4 ms·

It does look like knowledge work will go first, it was surprisingly the easiest. Taking control of $100,000 robots to manufacture does have value at the high-e

It does look like knowledge work will go first, it was surprisingly the easiest. Taking control of $100,000 robots to manufacture does have value at the high-end, but still can't compete with humans at the low end. I wonder if this can bring some human-friendly Manufacturing back to certain locations. The environmental impact of outsourcing is incredibly large, making things closer to home with newly freed hands is probably better. It might also help with the obesity crisis having people spend time sitting 10 hours a day less frequently.

dinvlad 3y ago
Nah, it’ll be just a new, higher level of abstraction (and provided we even solve the hallucination problem, which is not a given). Engelbart’s and Parkinson’s laws at it once again. In other words, the new work fill the same time, and we’ll just be doing 10x of it (if we can solve the current challenges with LLMs). The human mind is unsurprisingly adaptable to changing conditions and great at abstract thinking, which is what ironically distinguishes it from these machines. We also tend to severely underestimate our own abilities and overestimate the importance of the tools.
- rootusrootus 3y ago
  > We also tend to severely underestimate our own abilities and overestimate the importance of the tools. Even on HN, this is rampant. I wonder, is it because people are so used to interacting with most other humans solely through a chat interface that they perceive ChatGPT as being close to a real AGI, almost indistinguishable from human intelligence?
  - dinvlad 3y ago
    Pretty much :-) Text generation (and rudimental comprehension) is perceived to be a sign of higher intelligence, simply because it's our natural human UI. Ironically, we tend to limit our imagination on what _we_ could do that computers won't, while simultaneously overuse it to imagine what they _could_ do but won't. One very good example of this is humor - LLM models at present are terrible at generating anything but dad jokes, and are even worse at their attempts to explain "deep" humor. Gary Marcus had a recent episode on it, very interesting insights from top experts in the field: https://podcasts.apple.com/us/podcast/can-ai-make-you-laugh/id1532110146?i=1000613196164 https://podcasts.apple.com/us/podcast/can-ai-make-you-laugh/.... They make a case that LLMs were specifically optimized for text generation and not explanation, while humans are almost the opposite. I believe somewhere in there is also our tendency for confirmation bias and overlooking imperfections where we don't expect some. AI may generate nonsense details, but when we look at the overall text, we don't notice them since we're just awed by the output, and think if it looks flawless it must be flawless. Same thing as with mentalist tricks. And finally, we only think of our intelligence as thoughts carefully animated into sentences, while there's this whole "iceberg" level of feelings and unconscious cognition, which we don't even notice ourselves, but it would be extremely hard to reproduce in AI and we have not even the slightest idea how.
Closi 3y ago
> Taking control of $100,000 robots to manufacture does have value at the high-end, but still can't compete with humans at the low end. The problem with robots isn't the physical cost vs a human, it's the task flexibility & cost to program. This is something that could also theoretically be solved with AI in the (fairly close) future.
- ForestCritter 3y ago
  I worked in a semi automated assembly line. We,as workers were pretty good at keeping the robots running and knowing their quirks and my supervisor was even more knowledgeable. Despite this we still had to call up an engineer daily to trouble shoot why the robots were revolting. AI is not remotely near solving this.
  - Closi 3y ago
    Disagree - AI is getting much closer to solving this fast. Latest generation robotic arms with vision are taking on a greater variety of tasks (e.g. bin/case picking from a mixed tote), more reliably, with a better ability to adaptively recover failed tasks. The progress here are about not needing parts/items to be in a predictable/consistent location and still understanding how to handle/retrieve them and perform actions on them. This stuff is in sites right now.
    - ForestCritter 3y ago
      picking stuff from bins is a far stretch from a full assembly line production. Latest generation arms are surely being developed but the are far from fitting affordability profiles for the average American factory. It sounds great when you are reading it but fabrication is not like picking stuff from bins. The colossal amount of programming for every human task that is performed in a factory that builds compressors would be a dauntingly expensive enterprise even without counting the physical robotic equipement that would be needed. Theory and practice don't always correlate.
      - 3y ago
yoyohello13 3y ago
Whenever I hear about AI taking knowledge work, I always think of that show Dirty Jobs. Jobs like 'sewer cleaner' are not going to be automated anytime soon. So eventually all the fun jobs will be AI driven and humans will be left cleaning shit off the walls.
- HKH2 3y ago
  Then we will soon see what we really are.
- elforce002 3y ago
  Interesting. Everyone will be the same under this new era. This sounds awfully like communism to me, hehe.

dinvlad 3y ago

Nah, it’ll be just a new, higher level of abstraction (and provided we even solve the hallucination problem, which is not a given). Engelbart’s and Parkinson’s laws at it once again. In other words, the new work fill the same time, and we’ll just be doing 10x of it (if we can solve the current challenges with LLMs). The human mind is unsurprisingly adaptable to changing conditions and great at abstract thinking, which is what ironically distinguishes it from these machines. We also tend to severely underestimate our own abilities and overestimate the importance of the tools.

rootusrootus 3y ago
> We also tend to severely underestimate our own abilities and overestimate the importance of the tools. Even on HN, this is rampant. I wonder, is it because people are so used to interacting with most other humans solely through a chat interface that they perceive ChatGPT as being close to a real AGI, almost indistinguishable from human intelligence?
- dinvlad 3y ago
  Pretty much :-) Text generation (and rudimental comprehension) is perceived to be a sign of higher intelligence, simply because it's our natural human UI. Ironically, we tend to limit our imagination on what _we_ could do that computers won't, while simultaneously overuse it to imagine what they _could_ do but won't. One very good example of this is humor - LLM models at present are terrible at generating anything but dad jokes, and are even worse at their attempts to explain "deep" humor. Gary Marcus had a recent episode on it, very interesting insights from top experts in the field: https://podcasts.apple.com/us/podcast/can-ai-make-you-laugh/id1532110146?i=1000613196164 https://podcasts.apple.com/us/podcast/can-ai-make-you-laugh/.... They make a case that LLMs were specifically optimized for text generation and not explanation, while humans are almost the opposite. I believe somewhere in there is also our tendency for confirmation bias and overlooking imperfections where we don't expect some. AI may generate nonsense details, but when we look at the overall text, we don't notice them since we're just awed by the output, and think if it looks flawless it must be flawless. Same thing as with mentalist tricks. And finally, we only think of our intelligence as thoughts carefully animated into sentences, while there's this whole "iceberg" level of feelings and unconscious cognition, which we don't even notice ourselves, but it would be extremely hard to reproduce in AI and we have not even the slightest idea how.

Closi 3y ago

> Taking control of $100,000 robots to manufacture does have value at the high-end, but still can't compete with humans at the low end. The problem with robots isn't the physical cost vs a human, it's the task flexibility & cost to program. This is something that could also theoretically be solved with AI in the (fairly close) future.

ForestCritter 3y ago
I worked in a semi automated assembly line. We,as workers were pretty good at keeping the robots running and knowing their quirks and my supervisor was even more knowledgeable. Despite this we still had to call up an engineer daily to trouble shoot why the robots were revolting. AI is not remotely near solving this.
- Closi 3y ago
  Disagree - AI is getting much closer to solving this fast. Latest generation robotic arms with vision are taking on a greater variety of tasks (e.g. bin/case picking from a mixed tote), more reliably, with a better ability to adaptively recover failed tasks. The progress here are about not needing parts/items to be in a predictable/consistent location and still understanding how to handle/retrieve them and perform actions on them. This stuff is in sites right now.
  - ForestCritter 3y ago
    picking stuff from bins is a far stretch from a full assembly line production. Latest generation arms are surely being developed but the are far from fitting affordability profiles for the average American factory. It sounds great when you are reading it but fabrication is not like picking stuff from bins. The colossal amount of programming for every human task that is performed in a factory that builds compressors would be a dauntingly expensive enterprise even without counting the physical robotic equipement that would be needed. Theory and practice don't always correlate.
    - 3y ago

yoyohello13 3y ago

Whenever I hear about AI taking knowledge work, I always think of that show Dirty Jobs. Jobs like 'sewer cleaner' are not going to be automated anytime soon. So eventually all the fun jobs will be AI driven and humans will be left cleaning shit off the walls.

HKH2 3y ago
Then we will soon see what we really are.
elforce002 3y ago
Interesting. Everyone will be the same under this new era. This sounds awfully like communism to me, hehe.