in plain english, like the AI techbros you despise, you’ve given up your ability to think.
The unthinking AI haters are all over social media. They keep saying that AI can’t really think. But ironically, the “arguments” they usually use is the worst kind unthinking regurgitated groupthink slob.
but nothing they produce will live up to the hype
We just want the next book to be about the same as the previous books. There is not “hype” here?
NASA is too focused on creating a moon base which is dumb as shit.
Why dumb?
Even if you want a Mars base eventually, it seems like a good idea to get some practice building a similar moon base first. Many of the problems will be the same, but it will be much easier, cheaper, and safer to learn them in a place which is only days away from resupply.
No matter what they produce
Now, I agree that we have no right to demand anything from them.
But… their output rate has been genuinely abysmal. If we had a right to demand anything at all from them (we don’t), then it would be quite reasonable to ask them to at least finish one book every 10 years…
People here in the fediverse seems perfectly happy to downvote factually correct arguments. Or to upvote blatantly invalid or false arguments. Just because they prefer the fake conclusion. It seems much worse than on Reddit.
So I think part of the solution could be to somehow enforce against people arguing in bad faith. Timeouts or blocks, or whatever.
LLMs don’t have any intentions.
Eh. The output from LLMs is usually pretty goal-oriented, so it arguably has intentions.
The LLM is not designed to deceive though, so in that sense it is correct that it is not lies.
The question is not whether current battery technology poses that risk: it does. The questions are whether that’s an outsized risk relative to other car technologies and whether Tesla could have done more.
Googling, there also seem to be a consensus that the fire risk in petrol car is way higher than the risk for battery cars. E.g. from https://www.theguardian.com/business/2023/nov/20/do-electric-cars-pose-a-greater-fire-risk-than-petrol-or-diesel-vehicles
“All the data shows that EVs are just much, much less likely to set on fire than their petrol equivalent,” said Colin Walker, the head of transport at the Energy and Climate Intelligence Unit thinktank. “The many, many fires that you have for petrol or diesel cars just aren’t reported.”
After some googling, the “Why?!” in the image seems to be completely bullshit, though.
Electric cars have a way lower amount of deaths from fire than petrol cars. So why would there be outrage about Tesla fire safety?
The comparison in the image also completely ignores number of cars on the street, total miles driven. All car models will have fire deaths. Without scaling with number of cars, you are just always going to be outraged at the number of deaths for the most popular car model…
The outrage about the Ford Pinto was because it was a known specific defect which was covered up (fuel tank rupturing when rear ended), more than the number of deaths in itself. As far as I can tell, the number of fire deaths were actually not extraordinarily high for the Pinto. Hence comparing this not especially high number to Tesla for outrage purposes is meaningless.
Gold is fluctuating wildly in price. It is meaningless to talk about measuring commodity prices in gold.
Please tell us where your window is - how can we avoid climbing through your window, if we don’t know which window we are not supposed to climb through.
Sony loves making their own custom formats.
Stuff is just getting more expensive, because of demand competition with AI. There is no reason to think that production for non-AI computing will ever hit literally zero.
RAM used to be ~$4/GB. So 300*4=$1200. A price increase of $1200 is actually pretty darn affordable to get self driving, surely?
Sure, there are other components than RAM needed. But the RAM alone is not what would make it unaffordable.
The world is not that simple. There are too many combinations to try. And you risk hitting local maxima, even if doing the gradient thing.
I were simply unable to convince Codex to split a patch into separate git commits in a meaningful way. There are things that just doesn’t work.
Still useful for lots of stuff. Just don’t use it blind.
If used by an expect developer, then the combinations are not just random “lucky” choices.
Well, yes, that is a central point.
I am a senior programmer. LLMs are amazing - I know exactly what I want, and I can ask for it and review it. My productivity has gone up at least 3-fold, with no decrease in quality, by using LLMs responsibly.
But it seems to me that some people on social media just can’t imagine using LLMs in this way. They just imagine that all LLM usage is vibe coding, using the output without understanding or review. Obviously you are very unlikely to create any fundamentally new solutions if you only use LLMs that way.
only to find out you didn’t provide adequate requirements for your config.
Senior programmer. I know exactly what I want. My requirement communicated to the LLM are precise and adequate.
The database was an arbitrary example. A more relevant example would be tenserflow layers in a neural network. As I understand it, you can in some cases get a novel solution to a problem just by choosing a smart enough combination, with the right data.
ChatGPT absolutely knows how to help doing the grunt work setting up the tenserflow configuration, following your directions.
Even novel solutions are usually built out of smaller common building blocks. E.g. many novel solutions surely use a database. You can make the LLM help you set up and use the database, that your novel solution uses.
No. Not by any definition of “simulation” I know.