Are there any jobs that can't be automated?
Given what you know about robotics and machine learning, do you think that there are jobs that can't be automated? Are there one's that are going to be harder to automate than others? What are those jobs?
Comments (23)
Why? You need to tell an AI what to do and how to learn, they're poor at transferring learning in one problem space to another, poor at recognising 'new' problem spaces. And this is with as much data as they need to learn.
I believe they also aren't great at knowing why something is the case. They might be able to infer that someone who has gone to a Slayer concert is more likely to go to a Megadeth concert, but from that they couldn't learn of the underlying construct 'metalhead' without input.
Underlying all of this is access to data. Any data stream which somehow resists digitalisation, storage and propagation will be difficult to automate tasks using AI within. Transience is anathema to beings of repetition.
Anything that involves meaning and nuance and flexibility.
I hold the belief that there are limits to logic and mathematics in the sense that there are other aspects to being intuition, awareness, sensation, emotion, skills, and importantly habituation; aspects of being that cant be replicated in mathematics, logic, symbols, or material. I think attempting to do so is probably a fools errand but I could be overlooking something.
In any case, if all human action is automated through machines then its centuries off.
Fantabulous!
I would point out all of those jobs which need the "essence" of human nature fulfilled by emotions. For example: A writer or editor. Imagine writing a history about the holidays of this summer which you have spent with family and friends. A robot would not be able to do that because it is full of emotions and memories. I don't even imagine an engineer or programmer putting this information in an IA.
Complex feelings such as sadness and nostalgia are only upon us and how we use and make art with those feelings.
An AI editor, (?) even so far as a proof reader, could exist in the future. Consider the neonate forms like grammerly?
Some specific types of human services can be 'mechanized'. Machines can 'perform' in nursing homes, for instance, leading group exercise sessions. Machines can perform a kind of counseling service of listening and providing some level of listening and response.
Anyone who has been repeatedly frustrated by automated telephone systems where one speaks one's responses, has experienced the limitations of some current software abilities. (However, sometimes they work just fine.)
That said, I am not arguing that machines SHOULD be doing any of this, just that -- if standards are not very high -- they can. They are not being used because people prefer to interact with robots; they are used because corporations are usually trying to save money.
I am not a detail person, and in some work places, I would gladly have given my job over to a robot to perform paper processing. A good share of clerical work probably could be replaced by automated information processing.
The consequence of eliminating jobs is not trivial, however. There are detail-oriented people who can process paper all day without becoming remotely homicidal.
Write a text, yes but just copying some empty words. Whenever you try to write a book or poem you spread all your emotions in them. I could tell a machine my thoughts and order to copy it in a "Word" page. But they would still be my thoughts and emotions.
I can't see neither imagine a machine having emotions itself because this issue is inherent to us, humans.
We love to write stories or poems about autumn or sunset. A machine couldn't understand this feeling.
It is possible, indeed. But I think their feelings would be limited by his developer's. The machine would only be able to script a few feelings according to a specific human.
For example: I really enjoy rainy days. Probably the developer hates rainy days. Then, a machine would have a different feeling from mine not reaching a "standard" feeling because that's impossible.
1. Thinkers (IQ)
2. Feelers (EQ)
3. Workers (PQ)
Group 3 is most susceptible to being replaced/augmented by machines.
Work on how to automate group 1 abilities was begun in the 1950s, but apart from Deep Blue which defeated world champion Kasparov in chess and AlphaGo which beat, another world champion, Lee Sedol, there's not much to brag about in the AI field; group 1 is safe (for now).
Group 2 seems to be the last entry in our wish list; these guys and gals are probably gonna be the ones AI want as symbionts!
The real moot point, as Turing realized, is whether you can tell the difference between your man-made pot and the AI one. If you can't, the question is surely meaningless. Of course by extension you can consider any work of art in the same way. If AI can produce a believable Picasso, isn't it worth the same as a genuine one? If AI can produce a concerto indistinguishable from one by Mozart can you call it inferior?
No. That is no different to an art forgery by a person and those works are not worth the same as a Picasso. What we are paying for in a Picasso is not the art, so much, but the provenance and the fact Picasso did it. Can a fake sell for the same price as the real thing? Yes, if the right academics vouch for its authenticity. But the price is only high if it is believed Picasso made it.
Quoting Tim3003
It will definitely be an inferior Mozart because it won't be Mozart. I think there have been humans who have copied works in Mozart's style before and those works are generally experienced as derivative. I suspect that a musicologist will always be able to tell a Mozart from an ersatz Mozart, no matter how good the latter. The rest of us may not pick it up. But let's say no one can tell them apart. In this case what we have is a work people would say has many of the attributes of a real Mozart and it may well be enjoyed by many as such. I guess we could still say it is inferior in as much as no human made it. So the extraordinary effort of a single person's compressed creativity and innovation is not on show.
Psychotherapy?
When the best psychotherapists are AIs, we might be doomed.
Dont psychotherapists just sit there and listen? If anything can be automated psychotherapy is the easiest one of all as all they do is ask you and how did that make you feel every once in a while whilst youre lying on their couch :lol:
At least one: "god" (the) omni-providential agent.
None after AGI has been achieved (i.e. post-Singularity).
Automating a politician would be hard though now thats a future challenge for current AI :lol: it would crash and burn with contradictions