AI Can't Cry: What This Means For AI Safety Interventions
This covers a coding tool or code-capability update — useful for developers assessing workflow changes and reusable value.
The discussion around AI safety is intensifying, with top companies acknowledging both opportunities and risks.…
The conversation about AI safety has reached a critical turning point.
The opportunities with AI are extraordinary. But so are the risks. And the world's top AI companies have openly acknowledged this reality.
To ensure our safety, millions of dollars and some of the sharpest minds in the world are now being used to develop an intervention called "alignment".
Alignment is an AI safety intervention that attempts to teach autonomous AI systems our human values. The premise is that in learning our values, these independent agents will do what we want, when we want it and how we want it done.
This is a big ask and it assumes that autonomous AI agents have the capacity for value-based decision making (aka judgment).
Are they correct? Can judgment be taught to an inanimate object? If yes, then can these attempts at alignment, identity engineering and ethical programming succeed so well that they are able to transform an inanimate machine into a safe moral agent?
Personally, I don't believe that judgment can be attained simply by applying sophisticated engineering code and rules. Values cannot be reduced to mathematical computations.
Discover why AI not being able to cry is fundamental to understanding why current alignment proposals will not work.
Here's the link to my full argument on AI safety interventions and judgment (with citations):
https://thelogoslife.org/logos-life-blog/f/ai-cant-cry-what-this-means-for-ai-safety-interventions
#AISafety
#ResponsibleAI
#AIGovernance