Agreed, but these ninnies can't create a secure sandbox that is only allowed to spit out an image. The sandbox itself will have sandbox escape vulnerabilities.

They could make sure their file preview handlers are secure. Just peer review (I guess if your peers are fellow M$FT employees that may be a problem) and validate the parser line by line use a god damn checklist and various tools that exist for doing this sort of thing. Ensure all inputs to every function are validated for length and types. It's not even that hard, and certainly not impossible, especially for images. I guess something complex like PDF may be harder .. but they don't need to support everything. I mean, if they can't write a trusted parser, maybe they should get out of the OS business? Jesus.

First off, I believe I implied they should work with the AI .. while it can't fully design a bridge yet, it is not far from being able to do so. It will, once it can interface with tools like autocad and utilize templates etc.). It doesn't just do "pattern matching" it can reason also. You can ask it things like "what are the structural components of a suspension bridge?" or "what type of bridge is best across a 500 ft canyon that's 1000 feet deep?" It will suggest a type and tell you why that type is superior to other choices (economics etc.). That shows it is able to think a problem through (it can make a list of possibilities and required knowns/unkowns and then flow through a decision tree it created based on scenarios bridges are selected and designed for). It then asks you for geological context (it kind of already does .. try it in ChatGPT .. you'll see) and in the future give you a few workable designs and CAD files (I said in the future, as in within a decade depending on the market need and resources they put into it). Right now it's willing to provide some sketches stuff "If you can share intended use (highway vs pedestrian), expected loads, and whether the canyon walls are sound rock, I can sketch a preliminary concept (span/depth ratios, member sizes, and an erection sequence) tailored to your site."

Are you in favor of getting rid of cars and resorting to walking or running everywhere? I mean, the Olympics is for entertainment and showcasing natural human ability -- drugged up athletes don't meet that criteria. As I said before, if I need a bridge design, I am not going to deliberately handicap it. I am going to want the best possible design. Are humans are so dumb I have to repeat myself? Not a good sign. Maybe you should try to argue (with non-contrived examples or evidence) that the purely human designs will always be better than one that uses AI/computers, instead of giving me BS?

How so? If anything AI would increase human abilities.As for knowledge who cares if it's in your brain or encoded in silicon/electrons? If you can retrieve and use it at the opportune time that's good enough. If I ask you to give me the design for a bridge, if you present me a design based on your knowledge or one that AI produced .. it is the quality of the *design* that matters. I shouldn't actually care where you got it from (unless you stole it, in which case there are legal consequences), I should only care about the quality of the design itself.

Kids ARE using AI and they will continue to do so. Do what happened to math classes when calculators came out. Increase the breadth and amount of problems they are given to solve.

Uh, I am sure they can work it out. What do you think they'll hand the project to an intern? Besides there are non-flying solutions like an RC cone. If they can work out how to drive a truck autonomously on the highway I am sure they can figure out how to safely place a few cones on the road in the correct locations.

What makes you suspect that we'll have increased carnage getting to a self-driving world? What evidence? We have very strong evidence that contradicts that .. you can see the detailed stats for Waymo on this page: https://waymo.com/safety/impac... Tesla also has some of its own stats, you can google it.

I only know of one fatal self-driving car related accident, and that was back in 2016 when Uber thought it could save money by reducing the number of cameras its vehicle had, and the safety driver was watching Netflix instead of monitoring the driving. (Note: there have been a few Tesla autopilot (not FSD) related crashes). Autopilot is like glorified cruise control it was never intended as self-driving, you're supposed to pay full attention. I guess Tesla misnamed it.

We have had Waymo driving autonomously with no human intervention in multiple US cities without fatal self-driving caused accidents. So far, in 30 million miles of total driving (all Waymos have the same brain/sensors), there was only one serious accident a year or two ago that involved a Waymo (a woman was hit by a human driven car and thrown on the Waymo). Unlike humans the Waymo learned from that (even though it wasn't its fault the way it handled it could have been better).

If the rule is unreasonable and contributes nothing to safety then it's blocking. For example, if the existing rule said you can't be bald .. you'd have to agree that it's unreasonably blocking bald drivers.

How good are humans at those "organic" responses? 4000 fatal large truck accidents annually by human truck drivers in the US are not enough for you? You know, like this just yesterday: https://www.youtube.com/watch?...

That's a good idea, why couldn't they think of it? I was immediately thinking deploying wheeled remote control cones but drones are a cheaper and better idea. There are many solutions to this, you have to wonder about why Aurora didn't offer them up

False (obviously by 1994 the vast majority favored web over gopher .. but there were a few people who preferred gopher's simplicity). In fact there are still people today who still say Gopher is better: https://qr.ae/pC2X31 AND https://arstechnica.com/tech-p...

I had no idea "Slashdot Reader" was a bona-fide pundit credential up there with Nobel laureate. I should definitely put it in my resume.

He's waiting to decide after the Soviets prove that it's even possible.

"I don't support people getting unemployed because their work is stolen, mashed up and resold."

How can someone get unemployed because I look up something in AI? That's just dumb. Let's say I ask AI "how does an airfoil work?" I am getting to skip aerodynamics 101 and now some professor (not even the inventor of airfoils) is out of work? Maybe we shouldn't have people collecting a tax for sitting between knowledge?