Facebook PixelWhat would you have the world's most intelligent, creative, and resourceful people work on?
Brainstorming
Brainstorming
Create newCreate new
EverythingEverything
Sessions onlySessions only
Ideas onlyIdeas only
Brainstorming session

What would you have the world's most intelligent, creative, and resourceful people work on?

Image credit: fauxels / pexels

Loading...
Darko Savic
Darko Savic Sep 08, 2020
7
Creative contributions

Work against the major threats to humanity

Loading...
Subash Chapagain
Subash Chapagain Sep 09, 2020

Real-Time Health Supervision Technology

Loading...
J
Juran Sep 09, 2020

Eradicate Poverty and Hunger

Loading...
Mohammad Shazaib
Mohammad Shazaib Sep 08, 2020

Assign them under George Church to work on his lab's projects

Loading...
Darko Savic
Darko Savic Oct 05, 2020

AI-blocking firewalls everywhere?

Loading...
J
Juran Oct 20, 2020
Loading...
Povilas S
Povilas S2 months ago
I'm also a proponent of AI safety, but there's another side to your suggestion. By restricting AI from thinking and making decisions "on its own" we would take away the whole point of building an independent or in other words conscious being. And that's what AI research is essentially aiming for, making something more than just a tool. Or at least that's a huge part of what drives people's interest in it. Whether it's even possible to do it, that's another question, but I think people will keep aiming for it because it's a charming thing. Maybe it's better to approach solving the danger problem by other means, e.g. like formulated in this session: https://brainstorming.com/sessions/how-can-we-make-ai-comprehend-human-ethics/68.

Also, keeping AI capabilities strictly limited to the algorithm doesn't guarantee safety, because it's oriented towards progress and when the power of intelligence reaches very profound levels, we can no longer predict its "decisions" even though they are defined by the algorithm. Just like it happened with AlphaGo playing the Go game, at a certain point the developers couldn't understand why AI was making moves that it was making - it was beyond their capability to grasp it with their own intelligence even though they build the algorithm. And that's what people concerned about AI often point to, that the solution of AI that we can't predict might turn against us, for example, that the ultimate solution to "how to solve environmental problems" for AI might appear to extinguish humans. This is a very bold and predictable example, but it's entirely possible that a malevolent (from our perspective) solution might be offered to some apparently simple and unrelated problem. Of course, we might not allow the AI to act on the solution in the first place, but if it thinks several magnitudes ahead of us it might find ways how to bypass that problem as well.

The point here is that the very restriction to act within the frame of the limited quest, like just finding the solution to a particular problem, might lead to something unexpected and possibly undesirable. In this perspective maybe it's even better if AI is allowed to "think more broadly".
Loading...
Shubhankar Kulkarni
Shubhankar Kulkarni2 months ago
Povilas S Agreed. AI may need all the information there is to think and come up with a solution. Here are a few things that can be done to let AI be free to think up a solution and parallelly not be a menace to the human race:
1. Offer the AI the power only to know stuff and offer solutions. The AI should not be able to "act" on its own.
2. Disconnect it from the real internet. Create an exact copy of the internet and feed that data to AI. That way AI knows what is happening but cannot make changes to it on its own.
Loading...
Shubhankar Kulkarni
Shubhankar Kulkarni4 months ago
I agree. Every AI algorithm, probably, needs a kill switch.

Pick anything from the yearly Global Risk Report

Loading...
Dragan Otasevic
Dragan Otasevic Sep 30, 2020

Engineering happiness

Loading...
Povilas S
Povilas S Jan 16, 2021

Add your creative contribution

0 / 200

Added via the text editor

Sign up or

or

Guest sign up

* Indicates a required field

By using this platform you agree to our terms of service and privacy policy.

General comments