New HW4 Teslas do in fact include a front-facing radar, but it’s currently only used for collecting data - not for FSD.

Still, gotta give them credit for getting by with vision-only quite well. I don’t personally see any practical reason why you absolutely must include LiDAR. We already know driving relatively safely with vision only is possible - all the best drivers in the world do it.

And they always will. You need to look at the big picture here, not individual cases. If we replaced every single car on US roads with one driven by AI - proven to be 10 times better a driver than a human - that would still mean 4,000 people getting killed by them each year. That, however, doesn’t mean we should go back to human drivers and 40,000 people killed annually.

Broadly speaking, an AI driver getting stumped means it’s stuck in the middle of the road - while a human driver getting stumped means plowing into a semi truck.

I’d rather be inconvenienced than killed. And from what I’ve seen, even our current AI drivers are already statistically safer than the average human driver - and they’re only going to keep getting better.

They’ll never be flawless though. Nothing is.

If it’s connected to internet it can be hacked.

I think the interventions here are more like: “that’s a trash can someone pushed onto the road - let me help you around it” rather than: “let me drive you all the way to your destination.”

It’s usually not the genuinely hard stuff that stumps AI drivers - it’s the really stupid, obvious things it simply never encountered in its training data before.

I’ve seen a few videos of this thing in action, and while I like the concept - especially that you can use the same device to mow the lawn too with the lawnmower attachment - it’s still quite painful to watch it work.

Especially with snow blowing, it’s just so disorganized: driving all over the place and making quite the mess. If I’m dropping 5k on an automatic snow blower, I don’t want to have to clean up after it.

They’re leaving like redditors to Lemmy. There’s dozens of them!

Nobody knows what it actually takes to reach AGI, so nobody knows whether a certain system has enough compute and context size to get there.

For all we know, it could turn out way simpler than anyone thought - or the exact opposite.

My point still stands: you (or Cameron) couldn’t possibly know with absolute certainty.

I’d have zero issue with the claim if you’d included even a shred of humility and acknowledged you might be wrong. You made an absolute statement instead. That I disagree with.

It’s perfectly valid to discuss the dangers of AGI whether LLMs are the path there or not. I’ve been concerned about AGI and ASI for far longer than I’ve even known about LLMs, and people were worried about exactly the same stuff back then as they are now.

This is precisely the kind of threat you should try to find a solution for before we actually reach AGI - because once we do, it’s way, way too late.

Also:

There is factually 0 chance we’ll reach AGI with the current brand of technology.

You couldn’t possibly know that with absolute certainty.

If anyone who wasn’t the head of an LLM company spouted this drivel, they’d be locked away in a padded room and Axios would rightly be called out for exacerbating the mental health crisis of a paranoid schizophrenic.

Like Eliezer Yudkowsky, Roman Yampolskiy, Stuart Russell, Nick Bostrom, Yoshua Bengio, Geoffrey Hinton, Max Tegmark and Toby Ord?

This comes across more as a warning than a sales pitch.

If only I could convince myself to be as dismissive about the threats of AGI as the average user here seems to be…