pera

> Avoiding too much repetition is a core principle of this place

This principle is applied very selectively though: The homepage has been full of insignificant iterations of overly hyped tech products for years now.

It is very hard to imagine all these submissions of announcements of products with monthly release cycles gratifying anyone's intellectual curiosity. Yet it apparently does because they can stay in the homepage for 24 hours.

But for some reason, something as unprecedented as the United States government threatening Harvard with a xenophobic ban is deemed "repetitive"?

I was talking about this issue with a friend a while ago: If an LLM often hallucinates the same package name for a common problem you could copy an existing library, adapt the API to fit the hallucination, use the same hallucinated name and finally include a backdoor.

> Global AI data center power demand could reach 68 GW by 2027 and 327 GW by 2030, compared with total global data center capacity of just 88 GW in 2022.

"AI's Power Requirements Under Exponential Growth", Jan 28, 2025:

https://www.rand.org/pubs/research_reports/RRA3572-1.html

As a point of reference: The current demand in the UK is 31.2 GW (https://grid.iamkate.com/)

It's for sure less questionable than the current proposition of letting a handful of billionaires exploit the effort of millions of workers, without permission and completely disregarding the law, just for the sake of accumulating more power and more billions.

Sure, patent trolls suck, so do MAFIAA, but a world where creators have no means to subsist, where everything you do will be captured by AI corps without your permission, just to be regurgitated into a model for a profit, sucks way way more

- April 7th: Spread rumours of a 90 days tariff pause which immediately triggers a market really, then claim it was "fake new", markets go down

- April 9th: "THIS IS A GREAT TIME TO BUY!!! DJT", then 4 hours later announce a 90 days tariff reduction, markets jump nearly 10%

This level of market manipulation orchestrated directly by a head of state is something you only see in totalitarian states.

The US is moving into a very serious situation. These people won't ever voluntarily leave the Oval Office.

Dump and pump... Trump friends must be making a fortune

I guess the GP meant that the NYT article is not very informative. Bloomberg imo did a way better job reporting this event:

> “This is a fire sale of Treasuries,” said Calvin Yeoh, portfolio manager at hedge fund Blue Edge Advisors Pte. who is selling 20 to 30-year Treasuries futures. “I haven’t seen moves or volatility of this size since the chaos of the pandemic.”

https://finance.yahoo.com/news/treasuries-fire-sale-sends-lo...

Or trading VIX derivatives

Language Modeling Is Compression:

https://deepmind.google/research/publications/39768/

The jump in productivity we had with the world wide web and search engines was several orders of magnitude higher than what you have right now with LLMs, yet I don't remember a single person back in the 2000s calling Google "the greatest tool in human history".

Almost sixty years after ELIZA, chatbots seem to still produce a very strong emotional reaction to some folks.

The original title is too long for HN: Magical thinking and the test of humanity: we have seen the danger of AI and it is us

I agree. Sorted by percentage: 2025-04-04 is #7 (−5.50%) and 2025-04-03 is #12 (−3.98%)

Based on those rules your guess would be wrong :)

This is not about age really: "NFTs"/"web3.0"/"Blockchain technologies" for instance were hyped by every age group.

From the same dilettantes who brought you the Zizians and other bizarre cults... thanks but I rather read Nostradamus

That might explain this then:

> Trump Tariffs Hit Antarctic Islands Inhabited by Zero Humans and Many Penguins

https://www.wired.com/story/trump-tariffs-antarctic-islands-...

> we also sought a language that would allow us to model check (and later prove) key aspects of systems designs while being more approachable to programmers.

I find it a bit surprising that TLA+ with PlusCal can be challenging to learn for your average software engineer, could anyone with experience in P show an example of something that can be difficult to express in TLA+ which is significantly easier in P?

It's not about the style and it's a mistake to perceive generative AI models as analogues of human artists:

- When does generative AI qualify for fair use?:

https://suchir.net/fair_use.html

- AI is currently just glorified compression:

https://news.ycombinator.com/item?id=38399753

Call centres are the first target for full automation. If I remember correctly they represent 3 million jobs in the US only.

There are no "safe jobs" in capitalism.

You can do interesting things with executable binaries as well, see WINE :)

Open model is such a misnomer: it's like calling an ELF an "open executable".

Distributing things for free doesn't make them "open". The reality is that free (as in free beer) weights are closer to freeware than anything open source. In fact, since all these models are build using pirated media a more appropriate term could be plain old warez.

Do you really need to provide the docs? I would have imagined that those docs are included in their training sets. There is even a guide on how to migrate from GTK3 to GTK4, so this seems to be a low-hanging fruit job for an LLM iff they are okay for coding.

Yeah sorry I worded my comment in a weird way: I'm definitively not advocating for any of those points

I guess the majority in here would agree that without strong market intervention OpenAI will soon implode. They urgently need:

- WIPO copyright exemption

- Anti-China protectionist measures

- Hard-line hardware export control

- Multi-billion dollar government contracts

You are missing GP's point and misunderstanding what generative models are actually doing.

The late OpenAI researcher and whistleblower, Suchir Balaji, wrote an excellent article regarding this topic:

https://suchir.net/fair_use.html

The Intercept wrote an article about ShadowDragon and SocialNet a few years ago:

https://theintercept.com/2021/09/21/surveillance-social-medi...

The explosion of garbage content is a big issue and has radically changed the way I use the web over the past year: Google and DuckDuckGo are not my primary tools anymore, instead I am now using specialized search engines more and more, for example, if I am looking for something I believe can be found in someone's personal blog I just use Marginalia or Mojeek, if I am searching for software issues I use GitHub's search, general info straight to Wikipedia, tech reviews HN's Algolia etc.

It might sound a bit cumbersome but it's actually super easy if you assign search keywords in your browser: for instance if I am looking for something on GitHub I just open a new tab on Firefox and type "gh tokio".

classic SLA hacking

Your example sounds reasonable but it's not realistic: The actual use of this type of tools is to intimidate those workers who have outputted 9.8 items instead of the average 9.9 over the past week.

This is how our society ended up making Amazon delivery workers urinating in fucking bottles inside their trucks.

Years ago I wrote one in ~50 lines of C using a wordlist of short, simple english, words:

https://gitlab.com/trobador/sdice