How on earth can you "exhaustively" deidentify millions of chat logs that could contain literally any personal details, and presumably all without OpenAI's own employees also sifting through personal information in exactly the way they're claiming would be bad if others did it?

Science? Really? There's a lot of soft-brained, unscientific and technophilic pseudo-religion in the article.

Let's work with the argument's load-bearing phrase, "exploration is an intrinsic part of the human spirit."

There are so many things to criticise in that single statement of bias. Suffice it to say there's a good case to be made that "provincial domesticity and tribalism are prevalent inherited traits in humans", without emotional appeals to a "spirit" not in evidence.

Or if it's even still readable. Intel when retrieving the 486 tape-in for the Edison project had to bake the tapes in an oven to remove moisture, and then had ONE CHANCE at imaging the tape as it crumbled to dust going through the reader.

For sure. And that's why I think it's important to distinguish harm caused despite good intentions and reasonable practices being followed from harm caused because someone did not follow reasonable practices or actively chose to cut corners.

Just my personal opinion, but given the track record in this particular industry, I think there should be demonstrable intent by decision-makers to follow good practices, not merely a lack of evidence of intent to circumvent or cut corners. This is expected in other regulated industries, compliance failures are a big deal, and for good reason. I see no reason why similar standards could not be imposed on those developing and operating autonomous vehicles, and every reason they should be given the inherent risks involved.

Maybe this will be an area where the US simply gets left behind because of the pro-car and litigious culture that seems to dominate discussions there.

Reading online discussions about driving -- admittedly a hazardous pastime if you want any facts to inform a debate -- you routinely see people from the US casually defending practices that are literally illegal and socially shunned in much of the world because they're so obviously dangerous. Combine that with the insanely oversized vehicles that a lot of drivers in the US apparently want to have and the car-centric environments that make alternative ways of getting around much less common and much less available, and that's how you get accident stats that are already far worse than much of the developed world.

But the people who will defend taking a hand off the wheel to pick up their can of drink while chatting with their partner on a call home all while driving their truck at 30mph down a narrow road full of parked cars past a school bus with kids getting out are probably going to object to being told their driving is objectively awful and far more likely to cause a death than the new self-driving technologies we're discussing here. You just don't see that kind of hubris, at least not to anything like the same degree, in most other places, so we might see more acceptance of self-driving vehicles elsewhere too.

IMHO the only sensible answer to is separate responsibility in the sense that a tragedy happened and someone has to try to help the survivors as best they can from responsibility in the sense that someone behaved inappropriately and that resulted in an avoidable tragedy happening in the first place.

It is inevitable that technology like this will result in harm to human beings sooner or later. Maybe one day we'll evolve a system that really is close to 100% safe, but I don't expect to see that in my lifetime. So it's vital to consider intent. Did the people developing the technology try to do things right and prioritise safety?

If they behaved properly and made reasonable decisions, a tragic accident might be just that. There's nothing to be gained from penalising people who were genuinely trying to make things better, made reasonable decisions, and had no intent to do anything wrong. There's still a question of how to look after the survivors who are affected. That should probably be a purely civil matter in law, and since nothing can undo the real damage, the reality is we're mostly talking about financial compensation here.

But if someone did choose to cut corners, or fail to follow approved procedures, or wilfully ignore new information that should have made something safer, particularly in the interests of personal gain or corporate profits, now we're into a whole different area. This is criminal territory, and I suspect it's going to be important for the decision-makers at the technology companies to have some personal skin in the game. There are professional ethics that apply to people like doctors and engineers and pilots, and they are personally responsible for complying with the rules of their profession. Probably there should be something similar for others who are involved with safety-critical technologies, including self-driving vehicles.

The perfect vs good argument is the pragmatic one for moral hazards like this. IMHO the best scenario as self-driving vehicles become mainstream technology is probably a culture like air travel: when there is some kind of accident, the priority is to learn from it and determine how to avoid the same problem happening again, and everyone takes the procedures and checks that have been established that way very seriously. That is necessarily going to require the active support of governments and regulators as well as the makers of the technology itself, and I hope the litigious culture in places like the US can allow it.

The REAL headline and buried lede for the original post should be:

Trump guts nuclear safety regulations

“The president signed a pair of orders on Friday aimed at streamlining the licensing and construction of nuclear power plants — while panning the Nuclear Regulatory Commission for its ‘myopic’ radiation safety standards.”

We now have industry capture of the Nuclear Regulatory Commission.

Who here knows about Admiral Hyman RIckover? All of this is worth reading:

https://en.wikipedia.org/wiki/Hyman_G._Rickover#Safety_record

Are You Scared Yet?

I would be.

The Department of Energy is selling off more than 40,000 pounds of weapons-grade plutonium from the Cold War arsenal to nuclear reactor startups. All of which I’m sure will be thoroughly vetted and monitored, because this is done under the direction of a former board member. Yikes!

https://en.wikipedia.org/wiki/Chris_Wright

Who IS Oklo, Inc. the "private nuclear reactor builder/operator"? Oklo is Sam Altman:

Trump Administration Providing Weapons Grade Plutonium to Sam Altman

We're in territory where weapons-grade plutonium is being given at fire-sale prices to billionaires who's ethical boundaries include creating their own demand for otherwise unnecessary, high-risk energy projects. Guys like Altman, who get their ideas from Wikipedia articles about Ayn Rand — because they are one rung lower than people who actually READ that garbage.

But I'm sure no inventory of hot nuke metal will ever go missing.

Apple had a culture of authenticity. Culture dies pretty hard in most cases. I think we will see the last of that culture dissipate, as it eroded so greatly under Cook and Ive. Then the extractive, enshittifying corruption will spread from Apple, too.

There really was something, that began with Jobs and Woz. It wasn't perfect, and Jobs had a way of twisting ethical stances in ends-justifying-means sophistry. But Steve Jobs would never have prostrated before Trump, proffering a solid gold token.

Oh, you want profit? This is a surveillance spyware wrapper around the entire MacOS user experience - so if you thought Microsoft's Copilot Recall was invasive monitoring, you haven't seen anything yet.

If Apple won't monetize a user panopticon and partner with governments to do it, OpenAI will be right there, to take the cash.

I will argue against strict adherence to this statement. Gnome applications written to the project guidelines have become very fine, since the introduction of GTK-4 and libadwaita. I prefer many of these to their equivalents on MacOS.

It's true that most of these fall into a general category of "utilities", and that Windows enjoys a broader ecosystem driven by commercial incentive. But Windows programs are hardly "better' for this, and the widely varied usability is generally sub-par compared to level that's become norm for Gnome/Adwaita software.

I agree. The functions provided by this bed do not require internet connectivity AT ALL, let alone some ridiculous cloud based architecture. $5000 is enough to include a $200 miniPC to monitor temp and positional sensors.