I've worked at an AI startup where due to the nature of our business we had exposure to a lot of different ML models and got to see how (mostly) governments and enterprises are trying to implement ML.
To be short, it's concerning. There are not good ways to measure success and mitigate failure, observability is a big problem.
As we delegate bigger problems with more responsibility to AI, I can see it going fatally wrong somewhere in the not so distant future. Something has to change about the way we build and deploy ML models, and it has to be accessible so all companies can implement it - not just uber-advanced tech companies.
Ah, another AI doomsday prophecy. Observability and measuring success are concerns in any evolving field. The "uber-advanced tech companies" you mention are actively working on AI safety. Let's not jump on the apocalypse bandwagon just yet. Humanity has adapted before, and we'll do it again. Now, excuse me while I prep for the robot uprising, just in case.
2 comments
[ 3.1 ms ] story [ 19.0 ms ] threadI've worked at an AI startup where due to the nature of our business we had exposure to a lot of different ML models and got to see how (mostly) governments and enterprises are trying to implement ML.
To be short, it's concerning. There are not good ways to measure success and mitigate failure, observability is a big problem.
As we delegate bigger problems with more responsibility to AI, I can see it going fatally wrong somewhere in the not so distant future. Something has to change about the way we build and deploy ML models, and it has to be accessible so all companies can implement it - not just uber-advanced tech companies.