertainment will be early to go.

As soon as we can get realistic video and audio why employ 100+ people to do the show. It'll all be CGI.

One person can do Game of Thrones.

This is why I think CG audio will be an amazing jump for humanity whole new content will be created and make other content obsolete.

Okay, so the grunt work of entertainment (manually creating models, rigs, applying textures, shaders, etc) might eventually be reduced (although the current trend is the opposite, despite better and better software), but the creative aspects of entertainment (not just coming up with something 'different', but making it a meaningful difference) will be damn difficult to replicate in a consistent and repeatable way by computers for a long time to come.

Creative work also benefits from multiple people (up to a point) with multiple unique experiences of living life working together, as well. We aren't even trying to do that with computers right now, so I think their creative output would be more limited in scope (again, for awhile, not necessarily forever).

Computers may be able to do variations on an established formula, though, and there's a lot of formulaic dreck out there.

Computer aided storylines will be the start?

Marketing, which is basically applied psychology.

You're absolutely right, and sadly, this (IMHO) is to the detriment to humanity as a whole. The ease and accuracy with which the fitness function provides feedback (did this change result in more money being spent?) means that this field will grow to be very accurate in an extraordinary time-frame.

It's not just the fact that this will be used to drain the wealth from society. The same techniques used in marketing to get people to purchase a product can be applied to convince them to do anything. We've already seen how fake new can be used to convince people to bring rifles into pizza places full of young children.

If these techniques can be applied in an automated and targeted fashion, the results could be absolutely devastating to society. Imagine what happens when a company develops the technology that could mine social media profiles, identify people that would likely commit mass murders, then use automated tools the subtly influence their behavior in such a way as to promote that outcome. Then they sell it to the highest bidder.

I consider that to be a scary world to create. And I think it is totally plausible, if not inevitable.

I work in marketing and personally budget out 8 figures+ a year in direct marketing spend. The truth is it's really hard to persuade people of things. Direct Marketing is 50% about finding people who already agree with you and want your product, 25% getting their attention, and 25% is testing which product messaging most jibes with people.

It's not about if we can automate something, I believe it will come down to social forces preventing ai in certain areas.

Politics are not going away any time soon, nor is law.

Most jobs involving some amount of nurturing and empathy, although you can automate some of them (eg: massage therapist). Many people will want to be heard, seen, and understood by another human being.

Edit: I wouldn't worry about most jobs being replaced any time soon. It's a tougher problem than you probably think it is.

Counselling? Psychologist? Pastor? Something along those lines I think.

Sufficient wet & squishy to make it difficult to work out what you'd need to automate let alone doing so.

Non-intelligent computers have already been used as therapists since as far back as the 60s.[0] People are easy to fool when you're feeding them platitudes.

[0] https://en.wikipedia.org/wiki/ELIZA

The question is can AI industry be taken over by AI?

Fiction writers. I doubt AI will ever gain the proficiency to write a good, interesting, original story.

I can't remember exactly - but i recall reading an article that was about an fiction writing AI winning a price for a short-story. Oh - google my old friend... http://www.digitaltrends.com/cool-tech/japanese-ai-writes-no...

AI will have a hard time writing stories that are interesting to humans.

Law will take some time as often questions are more nuanced than simply applying the law to a problem.

That, and it's going to be really hard to extract the facts of a case through automated means.

There are two big benefits to AI in the legal realm -- automation of rote tasks, such as constructing simple contracts, that companies like LegalZoom already do; and predicting outcomes of trials based on the lawyers and judges involved in the case.

For sure, though Judicata (https://www.judicata.com/) are making good progress in this.

Automation of routine tasks will certainly reduce the number of lawyers as many are used to taking styles from their firm's systems and amending it to suit a client's needs, but for bespoke legal advice it will be hard to replace an experienced lawyer (for now).

And yes, case outcome prediction is a great tool but I see that being more limited to insurers deciding whether to fund cases etc.

You're assuming there's still a human audience in a position to demand human entertainment.

I think sports entertainment and sports fans will exist for many years to come.

Programming, hopefully.

last job will be politics, not because AI won't be able to make better decisions, but people will never give up power !

art but it is not an industry and also psycholgists. apparently it takes a human to understand a human better

Teaching.