4 ms·

Right. It's the AI that is the problem. I have another use case for LLM's, I haven't thought of before: absolution of responsibility. The public is already pri

Right. It's the AI that is the problem. I have another use case for LLM's, I haven't thought of before: absolution of responsibility. The public is already primed to focus on AI in such cases.

b112 3y ago
In the early 90s, when every business was moving full out into computing, I'd call for an issue. Sometimes the person helping me, would say "I'm sorry, the computer won't let me do that". This seemed to grow as a response, until, to protect myself, I started pushing back. "I don't care, you guys owe me $x, and telling me your computer won't let you do it is absurd. Pay me." Most people would then get a manager on the phone, who would ape the "computer was at fault" line, but after pushback.. surprise!, there was a way to resolve things. Point is, abdicating responsibility is an easy out for the lazy and for corporate greed. And AI will surely be used in this fashion, because people love an out. Amazon is already the worst of this, if you have an issue that doesn't fit in a premade solution, good luck getting any fix! Taking humans out of the chain is always bad. Just look at the UK post office scandal, for the dark side of this.
- eurekin 3y ago
  Calls very close to home... It really feels like playing a whack-a-mole, never really progressing towards any reasonable level of responsibility
dspillett 3y ago
> absolution of responsibility There are some still fighting the fight, but this is already happening¹ when it comes to failure to attribute and other licensing issues. “I didn't copy/misattribute/other your code, an AI trained in part by it was ‘inspired’ by it amongst other code”. ---- [1] unless you believe the assurances that an LLM will never regurgitate chunks of training material, like the image based models have been shown to

4 ms·

Right. It's the AI that is the problem. I have another use case for LLM's, I haven't thought of before: absolution of responsibility. The public is already pri

by eurekin 3y ago

Right. It's the AI that is the problem. I have another use case for LLM's, I haven't thought of before: absolution of responsibility. The public is already primed to focus on AI in such cases.

b112 3y ago
In the early 90s, when every business was moving full out into computing, I'd call for an issue. Sometimes the person helping me, would say "I'm sorry, the computer won't let me do that". This seemed to grow as a response, until, to protect myself, I started pushing back. "I don't care, you guys owe me $x, and telling me your computer won't let you do it is absurd. Pay me." Most people would then get a manager on the phone, who would ape the "computer was at fault" line, but after pushback.. surprise!, there was a way to resolve things. Point is, abdicating responsibility is an easy out for the lazy and for corporate greed. And AI will surely be used in this fashion, because people love an out. Amazon is already the worst of this, if you have an issue that doesn't fit in a premade solution, good luck getting any fix! Taking humans out of the chain is always bad. Just look at the UK post office scandal, for the dark side of this.
- eurekin 3y ago
  Calls very close to home... It really feels like playing a whack-a-mole, never really progressing towards any reasonable level of responsibility
dspillett 3y ago
> absolution of responsibility There are some still fighting the fight, but this is already happening¹ when it comes to failure to attribute and other licensing issues. “I didn't copy/misattribute/other your code, an AI trained in part by it was ‘inspired’ by it amongst other code”. ---- [1] unless you believe the assurances that an LLM will never regurgitate chunks of training material, like the image based models have been shown to