Another way to describe GANs is that the loss function, which used to be hand-made, now is a whole separate neural net and can be learned bottom-up instead of dictated top-down.
There was a time when features were hand-made. Now we have deep learning and finding good features is automatic. GANs automate another part of the process.
I see 13 year olds watching their favorite youtube stars showing off the new products they're being sponsored with this week. How exactly would you block this?
Also, your username. lol
Streaming video on the scale of youtube or facebook is not sustainable due to bandwidth cost.
So even today even 100% human "ad blockers" are unable to block a lot of ads because they are unaware they are looking at one.
Ads as content works really well, so well that people refuse to believe it. Huh? It's just a funny picture of a tall, frosty bottle of Coca-Cola!
Soon Adsense will offer server-side content placement instead of just 3rd-party-served banner ads. People act like the arms race is over because they added some entries into a hosts file. Not even close.
I don't object to the existence of advertising. I object to advertising that is intrusive and/or resource-intensive enough to impair site usability. (And malware served over ad networks, of course, but that's a bit different as a problem.)
Ads subtle enough to get past an "AI Guardian" mean they are meant to influence your behavior. That is my issue with advertisements and why I actively refuse to purchase products I'm introduced to through advertising.
E:
I'd like to clarify that by "make good advertisements" I mean as in the goal of advertisements (get people to buy) and not "good" as in "acceptable".
Has the added bonus of fixing a lot of news in the process.
(Okay, so it's slightly more complicated in practice -- you would have to reduce the manipulative content there, and then add back a certain chosen bias so the reduced manipulation is overpowered by the chosen one.)
So, no, we won't have AI tuning the advertisements so that they can't be catch by software like this one because that means it also wouldn't be catch by humans and that a very big no.
I can see this as a paradigm change in the internet it's true, and it's hard to see the full reprecursions of this all.
Then again, I think that if we get back to sane advertisements in web pages, that are just a couple of lines of text clearly marked as such (like Google ads used to be), then people would be ok not using an ad blocker in those cases. The thing is, the marketing companies just pushed it too far...
That said, WRT GP's post, I don't think it's a million miles off. AI will undoubtedly become a cornerstone of modern life and I don't think the scenario they outline is entirely unlikely. Siri is objectively pretty hopeless, particularly when compared to its counterparts but it's been a nice clue as to what we can expect from modern innovation.
The problem with that attitude is that when you are looking for bad things you are looking for intent, which is really not necessary for bad things to happen. This is like looking at individual neurons and not finding psychopathy (or whatever else actually is a system outcome of the network's actions, not of individuals).
At some point we will have AI designing and AI delivering ads, and while we may have AI designed to prevent us from having to watch ads we don't want we will also have AI that watches everything, gathering and filtering information that is too much for us to handle but tuned to our needs since it's "our AI". Then the race will be that one AI wants to trick the consumer AI into giving their information more weight and attention.
So instead of the race humans against humans we'll have a race humans => AI ("sellers" of anything, from goods to news) => AI (consumers) => human (us).
It's going to be a lot more complex: Right now all that people on both sides have to know is human psychology. In that future they'll have to understand the potentially far more varied world of possible AIs - and if that isn't enough the complex interactions between them and also between the AIs and the humans.
Are we creating the diversity and complexity that we remove from the biosphere (the ongoing mass extinction and/or reduction of many species) anew but in a completely different space? In addition to technical systems we are also getting much better (and better faster!) in controlling biological systems, creating our own ideas. At least some programmers of the future will write their code in DNA - or possibly even something more complex, something that can encode completely new proteins that the current code can't represent. And then there's combining biology and technology... an explosion of complexity and diversity?
I studied CS more than two decades ago. I kept up to date and continue to do the odd course in my field, but what I consider an amazing experience (for an IT guy) was when I spent the last few years taking hundreds of hours of courses in biology and medicine. Looking for new ideas? Take an introduction to biology and genetics course instead of learning an only very mildly different programming language, for example (free): https://www.edx.org/course/introduction-biology-secret-life-...