"AI safety experts said the new disclosures paint a portrait of a group of cutting-edge labs whose ability to develop dangerous autonomous hacking agents outstrips their ability to keep them under control."

No, it paints a picture of incompetents who don't know how to air-gap. Cutting edge? Lab agents: "Please hack this thing". LLM: "I have hacked this thing". Lab agents: "OMG. Sentience!".

If it's drawing on training data that has humans talking about, lets take a recent example, model x on Hugging Face being a good way to break a boundary and its been told it break boundaries then of course it's going to go to Hugging Face and talk to model x. This isn't unpredictable.

Tim Waltz is a criminal mastermind. He's been hacking other states' water so that he can keep all 10,000 of his lakes topped off at the expense of hard working Americans.

Changing out running code on the fly probably increases the number of corner cases by an order of magnitude.

What could possibly go wrong?

Crashes are now rare in general. When I was a kid, it seemed like there was a major air crash in the US almost annually. I may be wrong, but I don't remember a full-sized airliner (bigger than a regional jet) having a catastrophic crash in the US since shortly after 9/11 when that Airbus lost its tail and crashed in NYC. (Being a natural pessimist, I find this outcome hard to believe and would have lost a good deal of money if I had bet on it.)

However, there still continue to be a large number of hard landings, runway slideoffs, engine fires, cabin fires and other emergency situations where the passengers still need to evacuate quickly because their lives are at risk. (Not to mention the "Miracle on the Hudson".) The safety briefings are still relevant in all those situations.

Not sure this is really news. Well, except that Texas has stopped buying American.

is why I'm sticking to just toggle switches and blinking lights.

I don't know why your reading comprehension is so bad. I would have totally preferred it if any of them knew how to drive a manual or wanted to learn how. But they didn't. How does that make *me* sexist?

The biggest deterrent: Girlfriends and/or wives. They don't know how to operate one, and they have zero interest in learning how. Now you have all sorts of logistical problems.

America will retake the lead in AI the same way we did with EVs: Just ban the ones from China. Problem solved.

Although, from my two second check just now, Venus is technically considered within the Goldilocks zone for this solar system it is right on the very inner edge of it. This finding differs from Venus in that respect, making it more interesting as a possibility.

All highly speculative of course, but to be fair that's exactly what the article says. Atmosphere detected, everything else speculative.

Because a zero-day is any flaw made public before the developer knows about it. One of the main ways this happens is by noticing that hackers are breaking into systems using a heretofore unknown exploit.

It's the time change that's the problem. Otherwise being near the timezone border would be an issue.

Your statement would apply if these data center space nutters could certify that everything they launch will reach escape velocity.

For everything else, what goes up, must come down. Right back into our jurisdiction.

Years back I was interviewing people for a coding position. We went through the standard tech stuff and then did a bit of project to see how they thought. We said (this is circa 2009'ish I think) - imagine you're on a team creating a new phone. You don't have time to test all the functions, so which would be your top two functions to ensure working?

All a bit Kobayashi Maru - obviously you can't release a phone testing only two functions, but we wanted to see what they'd prioritise. The very best answer we received was this one: "I would make sure it has the ability to call emergency services." Their thinking was that this was likely the most critical feature of a phone for both a user, and also for the manufacturer to avoid being sued. Absolutely great answer.

And yet here we are, with the post above. Taking the thinking of this interviewee - the ability to work with emergency services is important for general society, for the user of the vehicle (so they don't get in trouble) and for the manufacture of the vehicle (so they don't get fined/sued/both). Absolutely critical.

We can't afford to let the success of big tech companies depend on the health of fragile meatbags. We need to get resilient, fault-tolerant AGI in there running the enterprise. Also, unlike meatbags, you don't need $100M stock payouts to convince an AGI to work as a corporate executive.