notnullorvoid

Why in the hell would you use agents to test user experience. That's next level incompetence.

I don't think reasonable usage has anything to do with it, it's not hard to find useful workloads to run overnight, and that will just become easier as models get better and need less supervision.

At current GPU pricing you're right, unless you highly value running uncensored models, data ownership, and/or privacy.

As a Canadian I'm disappointed in Cohere's behaviour here. It's rather clear that this was the agent trying to test the rule it was asked to create, and it did not attempt to push a diagnostic file as the subtitle claims, it attempted to attach a small test string that had no sensitive info.

I suspect had that test url been to a S3 bucket url rather than Alibaba, Cohere wouldn't have even bothered to write the article. This is the kind of FUD I would expect from a US based AI company.

Cohere should be attempting to build bridges with Chinese AI companies, not burn them.

> Now we can basically say that its not Facebook and Google doing this directly

Finding out that TVs are listening does not mean Google and Facebook aren't. I've had plenty of times in an old car, or on a walk talking about something only to have related ads pop up later.

I'm embarrassed, I searched for the vote, and looked at the results of wrong bill C-22:

https://www.ourcommons.ca/Members/en/votes/44/1/190?view=par...

I didn't even read the text or look at the date... Reading what that C-22 was actually about I'm glad that one got a near unanimous yes vote.

And seeing the actual results makes me more hopeful that the senate may not allow it through.

> except for the party in power.

The votes for this bill seem to indicate that nearly all MPs regardless of party voted in favour, and none of them voted against.

I've always viewed Fielder as someone who is deeply manipulative for the purpose of entertainment. Yet I do not find it entertaining in the slightest, it's disgusting. So a documentary by a manipulative person, of another manipulative person sounds like the worst thing ever.

Excited to see this applied to Qwen 3.8 27B. It's reasoning output tends to include lots of restatement, which I guess would be more likely to be kept in latent space with this hybrid reasoning.

I suspect they meant an internal monologue, which many do have, but it's not really clear how much it contributes to thinking. People without internal monologue clearly are still able to think though.

I have never seen a dog owner release their dog in a situation where they are scared. Usually if a dog owner is scared they try to keep their dog closer, to protect the dog from the scary thing.

Are you sure the dogs are not already off leash in your area when you are running by. There are many owners who simply have their untrained dogs off leash when they probably shouldn't.

Many of the games include histories within the current telling, histories which include past versions of Gannon, Link, and/or Zelda.

It's pretty clear that the cyclic nature is a part of the story.

Did you even scan the paper? There are benchmarks right there starting on the 9th page.

I don't expect Scion's design to be very good, and the implementation is likely pretty bad, unless it is a near 1 to 1 copy of existing open source.

The amount of training data for incremental/derived data systems is quite sparse in the grand scheme of things. I've seen little to no improvement over the last year when testing new models in that area, they always make some pretty bad mistakes, and what is correct is very inefficient.

How do you suppose an increase in connected IPs affect their ability to geolocate users behind a VPN?

Node only has experimental permissions support (making it not as good for local scripts), and no WebGPU (though you can import dawn wrapper). Deno desktop is way ahead of anything available for Node.

Deno is a excellent and quirky severless, server, CLI, library, and GUI technology.

CF workerd is only one out of those 5.

Kind of hard to figure out what sandboxing controls are available given the lack of docs.

Given it's server focus though, I'd be very surprised if it was as good as Deno for giving agents a local code sandbox.

No if that were true they wouldn't be stopping development of Deno, which as it stands sits in the best position for codemode given it's granular sandboxing.

> We will support the Deno runtime for another year with monthly releases containing bug fixes and security updates. After that year we will end our development of the Deno runtime.

Extremely disappointing...

The risk isn't about reversing the hash, it's about finding a hash collision. If the hash is the only reference to your public key on chain, one does not need to reverse the hash, they just need produce a key pair who's public hash collides.

Currently that's a large brute force problem. But my gut is telling me is that finding a forward pass method of restraining private key generation such that we know what kind of public key hash to expect at the end, should be easier than cracking elliptic curves.

I guess depending on your outlook some might view that as "reversing" a hash, but it's not like you are getting the original input which is a impossible problem (unless there is a bunch of info you already know about the input).

My gut is telling me pushing security model to hashes seems like a far far worse bet if you are worried about advances in maths.

What issue do people have with webp?

What helped my RSI:

- Using mouse less

- Ergonomic keyboard

- Learning to type better (I still suck, but speed is fast enough, and I don't get a soar neck or fingers from my previous glance at KB then back to monitor and smash keys)

- Positioning top of monitor to eye level

- Gradually transitioning from programming through rapid trial and error, to building and exploring logical models of programs in my head, then type out some code that usually works.

All my experiments and usage of AI if anything lead to more typing rather than less.

That must be new. It did not work when I last tested type stripping, earlier this year using pnpm as well. And to be fair bun didn't work either.

Glad if it's working now regardless though, I also see some experimental support for coverage in the docs now.

There is no "one thing", there's many things, the top of the list being:

- packages in workspaces can be used as dependencies without compilation (node's type stripping doesn't support this)

- a good permissions model for limiting access to system fs, net, etc.

- native WebGPU support (including building a windowed app without a web view)

- desktop gui builds that allow using Chromium embedded or WebView

- built-in linter, benchmark, coverage, etc.

It doesn't really matter if they "should or shouldn't" have an expectation of privacy IMO, they already do have that expectation.

> If they were to offer total privacy, is it ok for the public to use chat to get advice on _how_ to commit a crime?

Yes it should be, but it should be illegal for a company providing chat service to respond with anything other than a refusal when doing so.

That detection and refusal should be a private closed loop though, anonymizing any data that will be passed into a training pipeline, or ads targeting. This requirement for closed loop private chats should be mandated by law sooner than later. Otherwise we're getting into very tricky territory where the temptation of alerting on things like pre-crime grows too close.

I hope this highlights that many (most?) non teach people don't consider or know that their use of AI services is not private.

Which quantization are you using to reach those numbers?

Building a LEGO set is akin to building a model car. Adult collectors and builders of model cars, planes, trains, etc. were a thing before millennials.

> Anti-cheat systems have become controversial recently

Not recently, they have a long history of being controversial.

Ideally we'd have consumer protection laws that prevent companies from doing things like using HWIDs to ban accounts.