throwa356262

The future is already here :)

These 3bit quants of qwen are supposed to be 95% accurate as the full model and runs moothly on laptops:

https://huggingface.co/ISTA-DASLab/Qwen3.8-27B-GSQ-RCO-GGUF

There are multiple independent tests of this on YouTube.

These guys are no fun at all.

Is this really the same cowboys that recklessly broke into others networks and instead of apologising proudly told the world how they did it?

Every 6 months Anthropic publishes a report telling us how closer they have gotten us to extinction as if human life was just a game, yet now they draw the line at computer game modding??

Looks like Microsoft forgot to talk to hardware manufacturers about this feature...

    I did not leave X the platform, but I did leave X the interface

Left Facebook and twitter 6-7 years ago. If you just can't leave something without constantly trying to reach back to it in some shape, then you are no different from the junkie that thinks he can quit anytime if he really wanted to.

Rant aside, beside social media minimalism I am also a big fan of compute minimalism. I don't use VScode with 100+ plugins or a zsh maxed up with so much bells and whistles that the terminal looks like Christmas and respond with the speed of microsoft powershell.

But I guess most people in SF are in the opposite camp.

The API access is paid for B2B customers, which includes some of the big labs.

They haven't received much of investment.

This has not been my experience.

I often use the smaller flash models. These may need a nudge once in a while, but not significantly more than what I need to do with the much more expensive Anthropic models.

The end product is often the same. Sometimes a model surprises me with a specially clever solution but this is random and can happen with any model, open or closed.

Someone needs to show increased traffic to not get fired in the next 10% round.

First of all, I am a huge fan. I bought the book, but haven't read it yet.

I am not sure I belive the story about thr manuscript however. It surfaced just as the Internet was becoming a thing. Coincidence?

No, this is not true. Author is trying to make A and O look better than they really are.

Google and Facebook are doing this, but these two were also considered to be the worst in class among the US labs until very recently.

A and O can't do this. And as we recently saw with fable, Anthropic can seriously benchmax their releases.

Lets face it, contractors in many other countries have done the same.

NSA probably detected this leak while going through openai data to datamine other countries :)

    Andy Yen, the CEO of Proton, had a message for an employee in the company’s security Slack channel who urged the company to at least take a few days to pentest newly written code before pushing it live. “That’s bullshit and loser thinking...”

Ah, that explains the recent enshitification. Add some AI slop to this mess and you will soon be the centerpiece of haveibeenpwned

I dont know if this is an extreme opinion.

Elon helped Trump get elected, who has been hostile towards Canada and its economy.

Tesla dropped their prices significantly to turn the trend from previous year.

Is this AI security slop or a real issue? Can the phone actually receive these kind of messages from the network?

Also, I don't understand Googles position here. On android the phone ia not part of the OS and can be updated via an app or framework update. Why not just fix this as a functional bug?

FYI, first thing those programs check is your nationality.

Guess how many countries are on the not-naughty list?

On the contrary, problems with Teslas are underrepprted in media.

This is despite Tesla beeing involved in a significant number of high profile accidents, many with fatalities [1] [2].

According to NHTSA complaint records it is also one of cars that breaks down most often. Model Y is on the bottom of quality lists in multiple European countries [3].

[1] https://myelectricsparks.com/tesla-crash-germany-father-chil...

[2] https://apnews.com/article/car-crash-tesla-france-fire-be8ec...

[3] https://www.tuev-nord.de/de/wissen/ratgeber-und-tipps-mobili...

Their previous updates have sometimes been really bad for Chinese models. Either by exclusion, delays or simply weighting things differently so OpenAI looks better.

This feels like the opposite of benchmaxing. Maybe "minbenching" can become a word now?

That's not benchmaxing.

You need a method to show you RL is improving the model that is separate from the training data.

So it begins. First Chinese providers were evil and should not be trusted, next Chinese models are dangerous and must be avoided.

I had a closed LLM push a sensitive local project to a github repo for the world to see in direct violation of my instructions. LLMs are dumb and unpredictable, run them enough times and at least one instance will do something stupid and dangerous.

IMO we are already at the enshitfication stage.

Anthropic increased short term profits by jacking up their prices for enterprise accounts - by starting charging API prices instead of plan prices. It is possible we soon see customers fleeing Anthropic once they find a good alternative

I think a company like Moonshot or Z.AI could have sabotaged this behind repair by doing a 10-20B IPO before Anthropic.

That could have popped the AI bubble right there.

Yeah sure, why not. Lets choose the guy that ran Hungary economy into the ground with a GDP below Romania?

Even politics asides, this guy is not particularly bright. He was a net receiver of EU funds while actively sabotaging EU work. Guess what EU did in response?

This sounds like user-blaming.

A bit of side track, but the 2014 massive hack was due to bad passwords hygine AND a serios vulnerability in Apple icloud (backup restore would bypass f2a and login failure rate limiting).

Back to the subject. See the reports from citizen labs, Google TAG and others. You will see they are often due to undisclosed vulnerabilities not user stupidity

How do we know it is, besides apple marketing?

Given how many high profile politicians, journalists and businessmen have gotten their phones hacked, the iPhone is not unhackable.

Is this not the expected function when you use bixby?

(Not saying people should use bixby mind you)

Looking at their pricing page for China, K3.1 seems to be about 30% cheaper than K3.

--

Model K3.1 Pro K3.1 Pro Highspeed. K3

Input ¥13.00 ¥26.00 ¥20.00

Output ¥68.00 ¥136.00 ¥100.00

Cache hit ¥1.30 ¥1.30 ¥2.00

Cache write ¥13.00 ¥26.00 ¥20.00

Ah, the version available to us mortals is nerfed in cybersecurity just like the US models :(

Console as in the terminal :)

And yes, I clicked the link thinking this was about PS5 homebrew.

Big deal, they are bringing 0731 prices to the same level as 1017. Given their compute constraints (thanks to US sanctions) it makes sense to not support too many legacy models anyway.

https://www.alibabacloud.com/help/en/model-studio/model-pric...

I dont know, with the exception of SK, all these companies have announced reduced AI budgets and/or move to own inference.