3 ms·

People write fake news all of the time, why is this AI so "dangerous"? At worst it'll just add to an already existing mountain of news. At best I believe it wil

by bArray 7y ago

People write fake news all of the time, why is this AI so "dangerous"? At worst it'll just add to an already existing mountain of news. At best I believe it will prompt people to be more critical of what they read and where they read it from, after all, how many times do you need to be embarrassed by some auto-generated news sources?

bjourne 7y ago
Spam. How on earth would a spam filter determine if such an auto-generated text is spam or ham? Since it doesn't have a human brain it can't. So we get into an arms race again. Right now, spam filters seem to have the upper hand but that can change very quickly. You can also use this text generation for less nefarious purposes such as for submitting journal articles (https://pdos.csail.mit.edu/archive/scigen/ https://pdos.csail.mit.edu/archive/scigen/), creating fake facebook campaigns, creating fake github profiles, phishing on dating sites, fool Google's page ranking algorithms and so on.
- bArray 7y ago
  To be completely honest, a lot of these services need a wake up call anyway. Fake papers, fake Facebook campaigns, fake GitHub accounts, faking dating profiles and crappy search rankings are already a problem - but these services are currently unmotivated to solve these problems as they still work 99% of the time. We are already in an arms race for people's attention and data, that horse already left the stable. Kindly asking people not to produce fake news is so far not working out for us. This AI will be replicated, likely very soon, if not already.
  - bjourne 7y ago
    I don't think your argument is very convincing since this problem would affect almost every site on the internet and almost every text-based communication channel. If burglary increased 100-fold, I don't think many would see it as a wake-up call to install beefier home security systems.
    - bArray 7y ago
      > I don't think your argument is very convincing since this > problem would affect almost every site on the internet and > almost every text-based communication channel. Every text based channel is already compromised, if not by AI, then by bad actors. The only difference is that it's in lower volume than it might be with AI. It's only a matter of time before somebody else recreates this AI (I'm relatively sure I could with 2 solid months and some motivation) and the longer we delay, potentially the greater the threat. Consider this in terms of anti-virus software, do you respond to each incremental advance in virus programming or do you wait until the problem is overwhelmingly bad and you don't have years of incremental research to support yourself? > If burglary increased 100-fold, I don't think many would > see it as a wake-up call to install beefier home security > systems. Even if this increased spam 100-fold, this wouldn't happen overnight. But by locking it away, researchers can not actively work on a counter solution - so when somebody invents a better AI and releases it, they are extremely ill prepared. Not only this, the people themselves are unprepared too. I think in delaying the handling this problem, the potential for mass disruption increases - not decreases.

by bArray 7y ago

bjourne 7y ago

Spam. How on earth would a spam filter determine if such an auto-generated text is spam or ham? Since it doesn't have a human brain it can't. So we get into an arms race again. Right now, spam filters seem to have the upper hand but that can change very quickly. You can also use this text generation for less nefarious purposes such as for submitting journal articles (https://pdos.csail.mit.edu/archive/scigen/ https://pdos.csail.mit.edu/archive/scigen/), creating fake facebook campaigns, creating fake github profiles, phishing on dating sites, fool Google's page ranking algorithms and so on.

bArray 7y ago
To be completely honest, a lot of these services need a wake up call anyway. Fake papers, fake Facebook campaigns, fake GitHub accounts, faking dating profiles and crappy search rankings are already a problem - but these services are currently unmotivated to solve these problems as they still work 99% of the time. We are already in an arms race for people's attention and data, that horse already left the stable. Kindly asking people not to produce fake news is so far not working out for us. This AI will be replicated, likely very soon, if not already.
- bjourne 7y ago
  I don't think your argument is very convincing since this problem would affect almost every site on the internet and almost every text-based communication channel. If burglary increased 100-fold, I don't think many would see it as a wake-up call to install beefier home security systems.
  - bArray 7y ago
    > I don't think your argument is very convincing since this > problem would affect almost every site on the internet and > almost every text-based communication channel. Every text based channel is already compromised, if not by AI, then by bad actors. The only difference is that it's in lower volume than it might be with AI. It's only a matter of time before somebody else recreates this AI (I'm relatively sure I could with 2 solid months and some motivation) and the longer we delay, potentially the greater the threat. Consider this in terms of anti-virus software, do you respond to each incremental advance in virus programming or do you wait until the problem is overwhelmingly bad and you don't have years of incremental research to support yourself? > If burglary increased 100-fold, I don't think many would > see it as a wake-up call to install beefier home security > systems. Even if this increased spam 100-fold, this wouldn't happen overnight. But by locking it away, researchers can not actively work on a counter solution - so when somebody invents a better AI and releases it, they are extremely ill prepared. Not only this, the people themselves are unprepared too. I think in delaying the handling this problem, the potential for mass disruption increases - not decreases.