4 ms·

Deep learning / ML / AI have long-term implications to affect essentially all non-manual jobs, including the subsets of knowledge workers, office jobs and prog

Deep learning / ML / AI have long-term implications to affect essentially all non-manual jobs, including the subsets of knowledge workers, office jobs and programming. (That’s in-addition to putting 90-95% of truckers, the no. 1 job in the US, and ride-for-hire workers out of work within 10-20 years.) Self-programming systems, given enough computing power and tech progress, will be A Thing™. At that point, we will be very close to one or more technological singularities as well, which will have even larger implications and knock-on effects to productivity, job availability and technological era. We really need an universal basic income (UBI) at some point, or there were will be billions of people but very few, livable wages jobs as the billionaires and potential trillionaires vacuum up an increasingly skewed power-law distribution of the new wealth created, while billions of people live in favelas. See also: the TED talk on deep learning: https://www.ted.com/talks/jeremy_howard_the_wonderful_and_terrifying_implications_of_computers_that_can_learn https://www.ted.com/talks/jeremy_howard_the_wonderful_and_te...

thu2111 7y ago
I used to believe this but don't any longer. For 'cookie cutter' business software development you don't need AI to write the software. Better frameworks or even just a respin of tools like MS Access would be much more impactful. Yet we see relatively little progress in the space of RAD tools, which were a big thing in the 1990s but have largely been killed off by the web. How many frontend devs even use a visual UI designer these days? Mostly they write HTML either by hand or using mashups of JavaScript and XML like in React. If anything dev productivity has gone backwards, and expecting AI to fix that doesn't make much sense. To start with just one big problem - even the best AI can't really conceptually understand something like a business domain, and business customers often struggle to understand what software can do or articulate what they really want. AI will contribute in some areas, probably with small workflow and performance optimisations. It isn't going to radically change anything.

thu2111 7y ago

I used to believe this but don't any longer. For 'cookie cutter' business software development you don't need AI to write the software. Better frameworks or even just a respin of tools like MS Access would be much more impactful. Yet we see relatively little progress in the space of RAD tools, which were a big thing in the 1990s but have largely been killed off by the web. How many frontend devs even use a visual UI designer these days? Mostly they write HTML either by hand or using mashups of JavaScript and XML like in React. If anything dev productivity has gone backwards, and expecting AI to fix that doesn't make much sense. To start with just one big problem - even the best AI can't really conceptually understand something like a business domain, and business customers often struggle to understand what software can do or articulate what they really want. AI will contribute in some areas, probably with small workflow and performance optimisations. It isn't going to radically change anything.