zlacker

[parent] [thread] 24 comments
1. lannis+(OP)[view] [source] 2023-11-18 07:45:48
Doomerism when it comes to technology has always been such a weird mindset to me. "oh no, they're gonna take the horses/horse buggies/telegram/landlines away!"

"Goddamn, how dare they invent the bigger cannons!?" - Romans, 1453, in Constantinople probably. (One incident where I can use my exempt powers).

replies(3): >>waihti+g >>aidama+F8 >>dorkwo+P8
2. waihti+g[view] [source] 2023-11-18 07:47:13
>>lannis+(OP)
its not even that, they have literally one argument and its nanobots
replies(3): >>anonym+75 >>3cats-+g8 >>Feepin+4n
◧◩
3. anonym+75[view] [source] [discussion] 2023-11-18 08:31:30
>>waihti+g
The more realistic argument is that AI will be used as a power amplifier by the already powerful.
replies(2): >>waihti+Q5 >>Exoris+h8
◧◩◪
4. waihti+Q5[view] [source] [discussion] 2023-11-18 08:37:07
>>anonym+75
actually is the opposite, it would democratize power at an unprecedented scale, that's why corporations are funding these NGO's (useful idiots)
replies(1): >>anonym+u7
◧◩◪◨
5. anonym+u7[view] [source] [discussion] 2023-11-18 08:53:19
>>waihti+Q5
Through what mechanism would it democratize power? I thought the GPTs were already limited to regular end users due to computational constraints. Most people can't afford dozens of Nvidia GPUs and the API infrastructure to data mine.
replies(1): >>waihti+nf
◧◩
6. 3cats-+g8[view] [source] [discussion] 2023-11-18 08:59:42
>>waihti+g
Don't misrepresent the problem.
replies(1): >>waihti+sp
◧◩◪
7. Exoris+h8[view] [source] [discussion] 2023-11-18 08:59:49
>>anonym+75
Nah other way around. It would amplify the majority. That's why the powers-that-be consider it a huge potential problem.
replies(1): >>Generi+Oc
8. aidama+F8[view] [source] 2023-11-18 09:02:22
>>lannis+(OP)
Most of the time the technology is not building an artificial intelligent computer that is capable of superhuman reasoning.

Have you used gpt4? it's reasoning capabilities match human ability. The more you think about it, the more scary the reality becomes. GPT 4 can reason through any mental exercise as well as a human. The rest of the work to make it autonomous is simple in comparison.

replies(7): >>konsch+Qc >>waihti+zm >>jerpin+es >>skepti+lZ >>Davidz+Fe1 >>kevinv+Ug1 >>lannis+iA5
9. dorkwo+P8[view] [source] 2023-11-18 09:03:37
>>lannis+(OP)
I agree. Critics will say something like "but don't you think it's sad that a machine was trained on the work of artists without their permission, and is now decimating the lives of the very artists who made its existence possible in the first place?". And, I agree, on its face it does sound sad. Very sad. But what these critics need to remember is that progress has always been a good thing, all throughout history. If something was true in the past, it will continue to be true in the future, and while it may seem difficult to envision how this will create a better world for all of us right now, it's important that we have faith and keep marching forward regardless.
replies(1): >>jddj+8d
◧◩◪◨
10. Generi+Oc[view] [source] [discussion] 2023-11-18 09:38:24
>>Exoris+h8
Ah yes, because all those normal people will be able to run these powerful models on the devices that they currently own. Such a naive take.

The rich will ALWAYS get their piece of the pie, and once they've had their fill, we'll be left fighting for the crumbs and thanking them for their generosity.

AI won't solve world hunger, it will make millions of people jobless. It won't stop wars, it will be used as a tool for the elite to spread propaganda. The problems that plague society today are ones that technology (that has existed for decades) can fix but greed prevents it.

◧◩
11. konsch+Qc[view] [source] [discussion] 2023-11-18 09:38:30
>>aidama+F8
> GPT 4 can reason through any mental exercise as well as a human.

So can I.

And yet, people don't consider me an existential threat.

Mostly because I do not have nukes.

replies(2): >>Frustr+Id >>Feepin+Um
◧◩
12. jddj+8d[view] [source] [discussion] 2023-11-18 09:41:10
>>dorkwo+P8
Counterpoint: dogma often led to dark times.

We could try to think a little more deeply about things than "let jesus take the wheel"

◧◩◪
13. Frustr+Id[view] [source] [discussion] 2023-11-18 09:46:49
>>konsch+Qc
Yes. This point is almost always missed. 'Human Level' itself is not very high. And, 'Human Level' is still like Homer "doh, they unplugged me".

Current GPT doesn't have a physical threat.

But, take something like the movie "Colossus". Where they did give control of nukes. That was scary.

Now, go watch the Netflix show about AI. This GPT stuff is so far just fun apps.

The military already has AI that can out pilot a human in a F-16, you think it will stop there ? That is probably already old news.

◧◩◪◨⬒
14. waihti+nf[view] [source] [discussion] 2023-11-18 10:02:39
>>anonym+u7
computers used to fill rooms and now you carry one in your pocket
replies(1): >>anonym+nf1
◧◩
15. waihti+zm[view] [source] [discussion] 2023-11-18 11:01:41
>>aidama+F8
Yann LeCun disagrees with you, and I take his word on it
◧◩◪
16. Feepin+Um[view] [source] [discussion] 2023-11-18 11:04:34
>>konsch+Qc
Yes but you can't scale your brain by building a bigger womb.
◧◩
17. Feepin+4n[view] [source] [discussion] 2023-11-18 11:06:02
>>waihti+g
Nanobots are easy and convenient (for a superintelligence). But it's not like they're necessary. ASI can take over the world the old fashioned way, it just takes longer and is harder to explain.
◧◩◪
18. waihti+sp[view] [source] [discussion] 2023-11-18 11:24:51
>>3cats-+g8
what is the problem, in tangible terms?
replies(1): >>3cats-+4V
◧◩
19. jerpin+es[view] [source] [discussion] 2023-11-18 11:44:07
>>aidama+F8
gpt4 is capable of reasoning “in distribution”. Its reasoning drops when you go outside the goldilocks zone
◧◩◪◨
20. 3cats-+4V[view] [source] [discussion] 2023-11-18 14:44:41
>>waihti+sp
Machines have so far replaced us in physical tasks, which has forced us to move largely to menial office jobs, typing on a computer, doing things machines are bad at. Over 80% of jobs in the US are office-confined (or from home, but that's not the point). We're actually a poor fit for those monotonic, sedentary jobs, our bodies and minds are not designed for them. And from that the subsequent devastating effects on our physical and mental health. But you gotta have a job, or you can't exist. The system throws out parts that are not useful. It's the nature of the system.

Well here comes AI to take those jobs. What happens, you think? Where do we go next? Do you imagine we'll all just sit idle and give out orders for the AI to fulfill? Recall: the system throws away parts that are not useful. And we're not better at orchestrating this system than we are at implementing it. Most people already struggle to handle the complexity of modern life. So they'll be thrown out.

Now think what happens with a society where most people are unemployed, unhappy and hungry, and businesses are mostly, not ENTIRELY mind you, but most self-sufficient machinery that does the thinking and does the footwork?

But even that doesn't describe the problem alone. It's more of an end game. Before this we'll see not-so-superior AI pollute our web, media, public space with quickly generated content, as actual artists and thinkers are displaced, unable to compete. Our culture will die first. And then, eventually, we'll start dying.

As I'm describing this, note I don't say this from place of fear. I don't fear this. I see it more as an obvious place for our civilization to go. We can't help it, because we don't decide where this civilization goes any more than your cells decide where you go, or any more than the atoms of your cells "decide" where the cell goes.

We're not in control. That's just evolution.

Say, when you're sick and you have cancer, those cells are part of you, but they harm you, so you cut them out, apply chemotherapy, and then if there's a prosthesis to substitute the organ you removed with a machine, you do it, and you don't think twice about it.

What makes you think our society as a whole is different? If humans are not good at what society needs, it cuts those people out, and replaces them with working machines. It's so plainly obvious. We pay lip service to human rights and the value of an individual, but clearly that's not what we end up doing. A politician is chasing money and power, and they don't mind starting wars to get them if they can. A business chases profit, so they don't mind automating away any employee they can. It's always been this way. So now that you can replace the human thinkers, businesses won't need human thinkers. And since there's nothing left humans are good at, society won't need humans.

◧◩
21. skepti+lZ[view] [source] [discussion] 2023-11-18 15:12:13
>>aidama+F8
> it's reasoning capabilities match human ability

All research that I have seen disagrees with this take. Ask GPT-4 a few basic block world problems and see for yourself if it can match human ability.

◧◩
22. Davidz+Fe1[view] [source] [discussion] 2023-11-18 16:42:50
>>aidama+F8
Please. Next year maybe but have a little respect please
◧◩◪◨⬒⬓
23. anonym+nf1[view] [source] [discussion] 2023-11-18 16:46:33
>>waihti+nf
And yet rich companies and government agencies still have computers which fill rooms. https://www.energy.gov/supercomputing-and-exascale

And those which are carried in our pockets are no longer capable of being home brewed.

◧◩
24. kevinv+Ug1[view] [source] [discussion] 2023-11-18 16:53:45
>>aidama+F8
GPT 4 can reason through any mental exercise as well as a human

Lol try giving it any of the puzzles from here: https://momath.org/home/varsity-math/complete-varsity-math/

Don’t just accept its confident tone, read through and actually parse the logic. It totally falls apart.

◧◩
25. lannis+iA5[view] [source] [discussion] 2023-11-19 21:14:19
>>aidama+F8
>Have you used gpt4? it's reasoning capabilities match human ability.

I use it every day, and I have to often guide it like a 5 year old to come to the conclusion to help me the way I want it to.

>GPT 4 can reason through any mental exercise as well as a human.

So can my alcoholic neighbor. That should not be a benchmark of anything.

[go to top]