"If you wanted to create a tool that would enable the destruction of institutions that prop up democratic life, you could not do better than artificial intelligence. Authoritarian leaders and technology oligarchs are deploing [sic] AI systems to hollow out public institutions with an astonishing alacrity"

So in the first two sentences we have hyperbole and typos? Hardly seems like high-quality academic output. It reads more like a blog post.

Reading and interpreting datasheets: A- (this has gotten a LOT better in the last year)

Give netlist to LLM and ask it to check for errors: C (hit or miss, but useful because catching ANY errors helps)

Give Image to LLM and ask it to check for errors: C (hit or miss)

Design of circuit from description: D- (hallucinates parts, suggests parts for wrong purpose. suggests obsolete parts. Cannot make diagrams. Not an F because its textual descriptions have gotten better. When describing what nodes connect to each other now its not always wrong. You will have to re-check EVERYTHING though, so its usefulness is doubtful)

> and PRC's was to peak emissions by 2030s

This appears to be wrong. Peak is supposed to be before 2030. They will not hit it.

https://climateactiontracker.org/countries/china/targets/

Because fossil fuels have higher in/out losses this is number is larger than usage. This metric is generally used to track decarbonization.

Using the IEA number you can see the hydro+solar+wind production is about 9.5% of the total, not 18%.

ChatGPT or you favorite LLM can explain in greater detail, just send it the plot image and ask.

https://www.iea.org/countries/china

https://ourworldindata.org/grapher/coal-consumption-by-count...

If, for ethical reasons, fewer people were willing to take these jobs, then either salaries would have to rise or the work would be done less effectively.

If salaries rise, the business becomes more expensive and harder to scale. If effectiveness drops, the systems are less capable of extracting/using people’s data.

Either way, refusing these jobs imposes real friction on the surveillance model.

If you want a deontological answer:

You have a responsibility not to participate in unethical behavior, even if someone else would.

"Dollar General stores have failed more than 4,300 government price-accuracy inspections in 23 states since January 2022" Is this a lot or a little? There no context for national chain

> Dad journalism

"Bad" Typo, sorry

> You should not update on this article

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

> (…) a national report found that (…)

The article's thesis is that these stores are uniquely bad. Whether or not this is correct necessarily depends on a comparison with other stores.

You should not update on this article unless you have some outside knowledge of the industry.

I had AI look into it, it found a national report found that dollar stores had pricing errors at about twice (3.5%) the rate of traditional supermarkets (1.7%) but lower than convenience stores (4.9%).

https://cdn.ncwm.com/userfiles/files/Resources/Price%20Verif...

- you can record all manner of video in your store...

- but you can't process it in this particular way.

My view: I would think given how many code supply chain attacks we've see recently this would be be regarded as (at worst) a necessary evil. How much software used by large numbers of people does the open source community think will be done by anons?

Sidenote: The author implies SyncThing development was stopped due to author ID but the post linked does not say this and gives a completely different reason (forced updates)

But anybody relying on Gary's posts in order to be be informed on this subject is being being mislead. This isn't an isolated incident either.

People need to be made be aware when you read him it is mere punditry, not substantive engagement with the literature.

It is scientific malpractice to write a post supposedly rebutting responses to a paper and not directly address the most salient one.

https://arxiv.org/abs/2506.09250

Furthermore we have clearly seen increases in reasoning from previous frontier models to current frontier models.

If the authors could /did show that both previous-generation and current-generation frontier models hit a wall at similar complexity that would be something, AFAIK they do not.

My read of this is that the paper demonstrates that given a particular model (and the problems examined with it) that giving more thought tokens does not help on problems above a certain complexity. It does not say anything about the capabilities of future, larger, models to handle more complex tasks. (NB: humans trend similarly)

My concern is that people are extrapolating from this to conclusions about LLM's generally, and this is not warranted

The only part about this i find even surprising is he abstract's conclusion (1): that 'thinking' can lead to worse outcomes for certain simple problem. (again though, maybe you can say humans are the same here. You can overthink things)

It's effectively 6 years too. You only get to depreciate 10% in 1st year. This might have killed my company if it was around during first years.

See my comments on the previous discussion (Nov 2023) here: https://news.ycombinator.com/item?id=38145630