RSS

Limits to Control

TagLast edit: Jul 2, 2025, 7:39 AM by Remmelt

Limits to Control is a field of research aiming to discover and verify:

  1. Fundamental limits to controlling fully autonomous AI using any method of causation.

  2. Threat models of AI convergent dynamics that cannot be sufficiently controlled (by 1.).

  3. Impossibility theorems, by contradiction of ‘long-term AI safety’ with convergence result (2.)

~ ~ ~

Definitions

‘Fundamental limit’

A limit to the capacity to reach some states over other states. This limit is absolute – it is not just based on practical impediments, but on the physics or information signalling of the system itself.

‘Control’

‘Long term’

‘AI safety’

Ambient conditions caused by AI’s operations fall within the environmental range needed for the survival of humans (a minimum-threshold definition of safety).

‘Fully autonomous AI’

Any assembly of artificial components that persistently learns new functions from (and thus maintains and adapts its components across) the environment, without the need for humans or other organic life.

‘AI convergent dynamic that cannot be sufficiently controlled’

The learning/​adapting components propagate new environmental effects, where a subset of the dynamically explored state space is both unsafe and falls outside even one fundamental limit to control.

‘Verify’

To check whether an argument’s premises are empirically sound and whether its reasoning steps are logically valid.

Lenses of Control

WillPetilloOct 22, 2024, 7:51 AM
14 points
0 comments9 min readLW link

Limits to Con­trol Workshop

May 18, 2025, 4:05 PM
12 points
2 comments3 min readLW link

De­con­fus­ing ‘AI’ and ‘evolu­tion’

RemmeltJul 11, 2025, 1:44 AM
12 points
9 comments26 min readLW link

The Robot, the Pup­pet-mas­ter, and the Psychohistorian

WillPetilloDec 28, 2024, 12:12 AM
8 points
2 comments3 min readLW link

Pro­jects I would like to see (pos­si­bly at AI Safety Camp)

Linda LinseforsSep 27, 2023, 9:27 PM
22 points
12 comments4 min readLW link

The Con­trol Prob­lem: Un­solved or Un­solv­able?

RemmeltJun 2, 2023, 3:42 PM
56 points
46 comments13 min readLW link

Pro­ject Moonbeam

WillPetilloJun 27, 2025, 9:08 PM
14 points
2 comments6 min readLW link

Prin­ci­ples of AI Uncontrollability

WillPetilloAug 7, 2025, 9:10 PM
1 point
0 comments7 min readLW link

Why Re­cur­sive Self-Im­prove­ment Might Not Be the Ex­is­ten­tial Risk We Fear

Nassim_ANov 24, 2024, 5:17 PM
1 point
0 comments9 min readLW link

What if Align­ment is Not Enough?

WillPetilloMar 7, 2024, 8:10 AM
16 points
46 comments9 min readLW link
No comments.