Ask HN: What's the pain using current AI algorithms?
Here is my Calendly https://calendly.com/shakascchen/agi-user-interviews with Zoom url on confirmation.
Questions:
As an AI application developer, what's the pain using current AI algorithms or using trained machine learning models?
When people (may be you or others) don't think AI really understand a concept even if it shows good performance on testing data of your application, what problems would probably happen?
What problems are there when you want your AI accumulating systematic knowledge that can be re-used for fast transfer and adaptation?
How do you deal with rapid requirement changes, big or small, when building or maintaing an AI application?
To really understand a concept means to understand all related concepts and logic, which further expands to understanding all concepts. Do you agree so? On your opinion, does it mean the better performance an AI have, the closer it is to artificial general intelligence or human-level AI?