macleginn

10 ms·

1.
by macleginn 4d ago
"Moral panic" were the exact thoughts that came to my mind when I read this yesterday: https://www.theguardian.com/technology/2026/sep/11/mps-urge-...
2.
by macleginn 7d ago
The text is mostly about rustc_codegen_utc; maybe this should be reflected in the title? Also the use of italics there is rather jarring.
3.
by macleginn 9d ago
This is how deniable threats work. The promise of "not being nice" in conjunction with "ruining the career" is as clear a threat as there can be in writing.
4.
by macleginn 9d ago
I think this should be in the title of the post. 'OpenAI allegedly threatening to ruin a prominent researcher's career', or smth like that.
5.
by macleginn 11d ago
Recurrent depth and chain-of-thought are two completely different concepts. In the former approach, the output of the layer gets is rerouted as input to the same layer, potentially several times. This output/input is a fixed width time
6.
by macleginn 11d ago
This is why almost 100% of hard and social scientific discourse reads almost exactly like this these days.
7.
by macleginn 14d ago
‘But the idea that you’d need explicit self-referentiality before you could get convincing and world-changing conversational intelligence?’ — has anyone ever formulated this actual idea or anything logically equivalent? This looks like a mi
8.
by macleginn 2mo ago
I am duly impressed by the powerl of the nameless internal AI, but not a single human contributor's name listed anywhere? Did someone at least make this model a coffee?
9.
by macleginn 2mo ago
https://archive.is/VfemY
10.
Markets are getting the AI right(ft.com)
3 points by macleginn 2mo ago|3 comments
11.
Anthropic's Claude AI models hack into 3 outside groups during testing(ft.com)
2 points by macleginn 2mo ago|2 comments
12.
by macleginn 2mo ago
https://archive.is/dAMah
13.
by macleginn 2mo ago
What about writing code by hand?
14.
by macleginn 2mo ago
Array programming where we compare first and look for the first failure later will not help much here if runs are short because by itself it doesn't give you early termination and you may spend a lot of time on wasted comparisons.
15.
by macleginn 2mo ago
An unintended consequence may be the separation between master's and PhD programmes, like what is normally done in Europe.
16.
US postgraduate student visas will be limited to 4 years(dhs.gov)
8 points by macleginn 2mo ago|4 comments
17.
by macleginn 2mo ago
I find it fascinating that two of the most (if not the two most) original, inventive, and outright weird British writers of the 20th c., Ballard and Mervyn Peake, spent their early years in China.
18.
I tricked Claude into leaking your deepest, darkest secrets(ayush.digital)
674 points by macleginn 2mo ago|294 comments
19.
by macleginn 2mo ago
"It will help us solve some of the biggest problems society faces from accelerating drug discovery to developing new clean energy sources to creating novel advanced materials" — but these are not the real problems plaguing modern
20.
by macleginn 2mo ago
For context, I have access to MS Copilot through my workplace. To see what it looks like, I have tried to login through https://copilot.microsoft.com/ , where I was informed that my account, although recognised, is not yet
21.
UK Dynamism Fund(ukdynamism.fund)
1 points by macleginn 2mo ago|0 comments
22.
by macleginn 2mo ago
The overall resemblance to Ticket to Ride's Indian map is rather striking.
23.
by macleginn 3mo ago
Given the sheer volume of the output, it is conceivably difficult to maintain the quality of the plots/characters, even if the animation is stable or even improving. As a byproduct of that, more and more ambitious/batshit source t
24.
by macleginn 3mo ago
They are poor at generalising from a small number of examples; this is why the real generalisation power is achieved in pre-training.
25.
by macleginn 3mo ago
Unremarkable base model will remain an unremarkable fine-tuned model that memorised a couple thousand of input-output pairings.
26.
by macleginn 3mo ago
They normally don't do self-set-up real-time monitoring though.
27.
Bain tests software takeover targets by vibecoding AI replicas(ft.com)
32 points by macleginn 3mo ago|50 comments
28.
by macleginn 3mo ago
Ok, cool, depends on application, probably.
29.
by macleginn 3mo ago
Yes, they do have money to burn, and this will bring some improvements for sure, but active learning has never really worked out, has it? And even 10% of the educated population doing this for, like, 50 years is not that much data, while
30.
by macleginn 3mo ago
Not the worst way to make money, but if internet-scale data were not enough to reduce errors to a somewhat tolerable margin, how much data do they hope to collect in this manner?

More ›