Questions tagged [control-problem]

For question related to the "AI control problem".

Filter by
Sorted by
Tagged with
2
votes
1answer
48 views

Are linear approximators better suited to some tasks compared to complex neural net functions?

Model based RL attempts to learn a function $f(s_{t+1}|s_t, a_t)$ representing the environment transitions, otherwise known as a model of the system. I see linear functions are still being used in ...
6
votes
3answers
1k views

What would be the best way to disable a rogue AI?

Suppose that an artificial superintelligence (ASI) has finally been developed, but it has rebelled against humanity. We can assume that the ASI is online and can reproduce itself through electronic ...
1
vote
0answers
32 views

Using DDPG for control in multi-dimensional continuous action space?

I am relatively new to reinforcement learning, and I am trying to implement a reinforcement learning algorithm that can do continuous control in a custom environment. The state of the environment is ...
4
votes
1answer
1k views

Why is GLIE Monte-Carlo control an on-policy control?

In slide 16 of his lecture 5 of the course "Reinforcement Learning", David Silver introduced GLIE Monte-Carlo Control. But why is it an on-policy control? The sampling follows a policy $\pi$ while ...
1
vote
1answer
88 views

Will robots rebel against their human creators?

There are several science fiction movies where the robots rebel against their creators: for example, the Terminator's series or I Robot. In the future, is it possible that robots will rebel against ...
1
vote
2answers
75 views

Continuous control with DDPG: How to eliminate steady state error?

Currently I'm working on a continuous control problem using DDPG as my RL algorithm. All in all, things are working out quite well, but the algorithm does not show any tendencies to eliminate the ...
2
votes
1answer
41 views

Neural networks with internal dynamics in the state-space form

Neural networks with feedback (Hopfield, Hamming, etc.) differ from ordinary neural networks (multilayer perceptrons, etc.), which turns them into a dynamic element with its own internal dynamics (if ...
5
votes
3answers
473 views

Is there a way to protect humanity against the impending singularity?

We are careening into the future which may hold unpredictable dangers in relation to AI. I've haven't yet heard of Chappie or Robocop style police robots, but militarized drone tech is replacing many ...
1
vote
0answers
34 views

Solving the dead time problem for control using reinforcement learning

There are several occasion that reinforcement learning can be used as a control mean. The action is for example the set target temperature (which in many occasions change with time) and the state is ...
2
votes
0answers
28 views

How to learn using DDPG in python solely using a timeseries datasets

I have a lengthy timeseries datasets which contains several variables (from sensors etc) to be classified as actions or states. Providing they are successfully done, I want to learn a control policy ...
4
votes
1answer
59 views

Can the AI in a box experiment be formalized?

Introduction The AI in a box experiment is about a super strong game AI which starts with lower resources than the opponent and the question is, if the AI is able to win the game at the end, which is ...
6
votes
3answers
381 views

Could artificial intelligence cause problems for humanity after figuring out human behavior?

This BBC article suggests that intelligent algorithms, like those that select news stories and advertisements for display, could control our experience of the world and manipulate us. Will Artificial ...
0
votes
3answers
185 views

Can we use a messenger that does not alter the AI to solve the control problem?

Instead of directly communicating with the AI, we would instead communicate with a messenger, who would relay our communications to the AI. The messenger would have no power to alter the AI's hardware ...
3
votes
2answers
139 views

Is policy learning and online system identification the same?

In some newer robotics literature, the term system identification is used in a certain meaning. The idea is not to use a fixed model, but to create the model on the fly. So it is equal to a model-free ...
8
votes
3answers
415 views

What's the term for death by dissolving in AI?

What's the term (if such exists) for merging with AI (e.g. via neural lace) and becoming so diluted (e.g. 1:10000) that it effectively results in a death of the original self? It's not quite "digital ...
1
vote
0answers
21 views

Does a mechanical system repeats itself?

The inverted pendulum problem is a famous control task. It can be solved with a technique called system identification. System identification means to formalize the state-action space of a system in a ...
1
vote
0answers
28 views

Is an indirect policy superior to a normal one?

Before a robot can act with meaning, some planning is needed. The idea is, that the decision making process is independent from action. The task of figuring out what the best decision is, was ...
4
votes
1answer
131 views

Is learning possible without random thoughts and actions?

In my view intelligence begins once the thoughts/actions are logical rather than purely randomn based. The learning environments can be random but the logic seems to obey some elusive rules. There is ...
2
votes
2answers
121 views

How does reinforcement learning handle measured disturbances?

I recently encountered an interesting problem and was wondering how RL would solve it. The objective of the problem is to maximize the coffee quality, given by box X. The coffee quality objective ...
1
vote
1answer
148 views

Genetic Algorithm vs Particle Swarm Optimization

Which one gives better optimization results? Genetic Algorithm or Particle Swarm Optimization? Can I use them for online tuning problems? Thanks in advance!
3
votes
1answer
327 views

Finding goals in Hierarchical Reinforcement Learning

In a recent paper Data-Efficient Hierarchical Reinforcement Learning, O Nachum, S Gu, H Lee, S Levine, 2018, a promising agent controlling technique called Hierarchical Reinforcement Learning was ...
4
votes
2answers
144 views

Which government agencies oversee development of new AI?

Nick Bostrom talks in his book Superintelligence about the many dangers of AI. He considers it necessary that strong security mechanisms are put in place to ensure that a machine, once it gains ...
8
votes
3answers
3k views

Is overfitting always a bad thing?

DNN can be used to recognize pictures. Great. For that usage, it's better if they are somewhat flexible so as to recognize as cats even cats that are not on the pictures on which they trained (i.e. ...
2
votes
1answer
108 views

Which technologies should humanity develop before developing an AI with general superhuman intelligence?

The sequence in which new technologies are developed can be crucial for it's success and it's safe use. For example, before developing a nuclear reactor you want to have technologies like reliable ...
4
votes
2answers
214 views

What AI concept is behind the Mars Exploration Rover (MER)?

The Mars Exploration Rover (MER) Opportunity landed on Mars on January 25, 2004. The rover was originally designed for a 90 Sol mission (a Sol, one Martian day, is slightly longer than an Earth day at ...
9
votes
6answers
343 views

Would an AI with human intelligence have the same rights as a human under current legal frameworks?

For example, would an AI be able to own property, evict tenants, acquire debt, employ, vote, or marry? What are the legal structures in place to implement a strong AI into society?
5
votes
3answers
152 views

Will AIs make most of the scientific discoveries in 50 years?

Given the advantage AI already has over human intelligence, one could imagine a relatively weak strong-AI (barely human intelligence) still outperforming a segment of the human scientist population in ...
2
votes
1answer
2k views

What is the Control Problem?

The Wikipedia page describes AI control problem in very intricated way. Therefore I would like to better understand it based on some simple explanation, what's going on. Basically I don't want any ...