mikaraento

Do you know of a good ready-made implementation of such a proxy? I’ve been looking for one.

GitHub is also a worry in terms of exfiltration. You can’t block pushes to public repos unless you are using GitHub Enterprise Managed Users afaict.

Favelas

That might be somewhat ungenerous unless you have more detail to provide.

I know that at least some LLM products explicitly check output for similarity to training data to prevent direct reproduction.

You do know that something similar is true for JPEG, right? :)

JPEG is a compression method. Files with JPEG-compressed data are most likely to be in either JFIF or EXIF container formats. Both will almost always use the .jpg/.jpeg file extension.

Nit: RDP’s roots are more in multi-user windows like Citrix Metaframe than in remote administration. I’ve found it to perform better than the alternatives (remote X11, VNC, Chrome Remote Desktop) for remote GUIs. Nomachine is the only alternative that was close to its performance.

(And before somebody jumps in to correct me - in ancient times X11 performed quite well over the network but modern Linux GUI apps are no longer designed to minimise X11 network traffic)

I’m sorry you went through that

Not quite. You take Finnish pancake batter (unleavened, a bit thicker than French crepes) and add blanched, finely chopped spinach or nettles.

https://scandicuisine.com/stinging-nettle-pancakes/ looks quite reasonable though most Finns would not use a blender for this.

Relatively common in Finland to use young nettles like you’d use spinach in hot dishes (soup, blanched, pancakes).

I can be that someone this time. The ”repair and protect” version has helped my low-level toothache.

Heads-up to people trying this: Gmail will often put forwarded email into spam. Be careful especially in the beginning to check your spam folder. They may also reject the mail as spam, esp if your volume is large.

IIUC it’s hard to make forwarding to play nicely with DKIM and spf. There’s some disagreement on how to handle it. (I’m being purposefully vague as I did interact with folks handling this on the google side and don’t want to cause them trouble for helping me out).

Thanks for the links.

AFAICT these have not resulted in any shipping products.

Sorry, I meant to emphasize _different_ mechanical connections. That a sensor that detects pressure has a different mechanical linkage than the one detecting vibration. So you need multiple different manufacturing techniques to replicate that at correspondingly higher cost.

The “more than 10000” also has a large impact in size (sensors need to be very small) and cost (you are not paying for one sensor but 10000).

Of course some applications can do with much less. IIUC the article is all about a _universal_ humanoid robot, able to do _all_ tasks.

The article points out that the human hand has over 10000 sensors with specific spatial layout and various specialised purposes (pressure / vibration / stretching / temperature) that require different mechanical connections between the sensor and the skin.

Depends heavily on the use case. Indeed many tasks humans carry out are done without touch feedback - but many also require it.

An example of feed-forward manipulation is lifting a medium-sized object. Classic example is lifting a coffee cup. If you misjudge a full cup for empty you may spill the contents before your brain manages to replan the action based on sensory input. It takes around 300ms for that feedback loop to happen. We do many thing faster than that would allow.

The linked article has a great example of a task where a human needs feedback control: picking up and lighting a match.

Sibling comments also make a good point on that touch may well be necessary to learn the task. Babies do a lot of trial-and-error manipulation and even adults will do new tasks slower first.

Do you know of success stories here? Success of transferring models learned in physics simulation to the real world.

When we (ZenRobotics) tried this 15 years ago a big problem was the creation of sufficiently high-fidelity simulated worlds. Gathering statistics and modelling the geometry, brittleness, flexibility, surface texture, friction, variable density etc of a sufficiently large variety of objects was harder than gathering data from the real world.

Around 2008 a core step in search was basically a grep over all documents. The grep was distributed over roughly 1000 machines so that the documents could be held in memory rather than on disk.

Inverted indices were not used as they worked poorly for “an ordered list of words” (as opposed to a bag of words).

And this doesn’t even start to address the ranking part.

IIRC it was never used for any voice/call switching. I think it was used in one (line of) ATM switch(es).

Definitely lowered alcohol consumption for me.

I no longer get that dopamine rush of the first drink.

Some people go as far as describing it as anhedonia. I kinda get it.

Could you provide a pointer to docs for this? It wasn't obvious from an initial read of their docs.

I remember that years ago (mobile) Safari would aggressively use GPU layers and crash if you ran out of GPU memory. Maybe that's still happening?

Preview on a mac handles the file fine.

I remember being able to file support cases just by buying one for a couple of hundred dollars. They'd also promise that if it turned out to be a bug in the product the fee would be refunded.

(My case wasn't solved. It was something about variable delays in getting packets off the network and into userspace but we never got to the bottom of it).

So maybe not luxury, but premium. Like MB or Audi in cars. Pricing still plays out differently.

I guess text in images would be similar, and is indeed where image generation models struggle to get the details right.

E.g., making a greeting card with somebody's name spelled correctly.

SUBST was commonly used for source code directories. It served at least two purposes: making source appear at the same location on all machines and working around path length limits.

I've also used SUBST when reorganizing network drive mappings.

They didn't study patients using semaglutide for weight loss only.

I guess a Bayesian would increase their estimated probability of such a problem also for the weight loss case, but not by as much (as diabetes predisposes people to other ailments).

For compute offload, Google has indeed done that - the Tensor chips have Google's TPUs instead of Qualcomm DSPs.

Both on these TPUs as well as on pre-Tensor hardware that had Qualcomm DSPs, Pixels would not allow apps access to the kernel interfaces. Access would be blocked or mediated via a separate service process ('binderized HAL').

(Some) OEMs have repeatedly opened access to these kernel interfaces in order to trade security for performance.

(I used to work on compute offload at Google).

To be charitable, I think tptacek was referring to a variant of "accounts aren't closed because of credit risk" as not being true. The credit risk talked about in the article was ACH reversals with a bankrupt account owner falling on the bank, rather than overdrafts.

RFC 9413 referenced in a parent mentions HTML. It points out that formats meant to be human-authored may benefit more from being liberally accepted.

I also read that XHTML made template authoring hard, as the template itself might not be valid XHTML and/or different template inputs might make output invalid. (I sadly can't find the source of this point right now, but I can't claim credit for it).

Things that have helped me:

- nose mask rather than full face mask

- nasal spray if my nose feels congested

- upping the pressure. In theory a professional is supposed to set your pressure but I do tweak mine. If the setting is not visible it's probably in a hidden menu - Google for how to enable it for your machine.

- sleeping pills

I hope you find a solution. Sleep apnea is terrible.

To be fair, in the case of hearing aids you are both in the right.

Excessive regulation has created oligopolies and kept prices high in the US. The OTC hearing aid category is meant to help. Before that, low-cost devices tended to remain niche.

OTOH the regulation(s) were introduced due to blatant sales of substandard devices, esp in the 1970s. A high-amplification device runs the risk of further damaging your hearing. Many hearing aid users are vulnerable elderly.