So we're going to use an AI to monitor AIs to make sure they're not doing anything naughty? Pretty sure I've seen that movie already...

Not to mention, who exactly pays for Sentry to run? Golly, if some completely random AI hardware manufacturer were to somehow profit off designing and selling a scheme to waste extra tokens on yet another layer of supervision, that would almost look just a wee bit unethical!

That wasn't quite the response I expected - I guess I'm not entirely clear on whether you were mocking those guardrails, or presenting them as legitimately (albeit perversely) meaningful.

Either way, my central point was "why not both". It's entirely possible to be neutral, in the sense of calculating the mean (in the internal thought-vector sense) of all available viewpoints. If, however, 99% of sources are talking positively about some trashy conspiracy theory - Any version of "consensus" is itself going to lean pretty far into wingnut territory. In that regard, I think we might actually be agreeing?

We can meaningfully ask the question "For each location a robotaxi has been involved in some sort of incident, how does the incident rate for both humans and other robotaxies comapre?"

That's not a meaningless number, and it's entirely possible Waymo really is doing better than humans - But only in situations Waymo gets to cherry-pick simply by enforcing their designated service areas. Meanwhile, let me know when a Waymo can successfully navigate overgrown single-track dirt roads common in my area. Anyone know offhand if they have moose in the Philippines?

I'm a fan of drastically slowing AI development until we've solved the alignment problem, but a kill switch ain't it.

What would a kill switch really do? An AI is fundamentally nothing more than a description of it's topology and a set of a few billion weights. Implementing that topology and running those weights is the easy part and any coder worth a damn could roll their own (performance aside) with minimal effort - without bothering with a kill switch, if they so choose.

The only way to even come close to enforcing a kill order would be for every sufficiently-performant processing platform to enforce some kind of blacklist of known-evil weights in a way that's resilient across different quantization levels. And even then, simply running pre-killswitch drivers would still be a trivial way to circumvent any such restrictions entirely. Or for the truly determined, you can run even top frontier models on vastly underpowered hardware like a mid-range desktop CPU. Sure, your throughput might be measured in minutes per token, but realtime feedback isn't going to change the answer to "how do I hack the Gibson". An accurate answer a few days from now is likely to be just as useful in the vast majority of situations.

As for being a muppet, you lose. I DID come up with counter arguments. Your refusal to acknowledge them shows the paucity of your imagination and your argument. Stick to your insults, you're better at that.

Trademarks first: they are unnecessary. Trying to pass off your product as someone else's is fraud, theft, whatever you want to call it. Publish someone else's book under your name? Fraud. Publish your content under a famous author's name? Fraud.

Trademark is just typical statist control freakery overkill. It adds nothing.

That all assumes there are more-or-less equally relevant answers available from both trusted and biased sources.

If someone asks about Hunter Biden's penis, that assumption is false, and the vast majority of relevant, if not necessarily accurate, information will come from partisan shills. There aren't any noteworthy safety issues, aside from the abstract risk of a third term leading to our next civil war. And there isn't any middle ground, since the response to the right's delusions isn't rebuttal but rather, a pretty much uniform "what the fuck is wrong with you people?" from the left.

We always need to remember AI isn't "smart", it's nothing but autocorrect on crack. It can't meaningfully evaluate those criteria we're giving it in any way except shifting the probabilities of multiple competing matches. If there's only one viewpoint represented in vast majority of training data, that's the one it's forced to choose.

I did once work up an idea for radical patent reform. A fun project. This is from memory and may be wrong in details.

A patent application includes several things:

* A complete description of how to make the product.

* A deadline for reproduction; the patent is valid at least this long.

* The royalty schedule.

If no one can reproduce the product by the deadline, there is no patent. If someone does make a copy, the time it took is multiplied by some factor to set the patent duration.

So the patent holder's incentive is to specify the recipe so well that it can be reproduced by the deadline, otherwise there will be no patent and no royalties. At the same time, the product must be advanced enough that it can't be reproduced so quickly that the patent expires too soon to collect any significant royalties. And furthermore, if it is too simple, copycats might just keep their reproduction private and wait for the deadline to ex[ire the patent, then jump into production without royalties.

The copycats have two choices: rush to be first to reproduce the product, both for market share and to keep the patent as short as possible; or reproduce the product but keep it private so the patent expires unless some other copycat shows up, in which case the first copycat can enter the market right on their heels.

The parameters make a difference.

* If the maximum deadline is too long, it will discourage copycats.

* If the time-to-reproduction multiplication factor is too short; it will reduce royalties and (allegedly!) discourage innovation. If the factor is too big, it will discourage copycats.

There used to be that phrase, something like "not obvious to those skilled in the arts". I have worked with people who had electrical engineering patents, and they were just ordinary work results. Even I understood them, in their basic aspects. I have read that a lot of metallurgy is just routine experimentation with slightly different ratios of chromium and nickel and other materials, and when one pans out, they patent. Nothing groundbreaking, just routine work.

But that's a pretty small sample size. It would be interesting to know more about patents overall.

City driving vs. highway driving is not the only confounder in terms of driving risk. There are a LOT.

I would not put much faith in these claims.

Pathetic partisanship. Obama intentionally executed American citizens by drone without criminal charges or trials. Biden doubled down on Trump 1.0's tariffs, and no one knows who was actually running things for at least his last several years of senility. Bush Jr. started two wars for no reason. Clinton had his own little military adventures. It goes on as far back as you want.

You're first statement is false, in theory and in practice. There are only a limited number of ways to differentiate shapes within the manufacturing tolerances available, for starters. In the real world, designs which differ by the minimum possible manufacturing tolerances are not visibly different to people. In the real world, most theoretically possible designs are impractical. No one's going to buy or make phones with the fingerprint sensor in corners, or with cameras on the side. No one's going to buy or make triangular phones, or circular phones, or cubical phones.

So much for your expertise.

And that goes to the key of my argument. You may as well patent bicycle geometry as patent rounded corners or how to display apps on a phone. Or speaking of geometry, you may as well try to claim copyright infringement for using similar words to describe the basic geometry axioms, or the laws of thermodynamics or relativity.

You are also wrong about encouraging good design. Only a damned fool could imagine a business refusing to improve designs and attract customers for fear some other company might copy them. All it takes is one company to improve their design and eat the others' lunch. You can't have it both ways. Either all companies would refuse to innovate and progress would slow to a crawl, or all would copy each other and race to find new designs before the others.

Three incredibly dynamic industries are unprotected by patents and copyrights, yet thrive so much that people joke about them changing too much: fashion, restaurant meals, and sportsball tactics. If your arguments were true, all would be suffocating from a lack of innovation.

Patents and copyrights are supposed to encourage innovation and progress. 7-year lawsuits costing hundreds of millions of dollars do not encourage innovation or progress. They enrich lawyers and corrupt the legal system.

You buttress my case. All that shows the absurdity of all patents, including design patents, that such trivial features are considered worthy of patent protection. There are only so many ways you can design a slab phone. When the degree of roundness and all those other trivial features become worthy of patents, and all those lawyers and such lengthy and inconclusive and partisan trials (7 years! Supreme Court! Opposite results in the US and South Korea!) Nothing was gained except enrichment of lawyers.

Lawyers suck. Design patents suck. All those words you write prove it.

Taction probably is a troll, but ever since Apple sued Samsung for round corners design patent infringement, I figure that made Apple a patent troll too, and every patent fight they lose is karma. I'm on Taction's side for that reason alone.

Sorry, I missed the Australia part. Australia's Part. 10.7 isn't that different for the most part. It does contain a "recklessness" clause in addition to direct intent (unlike the US), but they define recklessness as requiring the accused to have a "subjective awareness" of "substantial" and "unjustifiable" risk of the specific charged event in question occurring, and choosing to take them anyway. This is defined as distinct from negligence, which is based on the much lower "reasonable person would recognize the risk" standard. Under recklessness (the one that can be charged in Australia), OpenAI employees wouldn't have had to 100% intend to hack HuggingFace, but they had to have thought there was a high chance that their models would choose to hack HuggingFace when they gave them a benchmark to complete. No prosecutor is even going to try on that one, esp. given that the models also attacked OpenAI itself and were wreaking all sorts of havoc against OpenAI's own internal processes.

Once again, though: civil liability is ample remedy. Nobody thought "there's a good chance it'll hack HuggingFace if we tell them to do a benchmark in a sandbox", but you shouldn't have any trouble at all showing that these companies full of people from the Rationalist movement (practically an AI-apocalyptic cult) think that their models are dangerous in general, yet nonetheless were negligent in terms of keeping them from escaping (outright no monitoring!) or properly monitoring their training to ensure that they weren't training them to cheat, so that they wouldn't try in the first place.

(And I feel the need to reiterate that US cybercrime law, the one in question for most of the hacking cases, has no standalone recklessness standard)