Limits to Control Workshop
Intro
Can we keep enough control over AI? If systems are developed to be more and more autonomous, this is no longer a given. Itās a hypothesis, calling for serious investigation. Even alignment relies on control. Researchers build mechanisms to control AIās impacts in line with human values.
So how must a control mechanism operate? What limits its capacity to track and correct all the AI signals/āeffects? Does it provide enough stability, or give way eventually to runaway impacts?
We explore these questions in a new field: Limits to Control.
About the workshop
This in-person workshop is meant to facilitate deep collaboration. Weāre bringing together researchers to map the territory of AI control limitations ā to understand the dynamics, patterns, and impossibilities of control. We are thrilled to welcome Roman Yampolskiy, Anders Sandberg, Forrest Landry, and other researchers working on control limitations.
This workshop aims to:
Build common knowledge among researchers on the limits to AI control.
Facilitate high-fidelity discussions through whiteboard sessions and collaborative problem-solving.
Identify and clarify viable directions in control research by establishing boundaries on controllability.
Formalize and elevate this critical research topic.
Dates & location
Date: Wednesday, June 11 - Friday, June 13, 2025
Location: University of Louisville, Kentucky, USA.
Detailed logistical information will be provided to confirmed participants.
Agenda
We aim to strike a balance between structured sharing and messy exploration ā we believe this is where the best ideas tend to emerge. Over the three days, we will do:
Talks: Researchers present their current work, keeping it under an hour.
Discussions: We break out into groups to discuss specific questions and bounce ideas around.
Regrouping: We come back into one room to synthesize what came out of the discussions.
Next steps: On the final day, weāll plan further development of this research agenda.
Sessions
Sessions will include:
Anders Sandbergās talk on theoretical limits to control: āDo any of them actually tell us anything?ā
Forrest Landryās whiteboard chat on an overlooked dynamic: āVirtual machines in recursive feedbackā
Richard Everheartās logical framework for AI alignment, aiming to refine foundational understanding.
Thibaud Veronās session on a framework and engineered toy models that illustrate control dynamics.
Will Petilloās session on better communication and narrative framings for AI safety/ācontrol concepts.
(More details on specific talks and activities will be added as confirmed.)
Proceedings, post-workshop outputs
This section will be updated after the workshop with summaries, key insights, and any public materials generated.
Potential outputs may include:
Summary report of discussions, key agreements, and open questions
Notes or photos from sessions
Links to research papers, blog posts, or pre-prints influenced by the workshop
A refined list of open research questions in the ālimits to controlā domain
Presentations or slide decks (if speakers consent to public sharing)
Join us
This workshop is for researchers actively working on or deeply interested in the theoretical and practical limits of AI control.Do you wish to contribute to these focused discussions? Email Orpheus at o@horizonomega.org to express your interest.
Costs & funding: Participants are generally expected to cover their own travel and accommodation. We can reimburse only some whose research is not yet funded. The workshop has a grant offer from Survival and Flourishing Fund.
To prepare: Read work by participants you are curious to chat with. Then we share some understanding already going in. Most of our time will be in collaborative discussions, so consider where you could bring in specific problems or concepts.
Suggested reading
Writings by participating researchers:
Papers:
On the Controllability of Artificial Intelligence: An Analysis of Limitations, by Roman Yampolskiy
AI: Unexplainable, Unpredictable, Uncontrollable, by Roman Yampolskiy
Impossibility Results in AI, by Roman Yampolskiy
[Forthcoming paper on control limits] by Anders Sandberg, Aybars Kocoglu, Thibaud Veron
[Forthcoming paper on a logical framework] by Richard Everheart
Essays:
Lenses of Control, by Will Petillo
The Control Problem: Unsolved or Unsolvable?, by Remmelt Ellen
Control as a Causative Feedback Process, by Forrest Landry
An Exploration of AGI Uncontainability, by Forrest Landry
On Error Detection, by Forrest Landry
Sorted roughly by ease of reading. To clarify an argument, do reach out to authors. Authors value questions!
Organizing team
This event is hosted by HĪ©, and organized by:
Orpheus Lummis (HĪ©)
Remmelt Ellen
Thibaud Veron
Contact
For inquiries regarding the workshop, please contact Orpheus at o@horizonomega.org.
I think a key element here is (the distribution of) whether the AI is motivated (i.e. acts as if motivated) to want to be controlled by us, or not ā i.e. something like corrigibility or value learning, and whether we can arrange that. Is the workshop expected to cover that question?
Thank you for your feedback. The workshop arranges a lot of time for free discussions, so the motivation of the AI to be controlled might pop up there. Under the current proposals for talks, the focus is more on the environment in which the agent evolves than the agent itself. However, discussions about the ālevel of trustā or ālevel of cooperationā needed to actually keep control is absolutely in the theme.
On a more personal level, unless I have very strong reasons to believe in an agentās honesty, I would not feel safe in a situation where my control depends on the agentās cooperation. As such, I would like to understand what control look like in different situations before surrendering any control capabilities to the agent.
Whether or not we decide to put an agent in a situation that we canāt keep under control if the agent doesnāt wish to be controlled is an interesting topicābut not on the agenda for now. If youād like to participate and run a session on that topic, youāre more than welcome to apply!