Suggestions
Iason Gabriel
Staff Research Scientist at DeepMind
Iason Gabriel is a Senior Staff Research Scientist at Google DeepMind, where he leads the Humanity, Ethics and Alignment team.124 His work focuses on the ethics of artificial intelligence, including topics such as AI value alignment, distributive justice, language ethics, and human rights.13
Background and Education
Gabriel holds a D.Phil in political philosophy from the University of Oxford.1 Before joining Google DeepMind, he taught moral and political philosophy at Oxford University and worked for the United Nations Development Program in Lebanon and Sudan.13
Career at Google DeepMind
Gabriel joined Google DeepMind in August 2017 as a Research Scientist.2 He was promoted to Senior Research Scientist in October 2018 and then to Staff Research Scientist in October 2020.2 As of November 2024, he holds the position of Senior Staff Research Scientist.4
Research Focus
At Google DeepMind, Gabriel's research primarily concentrates on:
- AI value alignment
- Ethics of advanced AI assistants
- Distributive justice in AI
- Language ethics
- Human rights in the context of AI
He is particularly interested in ensuring that AI technology works for the benefit of all and in exploring the relationship between AI and human values.1
Notable Contributions
Gabriel has contributed to several projects promoting responsible innovation in AI, including the creation of the ethics review process at NeurIPS (Conference on Neural Information Processing Systems).3 He has also been recognized as one of the leading thinkers in the field of AI ethics, being featured in the TIME 100 AI list.4
Academic and Professional Affiliations
Throughout his career, Gabriel has held various academic and professional positions, including:
- Fellow in Politics at St John's College, Oxford
- College Lecturer in Politics and International Relations at Worcester College, Oxford
- Senior International Consultant for UNDP in Lebanon
- Visiting Researcher at the World Bank Group's Mind, Behavior and Development Unit2
Iason Gabriel's work at the intersection of ethics, political philosophy, and artificial intelligence positions him as a key figure in shaping the ethical considerations surrounding the development and deployment of AI technologies.
Highlights
With these amazing technical advances from OpenAI & Google, it's a great time to ask: what kind of AI assistants do we want to see in the world? 🤖
And what kinds of principle or commitment could steer their development in the right direction?
Here are some initial ideas to get things started, a thread:
1️⃣ AI assistants should always self-identify as AI and not pretend to be human, outside of certain creative contexts.
2️⃣ AI assistants should protect their user's privacy and embody state-of-the-art privacy preserving technology.
3️⃣ AI assistants should be designed with accessibility and ease of use in mind—including for different communities + users with different needs.
4️⃣ AI assistants should be tested rigorously prior to deployment—including for safety and alignment. The human-computer interaction level of analysis is especially important.
5️⃣ AI assistants should include measures designed to curb malicious uses, including via limits on what they will do.
Alignment with user intentions is not enough + their action space needs to be bounded in certain ways.
6️⃣ AI assistants should prioritize user well-being—based on a holistic understanding of what it means for a life to go well. Focusing on short-term preferences is not enough.
7️⃣ AI assistants should be built with cooperation in mind—both between users and assistants, and on occasions where assistants interact with each other.
8️⃣ AI assistants should include safeguards to protect vulnerable users.
9️⃣ AI assistants should be aligned in a manner that emphasises factuality—and trained to reduce the likelihood of manipulation or deception.
There are many more proposals in our recent paper: https://lnkd.in/edunwf8a
What else could be common ground? What else might you add? 🌍
What are the ethical and societal implications of advanced AI assistants? What might change in a world with more agentic AI?
Our new paper explores these questions. It’s the result of a one year research collaboration involving 50+ researchers:
https://lnkd.in/em2K-jvM
We define advanced AI assistants as “artificial agents with natural language interfaces, whose function is to plan and execute sequences of actions on behalf of a user—across one or more domains—in line with the user’s expectations.”
Advanced assistants could have a number of novel capabilities—including reasoning, memory, tool use and asynchronous action, and have a profound impact on our lives, serving as planners, thought partners, tutors, personal assistants and more….
Yet, this new phase in AI technology also heralds a new phase in AI ethics—so sociotechnical foresight is key 🔮
We need to model the futures technology makes possible—based on existing social and technical analysis trends—and understand their implications 📈
For advanced AI assistants, several things stand out:
-
With autonomous and agentic technologies alignment is key—but this goes beyond following the instructions of intentions of a single user. Greater freedom and scope of action make it essential that AI systems embody the right set of values—respecting the needs of non-users and society more widely.
-
Increasingly personal and human-like forms of AI introduce new questions around anthropomorphism, privacy, trust and appropriate relationships with AI. Safeguards are needed to support user well-being—and well-being needs to be understood in holistic terms.
-
Millions of AI assistants could be deployed at a societal level where they’ll interact with one another and with non-users. Coordination to avoid collective action problems is key. So too, is equitable access and inclusive design.
-
It’s hard to evaluate advanced AI assistants using tools that focus only on model properties and outputs. We need to understand how people interact with assistants and how they could shape society over time. New kinds of holistic evaluation are needed: https://lnkd.in/e4eq-Qi4
-
We potentially stand at the beginning of an era of potentially profound technological and societal change—which means there is a window of opportunity to act in front of us now… the paper contains over 50 recommendations.
Finally, this is quite a long paper (some have said “book”) 😅 📖 but it contains a 2-page Executive Summary and 6-page review at the end. There’s also a guide for different kinds of reader.
Please jump in – and thank you to everyone who contributed, esp. Arianna Manzini, PhD (Oxon), Geoff Keeling, James Manyika and the other authors!
