The same way people generally equate luddism with anything. By entirely misunderstanding what it was to make a point that sounds snappy without all of that boring understanding history stuff.
You are simultaneously arguing for a more complex and nuanced testing approach that demands much higher quantities of data as a result, and also against RCTs, which perhaps rightly you've identified as having suffered from the same kind of cost disease as all other health care in the USA. I can't help but feel like you've identified the wrong root cause here.
The problem is that people are not listening to each other. Part of this is just working together is hard, but a bigger part is that there is status associated with being "in charge" of things as founder. The desire to feel like you are controlling the outcome and people need to take instruction and direction from you rather than working together to find a path through the forest.
100% this. I read these caveats in new models and all I hear is "we made sure this model has no idea about computer security." Such a weird thing to brag about.
Actually it's the opposite. Per mm of silicon it's massively less efficient and making enough chips and powering them is a major bottleneck right now. Worse, scaling to larger models requires more of our absolute best quality silicon manufacturing, where e.g. an H200 mostly just needs more memory.
Oh sure the EU is terrible at all technology policy but it also seems like this is sort of a don't interrupt your enemy when they are making as mistake situation.
If the US commit to handicapping their AI industry like this it's going to destroy the competitiveness of those companies globally. All of those US spending commitments on data centres etc are going to collapse, or americans will need to pay 2x the token cost of the rest of the world. Both very bad options.
There are legal forms of killing. It is only via the application of our legal institutions that criminality is decided. No specific act on its own is criminal, there is no Platonic Crime.
To be truly impartial I don't think you should even evaluate their claims directly. This allows them to focus the comparison on things they care about, not the things you care about. Instead you should decide what problems you need solving, and evaluate solutions to those problems against your own rubric.
A large part of vulnerability analysis is just having the time to crunch through enough possibilities. Expertise and smarts definitely speed this up but there's a lot of just turning the crank until something falls out. Even a relatively dumb model with some good prompting will find vulnerabilities if you ask it to and give it the time and resources to do so.