Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Microsoft Purview: Additional classifiers for Communication Compliance (preview)

pupuweb.com

Post reply on HN
Print view View original

201–210 of 317 posts

Re: Microsoft Purview: Additional classifiers for Communication Compliance (preview)

#201

Post by throwaway290 »

dijit wrote: post #159
fartcannon wrote: post #125

Earlier quoted context omitted.

Signal and WhatsApp aren't 100% trust worthy though. Why not pick something you can host yourself?

Right, I'm tired of this. E2EE doesn't mean anything if you have the same entity controlling the server as is controlling the endpoints. If you control both ends of an E2EE communication and they are closed then you gain nothing over normal TLS encryption, you still trust the authority. (Whatsapp is obviously closed and yes, signal can be considered effectively closed as their client is not reliably or reproducably b…

If you say something ML thinks is wrong in Teams, you can be fired (at will).

If you say something to your colleague via Whatsapp, the only scenario it can be used against you is if you commit an actual crime with reasonable evidence, they subpoena the records, and FB will be willing to go on record to the entire world as lying about Whatsapp E2EE, all in the name of putting you behind bars.

(Also, maybe we can imagine that products actually do what they do and it is not normal to fear lies and nefarious agenda behind every offering?)

Re: Microsoft Purview: Additional classifiers for Communication Compliance (preview)

#202

Post by runnerup »

davesque wrote: post #145
dang wrote: post #131

Submitted title was "Office 365 implementing AI to detect employees colluding, leaving and more". That broke the site guidelines: " Please use the original title, unless it is misleading or linkbait; don't editorialize. " - https://news.ycombinator.com/newsguidelines.html The proper place to include that sort of interpretation is by adding it in a comment in the thread. Then your interpretation is on a level playing…

I know it's policy to use original titles, but the editorialization in this case hardly seems sensational. Just look at the linked roadmap tickets: * https://www.microsoft.com/en-my/microsoft-365/roadmap?filter... * https://www.microsoft.com/en-my/microsoft-365/roadmap?filter... * https://www.microsoft.com/en-my/microsoft-365/roadmap?filter... * https://www.microsoft.com/en-my/microsoft-365/roadmap?filter... * https:…

> I know it's policy to use original titles, but the editorialization in this case hardly seems sensational.

It's interesting that the HackerNews guideline makes no statement about whether a custom headline is sensational or reasonable. It is: "Please use the original title, unless it is misleading or linkbait; don't editorialize." They probably have a slightly different reason for this rule than many people first imagine. And that reflects in the actual wording of the rule being slightly different than many people would first phrase it themselves.

Re: Microsoft Purview: Additional classifiers for Communication Compliance (preview)

#203

Post by creatonez »

21723 wrote: post #103

This is absolutely going to be used against unionizers, which is what's really meant by "colluding". In the US this is going to get a lot of people fired. In other parts of the world, it's going to get them killed. This kind of software is Zyklon B for the 21st century.

Sure, surveillance capitalism is pretty horrifying, but

> This kind of software is Zyklon B for the 21st century

is a bit of an over-the-top comparison

Re: Microsoft Purview: Additional classifiers for Communication Compliance (preview)

#204

Post by technion »

parasubvert wrote: post #168
outworlder wrote: post #120

Earlier quoted context omitted.

Except it's an AI making decisions and flagging people who may then be the recipient of adverse actions.

News flash: there used to be entire departments of humans that did this (and in some companies there still are). Your corporate comms are monitored and there is no privacy.

This is the part that I don't get.

15 years ago I worked at a place where there was an entire room of people who were hired to literally do nothing but read your internal mail all day. One of the deployment rules were to make sure they had unimpeded access to everything (except for executives of course). Please don't misinterpret me as suggesting I like the practice.

I can see why people are upset that this technology is being offered, I am too. But I can't see why people are suggesting it's a new low for corporations that have always been doing this.

Re: Microsoft Purview: Additional classifiers for Communication Compliance (preview)

#205

Post by bufferoverflow »

runnerup wrote: post #198
bufferoverflow wrote: post #185

Earlier quoted context omitted.

1) Nonsense. If you don't trust other people's code, you're screwed. You put yourself into the position where you have to audit your OS code, your CPU code, code of every driver that runs in your system. None of which you did. 2) Isn't WebRTC open source too? 3) Their code, their decisions.

You went from: > What's not trust worthy exactly? to: > Their code, their decisions. It's okay to be a fanboy! Evangelism is needed for any great product/company/ideology. But on HN you'll get typically called out for disingenuous or bad-faith lines of rhetoric. The person above gave you a perfectly reasonable answer to your original question of "What about Signal is not trustworthy?". It'd be kind to acknowledge tha…

> You went from:

>

> > What's not trust worthy exactly?

>

> to:

>

> > Their code, their decisions.

Two separate comments addressing two different points. One doesn't follow from the other. Stop arguing in such dishonest manner.

Re: Microsoft Purview: Additional classifiers for Communication Compliance (preview)

#206

Post by bufferoverflow »

fartcannon wrote: post #194
bufferoverflow wrote: post #185

Earlier quoted context omitted.

1) Nonsense. If you don't trust other people's code, you're screwed. You put yourself into the position where you have to audit your OS code, your CPU code, code of every driver that runs in your system. None of which you did. 2) Isn't WebRTC open source too? 3) Their code, their decisions.

Their code, their decisions is why it's bad. If the decide to start tattling to Microsoft too, what are you going to do? If it's open, we can fork it and move on with your lives. Free and open gives you and me the power to control our own communication.

> If the decide to start tattling to Microsoft too, what are you going to do?

That would be obvious in their source code, wouldn't it?

I would stop using them then.

Re: Microsoft Purview: Additional classifiers for Communication Compliance (preview)

#207

Post by dijit »

throwaway290 wrote: post #201
dijit wrote: post #159

Earlier quoted context omitted.

Right, I'm tired of this. E2EE doesn't mean anything if you have the same entity controlling the server as is controlling the endpoints. If you control both ends of an E2EE communication and they are closed then you gain nothing over normal TLS encryption, you still trust the authority. (Whatsapp is obviously closed and yes, signal can be considered effectively closed as their client is not reliably or reproducably b…

If you say something ML thinks is wrong in Teams, you can be fired (at will). If you say something to your colleague via Whatsapp, the only scenario it can be used against you is if you commit an actual crime with reasonable evidence, they subpoena the records, and FB will be willing to go on record to the entire world as lying about Whatsapp E2EE, all in the name of putting you behind bars. (Also, maybe we can imagi…

there were two points made:

1) Signal and whatsapp are not 100% trustworthy

Maybe the implication is that it will leak info to your employer, but I think this is more like a general statement; one that is likely an attempt to discuss why we still put our conversations into the hands of large companies with potentially unknown motives; and the questionable state of using "end-to-end" where one entity controls the network, access to the network and both ends of the exchange.

2) Why not use something you can host yourself

to which a reasonable reply is: network effects; I already have signal/whatsapp/telegram and I do not worry about them sending information to my employer.

Unless your employer is facebook, then I think that's a perfectly legitimate rebuttal, but one nobody is making.

In fact, people would rather argue that signal/whatsapp is the best privacy platform in the universe due to e2ee!

Re: Microsoft Purview: Additional classifiers for Communication Compliance (preview)

#209

Post by fartcannon »

hyperhopper wrote: post #137
efitz wrote: post #50

There are engineers who thought it would be a good idea to develop and train these models.

Are there? I think it's much more likely there are engineers that were told to work on this, and thought that working on this would be fun, let them have an ML project, be a learning opportunity, give them something good for their performance packet, give them good experience, be good for visibility in the org, and many things like that. I don't know why jumping to the most far reaching evil option is popular in thre…

Evil is banal.

Re: Microsoft Purview: Additional classifiers for Communication Compliance (preview)

#210

Post by marricks »

colechristensen wrote: post #170
marricks wrote: post #49

Earlier quoted context omitted.

Why are you saying this? It's what every single corporate pamphlet and forced talk says. No one is saying unions will bring everyone enlightenment and cure cancer. What people are saying is workers need a say in how the workplace is run and companies spending millions convincing folk otherwise should be forced to stop.

Few people are able to view unions objectively instead of picking a side, a union comes with tradeoffs not universal goods or evils. There are plenty of examples of unions fixing things and plenty examples of them making things worse, your opinion of a union shouldn't be based on keeping score but actually looking at the risks and rewards. Unless an organization is particularly nice, unions make a lot of sense for lo…

https://en.wikipedia.org/wiki/False_balance
Post reply on HN