Comment Re: Comet's Bold Strategy: Free AI Browser with Ma (Score 1) 27
(oops, by 'it', I mean some exploit you might encounter, not the browser itself.)
(oops, by 'it', I mean some exploit you might encounter, not the browser itself.)
Am I mistaken or is this a bit of closing the barn door after the horse has escaped? Is there no valuable info on your system that it could have exfiltrated other than your
Yeah. Ignoring this giant red flag seems foolish in the extreme. Don't even know if they keep current with the underlying (here Chromium) security patches, which is a basic necessity new browsers often miss.
"Perplexity classified it as 'no security impact.' " mandates sandboxing it (or using it on a toy system) at a minimum, or just taking a wide berth.
Major undisclosed conflicts of interest! See https://ancillary-proxy.atarimworker.io?url=https%3A%2F%2Fpubpeer.com%2Fpublicatio... (which you'd be aware of if you had the PubPeer plugin!) Lead author Giuseppe Battaglia is cofounder of a company commercializing underlying technology. So there's even greater risk this research, like all of that also based on the amyloid hypothesis, rests on fraudulent foundations. https://ancillary-proxy.atarimworker.io?url=https%3A%2F%2Fwww.statnews.com%2F2025%2F...
BINGO!
I seriously think Asimov's laws are pretty damn good.
Implementable? We don't know how to...yet.
But I am feeling my way toward some reasonable regulations. A first draft:
Any general purpose AI (or class of AI) that appears to have the ability to kill a human directly or indirectly must
1)be registered in a public database and
2)restricted to a controlled lab environment
unless it has been imbued with the restrictions of Asimov's three laws in a non-removable, non-overridable way. All instances of any class of AI that had been deemed to be thus restricted must be immediately recalled if an instance avoidably violated one of the laws anyway. A judge may order such a recall in the same way that a judge may sign a warrant. or issue an injunction? (Or maybe make it more like a subpoena?)
Just because we aren't currently able to build an AI that can meet such requirements does NOT mean such regulations aren't appropriate.
Classes of AI would include:
For self-driving vehicles, all of a company's vehicles, and vehicles of other companies that used that companies AI, would be in the same class.
...if they were willful, intentional, or reckless, and if they weren't, they still owe you $500 if you ask them to disclose to you any personal information disclosures and they don't. In either case, this only applies if you're Californian.
The relevant law is CALIFORNIA CIVIL CODE SECTION 1798.80-1798.84 which you can find here:
http://www.leginfo.ca.gov/cgi-bin/displaycode?section=civ&group=01001-02000&file=1798.80-1798.84
(1798.83 and 1798.84 are the most relevant.)
I'm currently (still!) suing TD Ameritrade for covering it up when they were hacked and the names, addresses, SOCIAL SECURITY NUMBERS, etc, of 6.4 MILLION customers were compromised. (See amtd.elvey.com.)
(b) Any customer injured by a violation of this title may
institute a civil action to recover damages.
(c) In addition, for a willful, intentional, or reckless violation
of Section 1798.83, a customer may recover a civil penalty not to
exceed three thousand dollars ($3,000) per violation; otherwise, the
customer may recover a civil penalty of up to five hundred dollars
($500) per violation for a violation of Section 1798.83..
The solution to a problem changes the nature of the problem. -- Peer