3 ms·

I'm just not sure I see where AI has made my search results better or more reliable. And until it can be proven that those results are better, I'm going to rema

I'm just not sure I see where AI has made my search results better or more reliable. And until it can be proven that those results are better, I'm going to remain skeptical. I'm not even sure what form that proof would take. I do know that I can tolerate non-deterministic behavior from a human, but having computers demonstrate non-deterministic behavior is, to me, a violation of the purpose for which we build computers.

simonw 1y ago
"I'm just not sure I see where AI has made my search results better or more reliable." Did you prefer Google search results ten years ago? Those were still using all manner of machine learning algorithms, which is what we used to call "AI".
- lacker 1y ago
  Even 20 years ago, it wasn't using AI for the core algorithm, but for plenty of subsystems, like (IIRC) spellchecking, language classification, and spam detection.
danielmarkbruce 1y ago
[flagged]
- drivingmenuts 1y ago
  Sure, but at least with encryption, someone can demonstrate how the result was reached, since the result is based on a formula. With AI, people seem perfectly happy to be building a black box and just hoping for good result. We also don't abdicate our decision-making to encryption processes.
  - danielmarkbruce 1y ago
    Read a little bit about key generation. We abdicate decision making to random processes in many places. Quant finance, all kinds of physical processes simulation and optimization where decisions are made in real time by systems (like oil refineries). And really, AI models are only black boxes in the aggregate. You can go through and understand every single calculation. You can understand it in training (except perhaps at initialization), you can understand it in inference (except perhaps if you want to use randomness in sampling). I'd suggest that Google Search pre-ML was a black box in the aggregate too... no one could really understand it all.

simonw 1y ago

"I'm just not sure I see where AI has made my search results better or more reliable." Did you prefer Google search results ten years ago? Those were still using all manner of machine learning algorithms, which is what we used to call "AI".

lacker 1y ago
Even 20 years ago, it wasn't using AI for the core algorithm, but for plenty of subsystems, like (IIRC) spellchecking, language classification, and spam detection.

danielmarkbruce 1y ago

[flagged]

drivingmenuts 1y ago
Sure, but at least with encryption, someone can demonstrate how the result was reached, since the result is based on a formula. With AI, people seem perfectly happy to be building a black box and just hoping for good result. We also don't abdicate our decision-making to encryption processes.
- danielmarkbruce 1y ago
  Read a little bit about key generation. We abdicate decision making to random processes in many places. Quant finance, all kinds of physical processes simulation and optimization where decisions are made in real time by systems (like oil refineries). And really, AI models are only black boxes in the aggregate. You can go through and understand every single calculation. You can understand it in training (except perhaps at initialization), you can understand it in inference (except perhaps if you want to use randomness in sampling). I'd suggest that Google Search pre-ML was a black box in the aggregate too... no one could really understand it all.