3 ms·

I have predicted this exact VERY predictable scenario this for years and got downvotes by AI enthusiasts who don’t want to even deal with any downsides of AI.

by EGreg 3y ago

I have predicted this exact VERY predictable scenario this for years and got downvotes by AI enthusiasts who don’t want to even deal with any downsides of AI. Examples: https://news.ycombinator.com/item?id=35688266 https://news.ycombinator.com/item?id=35688266 We are racing towards the abyss orders of magnitude faster than with climate change or nuclear proliferation, and even the overwhelming majority AI experts coming out and saying there is at least 20% chance of a global catastrophe or even risk of extinction earns a mere shrug: https://arstechnica.com/information-technology/2023/05/openai-execs-warn-of-risk-of-extinction-from-artificial-intelligence-in-new-open-letter/ https://arstechnica.com/information-technology/2023/05/opena... And CEOs: https://amp.cnn.com/cnn/2023/06/14/business/artificial-intelligence-ceos-warning https://amp.cnn.com/cnn/2023/06/14/business/artificial-intel... And yet even the most mild, libertarian-friendly proposal to mitigate the harm is utterly rejected by AI fans who gang up on any criticism, as the future botswarms will: https://news.ycombinator.com/item?id=35808289 https://news.ycombinator.com/item?id=35808289 I said the entire internet will turn into a dark forest, including all forums like Reddit and even HN. Swarms of bots will cooperate to “gang up on” opponents and astroturf to shift public opinion. And the worst part will be when humans start to PREFER bots the way organizations already do (eg trading bots replaced humans on wall street). The AI people are building a dystopian future for us and won’t face ANY accountability or disincentives, but rather the opposite. I expect this post to be downvoted by AI people chafing at any criticism. (Like the opposite of web3 posts.) The replies, if any would even appear, will be predictably “well, it was all already possible with human efforts”, ignoring the massive difference in scale and cost to malicious actors (well, the replies would have been that if I didn’t call it out just now, because they always are, and hardly any actual substantive discussion of the extreme dangerous outcomes that are only starting to come about in very early stages).

jprete 3y ago
I've made the same prediction. It was blatantly obvious to me what would happen as soon as I saw GPT 3.5 producing decent quality responses. I had hoped the finger problem of image generators would last longer, but there are a lot of people with absolutely no foresight on the potential downsides of technology. SORA and other video generators are absolute madness.
LargeWu 3y ago
Can anybody explain, specifically, what that 20% risk looks like? The most specific I ever see is "an adversarial AI will become sentient and wipe out humanity". It sounds like as much snake oil as the people pushing AI itself.
- EGreg 3y ago
  It doesnt need to become sentient to cause great disruption. 1. Bot swarms will simply disrupt everything about the Internet as we know it. Most people ALREADY barely scritinize chats and articles, so bots can EASILY produce those at scale to push opinion in any direction, or just sell shyt 2. Botswarms will outplay adversarrial games vs humans for karma / reputation points, as well as launch coordinated attacks on opponents organically trying to stop whatever viewpoint is being gradually pushed or sold, until they give up or are totally reputationally discredited 3. People start to PREFER bots to humans, just as they PREFER google maps to asking for directions etc. At that point most humans would be surrounded by 100-1000 bots and have no way of affecting other humans. 4. Physical world, cameras capturing all the info and cross correlating where you are. Maybe slaughterbots are mass-produced. Who knows. When the costs come down and scale goes up, it doesnt matter about AGI, the entire society is disrupted permanently. And that’s what AI is on track to so. It’s far easier to continually create a mess than to continually clean it up.
  - LargeWu 3y ago
    You see, the jump from "Bots take over all the karma points on social media sites" to "Slaughterbots" is a pretty wide chasm I'm having trouble getting over mentally. This is why I can't take such predictions seriously.
    - EGreg 3y ago
      Okay. So remove point 4 and it’s still very dystopian… Not to mention point 4 contains things that have been in place already for over a decade. It’s not even a prediction: https://magarshak.com/blog/?p=169 https://magarshak.com/blog/?p=169 But sure, take the one tiny thing you can caricature and ignore the rest. That’s one step up from strawman, I guess

by EGreg 3y ago

jprete 3y ago

I've made the same prediction. It was blatantly obvious to me what would happen as soon as I saw GPT 3.5 producing decent quality responses. I had hoped the finger problem of image generators would last longer, but there are a lot of people with absolutely no foresight on the potential downsides of technology. SORA and other video generators are absolute madness.

LargeWu 3y ago

Can anybody explain, specifically, what that 20% risk looks like? The most specific I ever see is "an adversarial AI will become sentient and wipe out humanity". It sounds like as much snake oil as the people pushing AI itself.

EGreg 3y ago
It doesnt need to become sentient to cause great disruption. 1. Bot swarms will simply disrupt everything about the Internet as we know it. Most people ALREADY barely scritinize chats and articles, so bots can EASILY produce those at scale to push opinion in any direction, or just sell shyt 2. Botswarms will outplay adversarrial games vs humans for karma / reputation points, as well as launch coordinated attacks on opponents organically trying to stop whatever viewpoint is being gradually pushed or sold, until they give up or are totally reputationally discredited 3. People start to PREFER bots to humans, just as they PREFER google maps to asking for directions etc. At that point most humans would be surrounded by 100-1000 bots and have no way of affecting other humans. 4. Physical world, cameras capturing all the info and cross correlating where you are. Maybe slaughterbots are mass-produced. Who knows. When the costs come down and scale goes up, it doesnt matter about AGI, the entire society is disrupted permanently. And that’s what AI is on track to so. It’s far easier to continually create a mess than to continually clean it up.
- LargeWu 3y ago
  You see, the jump from "Bots take over all the karma points on social media sites" to "Slaughterbots" is a pretty wide chasm I'm having trouble getting over mentally. This is why I can't take such predictions seriously.
  - EGreg 3y ago
    Okay. So remove point 4 and it’s still very dystopian… Not to mention point 4 contains things that have been in place already for over a decade. It’s not even a prediction: https://magarshak.com/blog/?p=169 https://magarshak.com/blog/?p=169 But sure, take the one tiny thing you can caricature and ignore the rest. That’s one step up from strawman, I guess