Message from @pratel

Discord ID: 472110545529733124


2018-07-26 18:30:59 UTC  

Yes it's a concept the same way the flying machines, and supersonic flight was a concept, yes.

2018-07-26 18:31:40 UTC  

Have you heard of No Free Lunch?

2018-07-26 18:32:00 UTC  

I know the saying.

2018-07-26 18:32:05 UTC  

It's a theorem.

2018-07-26 18:32:21 UTC  

That strictly limits the kinds of things that you can do.

2018-07-26 18:32:30 UTC  

Okay.

2018-07-26 18:32:48 UTC  

It's not as limiting as it sounds, but it does say there are fundamentally things you cannot do with any learning system.

2018-07-26 18:33:06 UTC  

There's also issues of adapting to the environment.

2018-07-26 18:33:17 UTC  

Like, you can't just take a car up an untamed mountain.

2018-07-26 18:33:38 UTC  

There's no point in AI making art for AI consumers.

2018-07-26 18:34:14 UTC  

no point != impossible

2018-07-26 18:34:44 UTC  

Sure, but it is something that AI isn't going to likely do.

2018-07-26 18:35:28 UTC  

Okay, so you don't think we're going to be turned into paper clips, got it.

2018-07-26 18:36:51 UTC  

ima head off from this convo, its rocket league time 😛

2018-07-26 18:36:55 UTC  

But I'm still not sure what this has to do with Artificial general intelligence, it would fixable enough to do any task a human can do for no ongoing cost.

2018-07-26 18:37:03 UTC  

Have fun.

2018-07-26 18:37:06 UTC  

ty

2018-07-26 18:37:44 UTC  

Let's quit talking about a mystical super strong AI that outpaces anything a human can do.

2018-07-26 18:38:10 UTC  

It's not mystical.

2018-07-26 18:38:14 UTC  

I'm saying by the time we reach that point, the conceptions of what a person is and what is doable will be radically different.

2018-07-26 18:38:45 UTC  

"Strong AI" right now is mostly good at a couple of tasks or learning straightforward new ones with lots of data.

2018-07-26 18:38:50 UTC  

I see so your argument is that it will be slow enough in development as to not be a problem?

2018-07-26 18:39:01 UTC  

One of my arguments.

2018-07-26 18:39:25 UTC  

No one has strong ai.

2018-07-26 18:39:31 UTC  

Correct.

2018-07-26 18:39:43 UTC  

Not in the "everything you can do the computer can do better"

2018-07-26 18:39:51 UTC  

General AI is not strong AI.

2018-07-26 18:39:57 UTC  

And FWIW, I'm not sure that will necessarily ever happen.

2018-07-26 18:40:01 UTC  

I didn't say better.

2018-07-26 18:40:08 UTC  

Per say.

2018-07-26 18:40:24 UTC  

Like everything I can do the robot will do way better.

2018-07-26 18:40:37 UTC  

But they will be better from an economic stand point.

2018-07-26 18:40:39 UTC  

But can it do all of them together better?

2018-07-26 18:40:55 UTC  

Integration is a fundamental and important task. It's also one that is too easily ignored.

2018-07-26 18:41:46 UTC  

Okay so we'll still have project managers?

2018-07-26 18:42:46 UTC  

But probably not even that, if we get a general intelligence, it should be able to integrate with it's self just fine.

2018-07-26 18:43:05 UTC  

And we'll still have park rangers, and we'll still have artists, and we'll still have government...

2018-07-26 18:43:14 UTC  

I disagree.

2018-07-26 18:43:40 UTC  

You have a hundred tools, you're saying just one more tool will necessarily make them all work together? And know what to build?

2018-07-26 18:43:54 UTC  

there's lots of weird rivets in these things.

2018-07-26 18:44:06 UTC  

Well when we have proff one way or the other we'll see, but the point remains that general AI doesn't have to be better than us as the job, just have to make more sense economically than us.