Sunday, June 01
Daily News Stuff 1 June 2025
May Flowers Edition
May Flowers Edition
Top Story
- CNN says that Anthropic's warning that AI could wipe out 50% of jobs is nonsense. (CNN)
For CNN, it's 100%.
Maybe more.
- AI agents tasked with running a simple vending machine business act like petulant children, declare bankruptcy for no reason, and try to call the FBI over $2 business fees that they have previously agreed to. (arXiv)
Or to put it another way, just like CNN anchors.
Tech News
- Gigabyte has listed memory supports speeds for AMD's Ryzen 9000G CPUs - which don't exist yet. (Tom's Hardware)
These are desktop versions of laptop chips like the Ryzen 370. That has four Zen 5 cores, eight slower Zen 5c cores, and 16 GPU cores - making it faster than the RX 580.
We'll have to wait a few months to see how these perform on the desktop but given that previous laptop-to-desktop generations have shown as much as 35% better performance with the increased power budget the 9900G (if that is what the new 12-core chip is named) could perform close to existing 12-core chips like the Ryzen 7900.
- Molex - the company that makes those ubiquitous hard drive power connectors - showed off a PCIe 7.0 cable running at 128GHz over a distance of a meter. (Tom's Hardware)
Well, technically it's 32GHz because it's using PAM4 DDR signals, but it's still a lot.
- How not to sort by average rating. (Evan Miller)
The simple approaches are wrong and the right approach is definitely not simple.
- Running Doom on an Apple AIX Unix server from 1996. (Old VCR)
Or indeed any other AIX 4.1 system, should you have a bunch of those lying around.
- GitHub added a Copilot feature to use AI to automatically open support tickets and Git repo owners are pissed. (GitHub)
Imagine doing tech support for free and you get the attention of a scam call center in Myanmar because Google handed them your 2FA phone number.
They are not happy.
- The Banana Pi BPI-R4 Pro has things. (Liliputing)
It's based on a four-core Arm Cortex A73 chip, so it should perform similarly to the Raspberry Pi 4.
But unlike the Pi 4, it has seven Ethernet ports (one gigabit, four 2.5Gb, and two 10Gb supporting both RJ45 and SFP ports), two M.2 slots for storage, three M.2 slots for 5G (I think these can also take SSDs), two mini-PCIe slots for wifi, plus 8GB of built-in storage and 4GB or 8GB of RAM. And a micro-SD slot. And it looks like three nano-SIM slots. And I think four USB ports including the debug port.
So if you want all of that, it has it. Would make a very flexible and powerful router if you need a very flexible and powerful router.
- Journalists keep embarrassing themselves. Is that normal? (The Verge)
Apparently, yes.
Musical Interlude
Disclaimer: Dun dun dun!
Posted by: Pixy Misa at
05:48 PM
| Comments (4)
| Add Comment
| Trackbacks (Suck)
Post contains 483 words, total size 5 kb.
1
Journalists have the basic fundamental quality issue of the way we train, recruit, and manage journalists.
Managers have the problem of the way we train, recruit and manage managers, which is still varied enough that we haven't managed to remove almost all of the positive value from managers. Though, even if we were more uniformly bad at hiring and developing managers, it would be hard for us to be as uniformly bad as we can achieve with occupations where we just get lots of money, and our results do not really matter so long as many people are pretending to be politically acceptable.
Anyway, I think lots of scholarship is terrible. Austrian economics is not some big secret, and yet basically academia is in the middle of doubling down on about a half dozen similar gambles where with one they proved in 2020 that they were not i) informed ii) honest proxies for the public iii) going to be inflicting only small bearable costs on the public. Academic consensus cannot have the information to make certain costly discussions on behalf of the public, and in trying they just continue the impression that they are motivated by malice against the 'uneducated' and other non-compliant persons.
AI is a case where there is a cult of unthinking 'experts' pushing it as a panacea that will radically transform the economy, and a cult of unthinking 'experts' pushing it as the at-least-as-unlikely end of humanity, and while some specific applications of neural net algorithms are situationally good, a lot of 'results' are as mediocre as existing human incompetents.
Managers have the problem of the way we train, recruit and manage managers, which is still varied enough that we haven't managed to remove almost all of the positive value from managers. Though, even if we were more uniformly bad at hiring and developing managers, it would be hard for us to be as uniformly bad as we can achieve with occupations where we just get lots of money, and our results do not really matter so long as many people are pretending to be politically acceptable.
Anyway, I think lots of scholarship is terrible. Austrian economics is not some big secret, and yet basically academia is in the middle of doubling down on about a half dozen similar gambles where with one they proved in 2020 that they were not i) informed ii) honest proxies for the public iii) going to be inflicting only small bearable costs on the public. Academic consensus cannot have the information to make certain costly discussions on behalf of the public, and in trying they just continue the impression that they are motivated by malice against the 'uneducated' and other non-compliant persons.
AI is a case where there is a cult of unthinking 'experts' pushing it as a panacea that will radically transform the economy, and a cult of unthinking 'experts' pushing it as the at-least-as-unlikely end of humanity, and while some specific applications of neural net algorithms are situationally good, a lot of 'results' are as mediocre as existing human incompetents.
Posted by: PatBuckman at Sunday, June 01 2025 11:46 PM (rcPLc)
2
"With yet another failed Starship test this week, in which the ambitious heavy rocket exploded once again, you
might reasonably suspect that luck has finally run out for SpaceX. "
Well, you might, if you are an 0bama-style jernolismer who is 27 years old and literally knows nothing, particularly not that that's how SpaceX has developed all their rockets, and especially about how they had 3 failures with Falcon 1 in a row, that left them one test flight, IIRC, from bankrupting the company. Or the number of times they lost Falcon 9s while trying to land them (and now can reuse them upwards of 20 times each.)
Well, you might, if you are an 0bama-style jernolismer who is 27 years old and literally knows nothing, particularly not that that's how SpaceX has developed all their rockets, and especially about how they had 3 failures with Falcon 1 in a row, that left them one test flight, IIRC, from bankrupting the company. Or the number of times they lost Falcon 9s while trying to land them (and now can reuse them upwards of 20 times each.)
Posted by: Rick C at Monday, June 02 2025 01:46 AM (1zWbY)
3
Musk hates having nothing productive to do. He got DOGE started, then left because routine operation was in good hands. He is almost certainly playing a long game on Starship, and it would be boring to skip tests to destruction now and then try to scale up operation later with a less matured design. But communists tend to declare core details finished while key concepts are incredibly suspect, and then be completely mystified when 'routine operation' is fatally catastrophic. Their failures are other people witching them, and right now Musk is clearly an important witch in their eyes.
Posted by: PatBuckman at Monday, June 02 2025 02:02 AM (rcPLc)
4
Uuuu Laaa!
Posted by: The Brickmuppet at Tuesday, June 03 2025 10:29 PM (3NtfN)
55kb generated in CPU 0.016, elapsed 0.1908 seconds.
58 queries taking 0.1802 seconds, 357 records returned.
Powered by Minx 1.1.6c-pink.
58 queries taking 0.1802 seconds, 357 records returned.
Powered by Minx 1.1.6c-pink.
