News for hackers. Stuff that matters.
Re:Most AI value will come from broad automation, not from R&D by ajb on Saturday March 22, 2025 @04:44PM (#43448464)
Yeah

The historic benefits of technology come from eliminating the waste of human effort:

- Motors and engines removed the waste of using our muscles as an energy source

- Computers removed the waste of people's time doing calculations and admin manually

- Electronic Communications removed the waste of carrying information in physical form

What's the most prevalent form of waste today? I think it's dishonesty and corruption. And AI seems set to make more of it.

Re:Most AI value will come from broad automation, not from R&D by sadeshmukh on Sunday March 23, 2025 @02:22AM (#43451188)
Does AI not also automate away repetitive tasks? This seems to be a disingenuous comparison.
Re:Most AI value will come from broad automation, not from R&D by bluefirebrand on Sunday March 23, 2025 @11:19AM (#43453505)
AI is strictly worse at automating repetitive tasks than traditional software, because AI output is non-deterministic

If a task is repetitive then the output is predictable so it should be automated deterministically, not with fuzzy AI bullshit

Re:Most AI value will come from broad automation, not from R&D by datadrivenangel on Sunday March 23, 2025 @12:53PM (#43454104)
LLMs are often worse, yes, but there are decision tree based machine learning models that can learn to be deterministic for classification if your training data accurately reflects the actual problem data.
Re:Most AI value will come from broad automation, not from R&D by bluefirebrand on Sunday March 23, 2025 @01:24PM (#43454333)
This sounds so stupidly over engineered for a majority of problems spaces

Like teaching an AI model to learn arithmetic instead of just building a calculator