> In other words, and paradoxically, work in labor-intensive jobs will be the last to be automated. Many used to think those jobs would be the first in line for automation, and creative jobs will be automated last.
This is surprising to me but probably intuitive considering the market forces. Creative jobs are highly specialized and therefore expensive. (Low supply + high demand = high cost). On the flip side, labor-intensive jobs are not specialized and therefore cheap (overall, I know supply is an issue in the US right now at least).
If the market is trying to reduce costs it will start with the highest first, especially those with the highest barriers to entry as far as skill development. It probably doesn't help that it appears computer vision is advancing slower than other areas of AI.
Yes, you correctly point out that high cost is a big factor. The other factor in why labor-intensive jobs are difficult, is that robotics has lagged far behind AI software. Even if Tesla or others mass produce robots starting in 2025 or 2030, it will take a very long time before there is enough robots to satisfy the labor-intensive demand globally. Thanks for your readership!
OpenAI estimates that Codex (what powers Github’s Copilot) can already complete 37% of coding tasks. DeepMind’s non-commercial AlphaCode is estimated to be 59% better at coding competitions than Codex. But how close are we to automation? There is a huge demand for developer jobs currently. In light of OpenAI’s DALLE-2 commercialization announcement yesterday, here is a new cross-disciplinary approach to determine how close we are to automation.
After reading the article how close, or far, do you think we are from software development jobs getting automated?
> OpenAI estimates that Codex (what powers Github’s Copilot) can already complete 37% of coding tasks.
This makes me a bit suspicious of the article.
Obviously this is just one developers experience now. I've been using Copilot for the last month, and it's an awesome bit of tech. But generally speaking, when I let it write a blob of code:
* There's a 50/50 chance it'll even compile
* It's always, _always_ wrong in some way, usually in ways tricky to spot
* And the tests it wrote _for it's own code_ are generally wrong too
As always, I think the bottleneck is in describing exactly what you want in a way an AI can successfully interpret it.
My experience with DALLE-2 and Copilot makes me think we are quite far off from AI being able to perform software dev on it's own. Both tools need a lot of oversight to pick out usable pieces from the gibberish that gets spit out most of the time.
That said, copilot is a great tool for generating boilerplate code, like writing a model class for a parsed JWT token or adding tedious annotations somewhere. I've used GPT-3 to generate javadoc headers for a codebase before with decent success. It gets way more fragile when there's an actual problem to be solved though, and is basically unusable about 100% of the time on complex areas of code.
From my experience, the real result will be that AI might replace interns and some people who are only doing the tedious data modeling parts of coding, but there's a very, very long way to go before it will be able to be more than just a tool for developers to code slightly faster.
I expect that in 2030, you will need 3 ppl to write code that now you need around 10. I think more devs will switch to more product owner roles or analysts that talk to ppl and spec but more high level - constraints etc. So basically these entry level coding that’s boring should get automated. Skilled individuals will get up to 10x leverage and be real rockstars.
That’s extrapolating just from current capabilities. I think there’s chance we will figure real reasoning in AI in next 10 years given interest/investment in field.
currently AI can only base its solutions on what's been done or shown as example before, especially in coding, so if you replace all software coders with AI then you have limits to what it can do
> > In other words, and paradoxically, work in labor-intensive jobs will be the last to be automated.
This I'm a little suspicious of, because of all tasks human did throughout history, labour-intensive stuff has _always_ been automated the most, and we didn't need AI to do it. Stuff like combine harvesters, sewing machines, more recently car-building robots, automated vast swathes of manual labor. Maybe AI won't be the deciding technology for automating what's left, but advances in robotics may well be.
Who tells - or, say, programs - the AI what to change every time stakeholders want a button moved or an endpoint to contain some new field on response? Is leadership going to do all that work too?
I see where you are going and may I suggest this thought experiment: let's say there is a team of 4 developers. AI automates 50% of their work, but there is still a lot of fine-tuning and certainly a lot of time in meetings with co-workers. But, this team has more free time so their company gives them more work. Great, so they are still working at capacity now. However, let's suppose AI gets to 80% automation, and there is no more additional work their company can offer them. That's when one or two, or more of them are at risk of losing their jobs.
I think that's a possibility, but as a senior dev myself (that makes me biased I guess) I've found that as time goes on coding is less and less of the job, and communication and creative discussions take over. These are just to figure out what stakeholders want. I have a hard time imagining an AI automating this part of my job, although I suppose it could happen.
mmh. Maybe it's time to start training AI USER-E things, which can demand and consume whatever the Dalees produce.. like Douglas Adams' Electrical Monk - then link these into each other, and go on with our lives :/ if there's any energy left..
This is surprising to me but probably intuitive considering the market forces. Creative jobs are highly specialized and therefore expensive. (Low supply + high demand = high cost). On the flip side, labor-intensive jobs are not specialized and therefore cheap (overall, I know supply is an issue in the US right now at least).
If the market is trying to reduce costs it will start with the highest first, especially those with the highest barriers to entry as far as skill development. It probably doesn't help that it appears computer vision is advancing slower than other areas of AI.
Great post. Really enjoying your new writing.