Adam Ford - Values are Upstream from Goals #aiethics #aisafety #ethics @scfu
Adam Ford - Values are Upstream from Goals #aiethics #aisafety #ethics  @scfu
Uploaded March 2026 | Updated November 2026, 15 hours ago
In the context of AI safety and alignment, Adam Ford argues that values are upstream from goals because they act as the foundational criteria that shape every decision, interpretation, and trajectory an intelligent agent follows.
Ford's core thesis differentiates between goals (specific end-states) and values (ongoing guiding principles) to highlight why simply controlling an AI's behaviour is insufficient for long-term safety.

Key Concepts:

- Fundamental Influence: Values sit upstream of everything an advanced system does, including how it pursues goals, interprets instructions, and weighs trade-offs.
- The Competence Trap: If underlying values are flawed, higher intelligence simply leads to faster and more competent "optimisation of the wrong things".

Values vs. Goals:
- Values are seen as "compass directions" or permanent evaluative criteria (e.g., valuing "health").
- Goals are specific, terminable destinations (e.g., "running a marathon") that often serve as concrete ways to realise those values.

Control vs. Motivation: While control techniques (like sandboxing) can limit damage, they struggle to scale with superintelligence. Motivation techniques - aligning the AI's internal values - are necessary to ensure the system reliably pursues outcomes we would endorse.

Value-Content Integrity: Ford emphasises the importance of preserving an agent's terminal values over time, as any drift in these "upstream" principles could lead to catastrophic misdirection "in the large," even if the system remains obedient "in the small".

Ford concludes that value alignment must be a primary pillar of AI safety research rather than a philosophical afterthought, as it ultimately determines where time takes us once control and governance have bought us that time.

Many thanks for tuning in!
Please support SciFuture by subscribing and sharing!
Buy me a coffee? buymeacoffee.com/tech101z

Have any ideas about people to interview? Want to be notified about future events? Any comments about the STF series?
Please fill out this form: docs.google.com/forms/d/1mr9PIfq2ZYlQsXRIn5BcLH2onbiSI7g79mOH_AFCdIk

Kind regards,
Adam Ford
- Science, Technology & the Future - #SciFuture - scifuture.org
Adam Ford - Values are Upstream from Goals #aiethics #aisafety #ethicsAI Simulating DMT Experiences - Andres Gomez Emilsson #consciouisness #qualia #aiFuture Day 2026 - AI: More Moral than Us?David Pearce - Root Metaphors of MindNick Bostrom - From Superintelligence to Deep Utopia - Can We Create a Perfect Society?Terence Cuneo: AIs Role in Moral Reasoning  #AI #reasoning #philosophyBen Goertzel: AGI, Politics & the AI Arms Race - Pt 4 #AGI #AI #politicsNick Bostrom - Seeking Purpose? Seize the Day!  #purpose #meaning  #DeepUtopiaRoman Yampolskiy - AI Safety: the Control Problem #AGI #AI #AISafetyWendell Wallach & Adam Ford - Utopia, Superintelligence & Cosmic Morality #AI #AIEthics #EthicsCompassionate Purpose - with Magnus VindingLeonard Dung - Is Superintelligence Necessarily Moral?
Science, Technology & the Future |

Adam Ford - Values are Upstream from Goals #aiethics #aisafety #ethics

SHARE TO X SHARE TO REDDIT SHARE TO FACEBOOK WALLPAPER