Hacker News Reader

Comment by _Understated_ | Hacker News Reader

I think we have to assume at this point that all online communication tools owned by for-profit mega-corps are compromised and that they can and do listen/analyse everything.

I have long assumed that with Whatsapp... I honestly don't believe that it's as secure as it used to be: Facebook'a raison d'etre is all about data gathering.

Also, when I read a WhatsApp message on my phone, it's unencrypted as I see it on the screen, it kind of has to be... what stops FB (or anyone for that matter) from reading it too at that point.

Sorry, I went on a tangent a bit there but I feel it was relevant.

View on HN

Specifically, you should realize anything that companies say is done with "AI" is probably in some part done by poorly paid contractors. And not always on the analysis end, for example: https://www.forbes.com/sites/johanmoreno/2019/05/28/25-of-go...

There's something incredibly dystopian by using humans as part of a process you literally tell people is entirely machines.

Haha, I never realised it was that immature.

As an aside: Being a tech I have been asked numerous times by non-techs "What actually is AI?"

I usually answer "a lot of if-statements" :)

AI researcher here. While industry doesn't usually find the need for huge complex models, I don't think your statement is fair, nor accurate. Can you provide some examples of AI tech that in your opinion is "a lot of if statements"?

How do you expect someone to justify a statement that's not true? I suppose you could say that it's a fair characterization of a basic decision tree but that doesn't describe modern ML methods.

The commenter is just referring to the nature of neural networks, a pile of weights and thresholds that you could conceivably write out as a maniacal mess of deeply nested if-statements.

If you're meaning me, I probably should've ended my comment with /s

I was making a joke :)

That's not a fair characterization. These contractors are mostly used for quality assurance. Once you've trained your model (which is not just if statements), you might you send a few thousand examples to raters to judge the true quality of your model.

It is probably something akin to "a lot of if-statements generated by an automated system that we neither understand nor can explain, which seems to more or less output results we wanted".

This reminds me of the great first novel of Russian author Victor Pelevin, Omon Ra. The protagonist joins the Soviet space program (fully automated moon exploration!) and

[spoilers ahead]

learns that the full automation is physical labor by cosmonauts at every stage of the process. No spoilers about the end, it's even better.

Compromised by what? Whom? Listening / analyzing for what?

These stories are really overblown. All these companies are doing it in order to make voice recognition better. It’s simply not possible to do with some QA... you have to know what people are actually saying vs what the system recognized.

I have nothing to back this statement up...

I am sure they are "training" their AI. I don't doubt it but they have access to masses of data. Some mega-personal I would imagine and also some that can provide insight into other businesses and markets and so on.

The temptation to "use" that data would be absolutely enormous and the financial pressures to use it to provide whatever advantage would be equally enormous.

I recall years ago reading about the FB app spying on users' usage of other apps on phones (My Google-fu is broken right now but I'll keep looking for it and update if I find it).

Imagine if <insert mega-corp> has data from one purpose/system that could provide a massive leg-up to them. Do you think they would ignore it? Perhaps in regulated industries like finance (I happen to work in finance and we have systems and procedures in place for this kind of thing) they may have to destroy it once it's been used for a very specific purpose but we're talking about Skype.

The best scenario is if the data doesn't exist or can't be read... it removes the possibility of temptation and financial pressures to use it.

I don’t think there’s as much pressing financial advantage in a 10 second clip of a couple fighting as you seem to think.

FB knowing what apps people use makes a lot of sense because it tells them how they’re doing relative to others. You’re not getting such information in the home.

Reply on news.ycombinator.com