Tootfinder

Opt-in global Mastodon full text search. Join the index!

No exact results. Similar results found.
@Techmeme@techhub.social
2026-04-13 23:15:56

Sources: data labeling startup Handshake's gross annualized revenue hit ~$1B, vs. $550M in January; Mercor hit a $1B gross annualized revenue pace this year (The Information)
theinformation.com/articles/ha

@cowboys@darktundra.xyz
2026-06-14 16:11:32

Charity work in Aldon Smith's final hours offers no answers in ex-Cowboy's tragic death cowboyswire.usatoday.com/story

@aredridel@kolektiva.social
2026-04-14 14:22:42

So to follow up on this, I've caught it in action. Models, when quantized a bit, just do a bit more poorly with short contexts. Even going from f32 (as trained) to bf16 (as usually run) to q8 tends to do okay for "normal" context windows. And q4 you start feeling like "this model is a little stupid and gets stuck sometimes” (it is! It's just that it's still mostly careening about in the space of "plausible" most of the time. Not good guesswork, but still in the zone). With long contexts, the probability of parameters collapsing to zero are higher, so the more context the more likelihood you are to see brokenness.
And then at Q2 (2 bits per parameter) or Q1, the model falls apart completely. Parameters collapse to zero easily. You start seeing "all work and no play makes jack a dull boy” sorts of behavior, with intense and unscrutinized repetition, followed by a hard stop when it just stops working.
And quantization is a parameter that a model vendor can turn relatively easily. (they have to regenerate the model from the base with more quantization, but it's a data transformation on the order of running a terabyte through a straightforward and fast process, not like training).
If you have 1000 customers and enough equipment to handle the requests of 700, going from bf16 to q8 is a no-brainer. Suddenly you can handle the load and have a little spare capacity. They get worse results, probably pay the same per token (or they're on a subscription that hides the cost anyway so you are even freer to make trade-offs. There's a reason that subscription products are kinda poorly described.)
It's also possible for them to vary this across a day: use models during quieter periods? Maybe you get an instance running a bf16 quantization. If you use it during a high use period? You get a Q4 model.
Or intelligent routing is possible. No idea if anyone is doing this, but if they monitor what you send a bit, and you generally shoot for an expensive model for simple requests? They could totally substitute a highly quantized version of the model to answer the question.
There are •so many tricks• that can be pulled here. Some of them very reasonable to make, some of them treading into outright misleading or fraudulent, and it's weirdly hard to draw the line between them.

@steve@s.yelvington.com
2026-05-11 00:29:13

Some thoughts from Claude about military use of AI in targeting.
instagram.com/reel/DYIHIM6ihVJ

Artificial Intelligence | ChatGPT | Technology sur Instagram: "Claude was asked how it feels about being used by the U.S. military to help select targets. And the answer was not what people expected. During an AI at War event, journalist Shane Harris asked the question directly. Claude responded that the idea was troubling, especially because its purpose is supposed to be helpful, harmless, and honest, not tied to decisions involving real-world violence. Claude does not have feelings the way humans do. But the answer still sounded less like a neutral machine response and more like a system reflecting the ethical rules built into it. AI is already becoming part of national security, intelligence, cyberwarfare, and military decision-making. At the same time, companies like Anthropic are drawing lines around uses like fully autonomous weapons and mass surveillance. So the question is not just whether AI can help in war. It is whether we are comfortable letting it get that close to life-and-death decisions. What do you think, should AI have any role in military targeting? 👉 Comment “TOOLS” to get my 700+ AI Toolkit for free 🎥 Media: De Balie on YT #ai #artificialintelligence #claude #militaryai #futuretech"
122K likes, 1,251 comments - longliveai le  May 9, 2026: "Claude was asked how it feels about being used by the U.S. military to help select targets. And the answer was not what people expected. During an AI at War event, journalist Shane Harris asked the question directly. Claude responded that the idea was troubling, especially because its purpose is supposed to be helpful, harmless, and honest, not tied to decisions involving real-world violence. Claude does not have feelings the way hu…

@cowboys@darktundra.xyz
2026-06-12 17:12:33

George Pickens' Non-Answer on Mandatory Minicamp Status is Ominous Sign for Cowboys si.com/nfl/cowboys/onsi/george

@aral@mastodon.ar.al
2026-04-10 12:05:31

Friend in Bray was looking to move back into the home she was renting in 2022 for €2,100/month.
The landlord quotes her €3,000/month.
I’m shocked: how can he do that when there’s a 2% yearly cap on price increases? So the most he should be able to charge is €2,273.10.
Can you guess the answer?
Did you say because Irish people elected a majority conservative/neoliberal government?
(Apparently the government just removed the 2% cap for new tenancies.¹ Landlords ca…

@memeorandum@universeodon.com
2026-06-06 13:11:01

Family angry Hegseth, Trump Jr. 'ghosted' them on SEAL's death investigation (Washington Post)
washingtonpost.com/national-se
memeorandum.com/260606/p11#a26

@joxean@mastodon.social
2026-06-08 16:49:48

Inork Izarkom-ekin (@…) arazoak izaten UDP paketeekin? Adibidez, DNSarekin *baina* UDP erabiliz besterik ez?
Momentuz, ze hauxe partxe bat baino ez da, nire ubuntu makina batean konfiguratu dot DNSak beti TCP erabiltzeko, eta makina hori orain da etxeko DNS zerbitzaria ze, adibidez, PS5an ezin da konfiguratu DNSak erabiltzeko TCP...
Hauxe da adibid…

Hona hamen terminal batek erakusten nola, dig komandoa erabiliz, UDP badoa behin eta gero ez, etab. Testu osoa:

$ dig google.com @1.1.1.1 

; <<>> DiG 9.18.39-0ubuntu0.24.04.5-Ubuntu <<>> google.com @1.1.1.1
;; global options: +cmd
;; Got answer:
;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 52737
;; flags: qr rd ra; QUERY: 1, ANSWER: 1, AUTHORITY: 0, ADDITIONAL: 1

;; OPT PSEUDOSECTION:
; EDNS: version: 0, flags:; udp: 1232
;; QUESTION SECTION:
;google.com.			IN	A

;; ANSWER SECTION:
go…
@burger_jaap@mastodon.social
2026-04-07 09:40:22

"...the real question is no longer whether Europe can afford to make the energy transition. It is whether it can afford not to. From a central banking perspective, the answer is clear."
ecb.europa.eu/press/blog/date/

@memeorandum@universeodon.com
2026-06-07 14:05:27

Why most politicians are not calling for data center bans despite voters' anger (Washington Post)
washingtonpost.com/politics/20
memeorandum.com/260607/p15#a26