4 ms·

>Some estimates suggest that half of all white-collar jobs will disappear as artificial intelligence advances I find this so odd. Is there an imagining that th

>Some estimates suggest that half of all white-collar jobs will disappear as artificial intelligence advances I find this so odd. Is there an imagining that there is some imminent trivial threshold at which AI will stop improving? If AI improves at the current pace, all jobs will disappear as the very minimal floor of change.

AnimalMuppet 1y ago
All jobs will disappear, possibly within one year? Um, no. No way. Not even all white collar jobs, and certainly not all jobs.
- catigula 1y ago
  Unfortunately this is an argument from incredulity.
  - AnimalMuppet 1y ago
    Well, you don't present much more than an argument from credulity, which isn't all that convincing either. Even if I accept your premise (AI continues to improve at the existing pace), I don't see how AI is going to replace, say, welders, in a year. Or even bankers. (Yeah, an AI might be able to do the financial part. But bankers - as opposed to tellers - are about relationship with customers, and the AI isn't going to substitute there.)
    - catigula 1y ago
      I made two basic arguments: 1. The exponential curve of AI progress 2. The lack of any mechanism that will immediately wall off AI progress at this specific exponential part of the curve >I don't see how AI is going to replace, say, welders, in a year. This article is about white-collar jobs. However, if human-level intelligence (let alone surpassing human-level intelligence) just becomes a matter of compute, robotics will quickly follow. That being said I'd expect this to be years, not a year. > But bankers - as opposed to tellers - are about relationship with customers, and the AI isn't going to substitute there Are you sure about that? Because even GPT-4o is an example of how people are very hungry for relationships presented by these AIs.
josefritzishere 1y ago
There is one reasonable basis on which AI will not improve signifigantly. AI already effectively has access to all data ever created and it can't tell you how many Bs are in Blueberry.
philipallstar 1y ago
> Is there an imagining that there is some imminent trivial threshold at which AI will stop improving? AI seems to need more and more power and expense thrown at it to get from say a 70% answer to an 80%, then just as much again to an 85%, then just as much again to an 87.5%. Speed of progress in the lower percentages is not an indicator of speed in the higher percentages.
- catigula 1y ago
  Whatever the power/expense GPT-4 -> GPT-5 is such a large improvement that I can scarcely say that it is very different from GPT-3 -> GPT-4. If this happens only one more time, you're going to see really large changes.
taylodl 1y ago
If AI improves at the current pace, then few jobs will disappear. Yes, I'm well-aware that some companies have engaged in cyclic belt-tightening and are trying to sound forward-thinking and say it was a result of AI, but there's no evidence that AI actually caused any of those job losses.

AnimalMuppet 1y ago

All jobs will disappear, possibly within one year? Um, no. No way. Not even all white collar jobs, and certainly not all jobs.

catigula 1y ago
Unfortunately this is an argument from incredulity.
- AnimalMuppet 1y ago
  Well, you don't present much more than an argument from credulity, which isn't all that convincing either. Even if I accept your premise (AI continues to improve at the existing pace), I don't see how AI is going to replace, say, welders, in a year. Or even bankers. (Yeah, an AI might be able to do the financial part. But bankers - as opposed to tellers - are about relationship with customers, and the AI isn't going to substitute there.)
  - catigula 1y ago
    I made two basic arguments: 1. The exponential curve of AI progress 2. The lack of any mechanism that will immediately wall off AI progress at this specific exponential part of the curve >I don't see how AI is going to replace, say, welders, in a year. This article is about white-collar jobs. However, if human-level intelligence (let alone surpassing human-level intelligence) just becomes a matter of compute, robotics will quickly follow. That being said I'd expect this to be years, not a year. > But bankers - as opposed to tellers - are about relationship with customers, and the AI isn't going to substitute there Are you sure about that? Because even GPT-4o is an example of how people are very hungry for relationships presented by these AIs.

josefritzishere 1y ago

There is one reasonable basis on which AI will not improve signifigantly. AI already effectively has access to all data ever created and it can't tell you how many Bs are in Blueberry.

philipallstar 1y ago

> Is there an imagining that there is some imminent trivial threshold at which AI will stop improving? AI seems to need more and more power and expense thrown at it to get from say a 70% answer to an 80%, then just as much again to an 85%, then just as much again to an 87.5%. Speed of progress in the lower percentages is not an indicator of speed in the higher percentages.

catigula 1y ago
Whatever the power/expense GPT-4 -> GPT-5 is such a large improvement that I can scarcely say that it is very different from GPT-3 -> GPT-4. If this happens only one more time, you're going to see really large changes.

taylodl 1y ago

If AI improves at the current pace, then few jobs will disappear. Yes, I'm well-aware that some companies have engaged in cyclic belt-tightening and are trying to sound forward-thinking and say it was a result of AI, but there's no evidence that AI actually caused any of those job losses.