Thing is, fingerprinting in general is not a new field, and people have been throwing neural nets at it for a long time. It has always been a pretty sketchy technique and sensitive to overfitting. You don't need people trying to shield themselves, things just change in ways that screw up fingerprints. They are notoriously difficult to keep up to date and mostly seem to survive as part of packages sold to big IT departments as 'this will detect things!', .

If you alter the code and distribute a binary, you are required to distribute the code too. You can only get around distribution if you only use the code internally, which is how so many 'internet' companies get around GPL since they keep everything server side.

These companies just love 'move fast and break things', unless the moving fast might break their things.
 
By their own logic, we should regulate now and work out problems later.

Which we probably get for free with simple network behavior.

I am not sure which side would really be 'always one step ahead' here. In order for this to work they would need to constantly retrain their CNN since any change in any application or websites's data usage pattern would throw it off. OS updates would also require retraining.
Boiler plate code would also throw it off, and I imagine the majority of websites would be indistinguishable. Things like user settings or adblockers would also throw it off.

So.. that sounds more like they developed a way for two isolated browser tabs to communicate with each other, rather than a sandbox sneaky figuring out what is going on outside of it?

So... having anything other than the web browser running?

All I really see there is that closer you are to things where getting the right answer does not matter, the more useful AI is. No the ringing endorsement you think it i .

Which is not something to look forward to. requirements engineers are already poorly paid and under hired for their workloads. Industry isn't about to start respecting their work just because AI cuts out other types..

That age group got brain damage from lead poisoning instead.

Hrm.

While it would not have a snowballs chance in hell of working, I would be curious to see someone try some kind of unjust enrichment claim. Contracted work was preformed for compensation, but also preformed under the company's written policies, and one could argue that if they would not have performed the work under the new policies, and the company continues to benefit from the work performed, that the benefits were fraudulently gained. Would not work, but it would be interesting to see the legal arguments play out.

I wonder how well they handle less common cases though. The classic problem with these systems, while they are always improving, is they follow the 'most of the time it is the most common case' line of reasoning, so most of the time they are right.
 
Though the recurring problem with these tools is that since they tend to give the most common answers, they can easily become a source of bias, making changing course even more difficult.

Eh, all it did was codify how law enforcement has always operated in the US, all the way back to the founding and writing of the constitution. The police in the US were a combination of essentially private corporate armies and regional gangs that went 'legitimate' since there was nothing to fight them with.

So basically anything that makes him or his followers feel stupid?

Thing is, I would argue 'iffy funding' is part of the point of the NSF. It handles funding for all the experimental stuff that the private sector or DARPA is not interested in, It fill the gaps and takes on risk that other institutions and funding vehicles handle poorly.... and also spackles over other problems like, as you point out, keeps grad students employed since otherwise it would making such education little more than a hobby of the already wealthy like it used to be.