Hacker News new | past | comments | ask | show | jobs | submit | semicolon_storm's comments login

The censorship is not present in the distilled models which you can run locally

Are you referring to the distilled models?

yes, they are not r1

Can you explain what you mean by this?

For example, the model named "deepseek-r1:8b" by ollama is not a deepseek r1 model. It is actually a fine tune of Meta's Llama 8b, fine tuned on data generated by deepseek r1.

Perhaps because that's a strawman argument. "Scaling" doesn't mean double the investment and get double the performance. Even OpenAI's own scaling laws paper doesn't argue that, in the graphs compute increases exponentially. What LLM scaling means is that there hasn't been a wall found where the loss stops decreasing. Increase model size/data/compute and loss will decrease -- so far.


In security, industry standard seems to be about the same as military grade: the cheapest possible option that still checks all the boxes for SOC.


Military grade has different meanings. I’ve worked in the electronics industry a long time and will say with confidence that the pcbs and chips we sent to the military were our best. Higher temperature ranges, much more thorough environmental testing, many more thermal and humidity cycles, lots more vibration testing. However we also sell them for 5-10x our regular prices but in much lower quantities. It’s a failed meme in many instances as the internet uses it though.


Basically, whatever the liability insurance wants for you to be in compliance, than that’s the standard.


Hot take, this is the way it should be. If you want better security then you update the requirements to get your certification.

Security by its very nature has a problem of knowing when to stop. There's always better security for an ever increasing amount of money and companies don't sign off on budgets of infinity dollars and projects of indefinite length. If you want security at all you have bound the cost and have well-defined stopping points.

And since 5 security experts in a room will have 10 different opinions on what those stopping points should be— what constitutes "good-enough" they only become meaningful when there's industry wide agreement on them.


There never will be an adequate industry-wide certification. There is no universal “good enough” or “when to stop” for security. What constitutes “good enough” is entirely dependent on what you are protecting and who you are protecting it from, which changes from system to system and changes from day to day.

The budget that it takes to protect against a script kiddy is a tiny fraction of the budget it takes to protect from a professional hacker group, which is a fraction of what it takes to protect from nation state-funded trolls. You can correctly decide that your security is “good enough” one day, but all it takes is a single random news story or internet comment to put a target on your back from someone more powerful, and suddenly that “good enough” isn’t good enough anymore.

The Internet Archive might have been making the correct decision all this time to invest in things that further its mission rather than burning extra money on security, and it seems their security for a long time was “good enough”… until it wasn’t.


Yep. And worse, now matter how much you pay for security it is still possible for someone to make a mistake and publish a credential somewhere public.


> since 5 security experts in a room will have 10 different opinions

If that happens you need to seriously rethink your hiring process.


This ^

We can’t all have the latest EPYC processors with the latest bug fixes using Secure Enclaves and homomorphic encryption for processing user data while using remote attestation of code running within multiple layers of virtualization. With, of course, that code also being written in Rust, running on a certified microkernel, and only updatable when at least 4 of 6 programmers, 1 from each continent, unite their signing keys stored on HSMs to sign the next release. All of that code is open source, by the way, and has a ratio of 10 auditors per programmer with 100% code coverage and 0 external dependencies.

Then watch as a kid fakes a subpoena using a hacked police account and your lawyers, who receive dozens every day, fall for it.


[flagged]


No, it’s your demeanor that is unbecoming and not worth engaging with. Villianizing your poor behavior not successfully baiting people into replying as you want is childish too. Take a breather.


No mention or comparison with phi-3 seems odd. Isn't phi-3 leading the other models by a bit?


ϕ-3 isn't in the 7B league.


Gemma2-2B shows that Phi isn't even in the 2B league.


Phi-3 small is


Water on the leaves isn’t bad if you’re watering in the morning where it’ll soon evaporate in the sunlight.


The evaporation is the thing you want to avoid.


I was told that water beads on leaf surfaces act as lenses, creating burnt spots on the leaves. So water at night.


This is 99% urban legend. You can just barely create it in a lab with just the right plant (with thick hydrophobic trichromes) under just the right light with no wind... but that's not what happens in nature.


> but that's not what happens in nature.

Exactly. When it rains in nature, 95% of the times a) there isn't enough sunlight for the droplets to focus and make a burn spot, and b) the droplets don't stay on the leaf but flow down instead.

The original advice is solid and not an urban legend, but it applies to cases like watering plants in your balcony when the sun is out, bright and hot. Source: I have caused burn spots in plants of my own.


Those “burn spots” are almost certainly from a fungal disease, not from some magnifying glass effect. https://s3.wp.wsu.edu/uploads/sites/403/2015/03/leaf-scorch....


I was told to water at night because it doesn't make much sense to pour water on the ground only for it to evaporate before it goes where it's needed.


I believe watering at night will generally lead to more fungal rot problems. Better to water early morning, when the water will have a chance to sink into the soil, but will be pulled up into the plant by evaporation of water from the leaves (the leaves’ own water, not water you applied)


Well, that obviously depends from how sunny and warm it is.



And install all their dependencies from non, which Microsoft also owns


npm*


Liberland had a government. Apparently no one ever told them they didn't need it.


Why? Apple is a “premium” offering and doesn’t have a monopoly in any area. If there was no other options I would agree


Only if you use their definition of monopoly at minimum they form an Oligopoly in phones. (They repeatedly have claimed in court that all computing devices are the market they compete in with iPhone)

If you are an Apple user what real choice do you have? The cost of switching is massive for established users as everything Apple only supports Apple.

Sure an unaffiliated user can choose but once you are in one camp it requires a massive change to switch.

They utilize this to raise their prices holding the difficulty to switch against their users.

On a similar note they use heavy vertical integration to avoid options for users. You want their OS you need to buy their hardware.

"Doesn't have a monopoly" only makes sense in a make believe world where you have 95% market share or enough competition.


What massive costs of switching?


Unless you never touched the App Store there are real costs there and you could spend hours if not days transferring stuff. (Accounts cloud data etc.)

Unless you get a new same manufacturer device and then it takes minutes and you keep everything.


What “cloud data”? Most apps that are available for iOS are also available for Android and are subscription based these days.

Copying data from iCloud Drive to another cloud provider like Google Drive can be done in the Files App.

Music you bought has been DRM free since 2009.

Apple works with MoviesAnyWhere. Four or five of the major studios participate too. Meaning any movie you buy from one of the participating studios can be synced to Amazon Movies, Google Play Movies (or whatever it’s called these days), Vudu etc.

Photos can be synced to Google Photos, Dropbox, OneDrive.

Bookmarks can be synced between Chrome and Firefox and Safari if you have a Windows PC using Apple’s extension for those browsers.


Each of those steps is distinct and requires it's own process.

Those in the know can switch.

Those who just have a phone are looking at a hard to judge effort that might take months of "where was that?"

Don't just discount the literal day it would take a tech person to switch their primary phone as meaningless. Nor should you discount how many people don't even know it is meaningfully possible.

I am pretty sure I knew people back in the early days who wanted to switch so they just lost everything from before.


Is it really difficult to say “I have Spotify/Apple Music/Office 365, random other subscription. Let me download the same app on an Android phone”?

People have been using copy and paste to copy files from one drive to another since Windows 95. It works the same way with the Files app. You download the other storage provider’s app and they show up in the files app. Heck you can literally attach a USB drive to your iOS device and it shows up in the Files app like it would hold computer.

Syncing photos is as simple as downloading Google photos.

You can literally Google all of this stuff.

As far as it taking “a day”. It takes me about as long to set up a new Windows PC.

And it’s not worth a day of your life to jump over Apple’s “walled garden” where there is a big ass door that you have to push on a little bit?


This still sounds like something incredibly simple compared to most actually difficult things.


Consider applying for YC's Spring batch! Applications are open till Feb 11.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: