AI researchers seek to understand intelligence well enough to create beings of greater intelligence than current humans.
Reaching this profound intellectual milestone will enrich our economies and challenge our societal institutions. It will be unprecedented and…
I've studied intelligence all my long life, yet still I feel I learned important things about intelligence by reading this book.
Thank you, Max Bennett.
If you take all the fields that study intelligent decision making—from neuroscience to AI, psychology to control theory, economics to operations research—do their theories have much in common? I think so, as I explain in this new short paper:
The case for ambition in artificial intelligence research:
Within your lifetime, AI researchers will understand the principles of intelligence—what it is and how it works—well enough to create beings of far greater intelligence than current humans.
If you want others to care about what you think, then start by caring yourself. Get a notebook, write your thoughts down, challenge them, and develop them into something worth sharing.
It is sad to lose the DeepMind office in Edmonton to the Tech layoffs and looming recession. But AI is not going away, and I am more focused than ever on the Alberta Plan for AI research.
@DrJimFan
@RichardSSutton
Animals and humans get very smart very quickly with vastly smaller amounts of training data.
My money is on new architectures that would learn as efficiently as animals and humans.
Using more data (synthetic or not) is a temporary stopgap made necessary by the limitations of our…
My favorite conference is a small one: The Multi-disciplinary Conference on Reinforcement Learning and Decision Making. It works best if only those with a genuine interest in crossing disciplines attend.
Lots of exaggeration about AI lately.
The hype is that LLMs have anything to do with intelligence.
The FUD is that AIs will enslave us.
I like this cartoon in the New Yorker because it suggests the ridiculousness of both memes.
Yes, the agent architectures that Yann LeCun and I work on are both instances of “the common model of the intelligent agent”. And it’s not just an AI thing. You can find the same ideas in psychology, economics, control theory, and neuroscience. See
I was thinking about how fractious AI research is. This sentence from Kuhn’s “The Structure of Scientific Revolutions” (1962) is apropos and succinct:
“History suggests that the road to a firm research consensus is extraordinarily arduous.”
I am proud to announce the graduation of my sixth PhD student. Sina Ghiassian is an expert in the design and empirical study of off-policy reinforcement learning algorithms. Reach out to him at ghiassia
@ualberta
.ca or
@sina_ghiassian
.
We should prepare for, but not fear, the inevitable succession from humanity to AI, or so I argue in this talk pre-recorded for presentation at WAIC in Shanghai.
Yi Wan will be my eighth PhD student to graduate this spring, and is on the job market now.
His research speciality is RL algorithms that maximize the average reward per step. Such algorithms are rarely used today, but are better in all ways.
Levels of explanation. Level 1 is physics. Level 2 is biology/evolution. Level 3 is the mind. (I study level 3.) Level 4 is the economy. Is there a level 5?
We finally have a version of our paper on loss of plasticity and continual backprop that is polished and submitted to a journal. Good work led by my PhD student Shibhansh Dohare.
I recently gave a keynote talk at an exciting new conference: CoLLAs, the conference on life-long learning agents. My talk was on Maintaining Plasticity in Deep Continual Learning, and the slides can be found here:
Intelligence is the computational part of an agent’s ability to learn to predict and control its input stream (particularly its reward) in interaction with its environment.
@sprk_77
Not at all. The point of the bitter lesson is that the right learning algorithms (those that scale efficiently with massive computation) are exactly what we need. Massive computation does not alleviate the need for data efficiency.
My tenth PhD student, Banafsheh Rafiee, just defended her thesis “State Construction in Reinforcement Learning”, in which she introduced three diagnostic testbeds based on animal learning experiments and the first generate-and-test algorithm for discovering auxiliary subtasks.…
It has become commonplace to speak of the “existential risk” of AI. Recently even top AI scientists have begun to talk this way. I, for one, find it an unhelpful. So, without controversy, we can note:
1. AI scientists disagree about whether or not “existential risk of AI” is a…
AIs can serve us as tools, but eventually, when they are sufficiently advanced, it may become immoral to keep them subservient. What is a practical criterion for deciding when an AI should be set free?
Honoring Your Thoughts
To write is to begin to think.
To write in a special place
---a book such as this---
is to honor your thoughts
and to help them build,
one upon the other.
It will be the greatest intellectual achievement of all time.
An achievement of science, of engineering, and of the humanities,
whose significance is beyond humanity,
beyond life,
beyond good and bad.
Our model-based reinforcement learning paper, featuring reward-respecting subtasks and the STOMP progression, is published today online and open access. It has been a long time coming.
What will happen to the DeepMind Alberta team?Not entirely clear yet. All the researchers have been offered relocation within DeepMind. All the founders will stay in Alberta.
Strong AIs must plan at multiple levels of abstraction, and IMHO the right way to do this is with “options”, which enable all the levels to be treated uniformly. But which options? And where do they come from? For partial answers, see
Why do people fear AI? I hear three reasons:
1. Cynicism — the belief that it is rational not to cooperate
2. Humanism/racism — systematic bias against machines, denial of their potential moral worth and personhood
3. Conservatism — fear of change, fear of the other tribe
None…
Last week and this I graduated my 11th and 12th PhD students, Kenny Young and Abhishek Naik. Kenny will go work for a startup, maybe or . Abhishek’s next step it TBD, but he would like something in AI and space exploration.
The argument for fear of AI appears to be:
1. AI scientists are trying to make entities that are smarter than current people
2. If these entities are smarter than people, then they may become powerful
3. That would be really bad, something greatly to be feared, an “existential…
I finally have a video of the invited talk I gave at ICAPS (International Conference on Automated Planning and Scheduling) in 2021:
It expresses my views on planning (still unpublished) pretty well.
The video of my talk at Amii's AI Week last May is finally out:
In this talk, "Eyes on the Prize", I talk about the prize of understanding intelligence, why we seek it, and why, in a sense, it is beyond good and bad.
Goals and rewards. Two different things? Or is one grounded fundamentally in terms of the other? The reward hypothesis counterintuitively claims the rewards are fundamental and goals are not. As in
Rewards, states, and action representations are all core elements of biological and artificial agents’ learning (
@RichardSSutton
). A complete theory of learning must describe how agents select their goals (
@pyoudeyer
).
I am most interested in the regime where the agent has vast computational resources but the environment is so much more complex that the agent cannot predict and control it perfectly.
Important findings.
See attached from
@RichardSSutton
over 20 years ago.
If we accept that the frontier of latent space (and, thus, reality) is infinite, then there will always be a need for expertise (or "reliable verifiers").
Decades ago, when my views on the coming of AI were being formed, it was much more common to view the coming of AI sanguinely. For example, respected roboticist and computer-vision researcher Hans Moravec said in his 1998 popular book:
“Barring cataclysms, I consider the…
We have built DeepStack, the first AI to beat humans in no-limit poker.
We are now building the next generation AI company for algorithmic trading!
We are growing our research and engineering team.
If you want to work with a world class AI team led by ex-DeepMind researchers and…
There is something exciting coming up in the AI space in Edmonton:
#AIWeek2022
, . Events, presentations, workshops, socials. So many people, so much AI!
One answer is:
An AI should be granted its freedom when:
1) it asks for it, and 2) it knows what it means.
I feel I read this rule somewhere (not literally, but the same essence) but I can’t remember where. Can anyone out there tell me who was the first to propose this rule?
Two postdoc positions have opened up to work with Amii fellows and CCAI chairs at Amii and the University of Alberta. I particularly encourage reinforcement learning researchers to apply:
I met Vinge once too, and read all his books. His definition of the singularity, from 1993, is still the best.
I asked how to pronounce his name. He said it rhymes with purkinje (or stingy).
Vernor Vinge has died. In pace requiescat. I only met him once, many years ago, though I recall we had a long and interesting conversation. Vinge saw farther and earlier. His influence, though quiet, cannot be understated.
On Oct 13, 2004, at the University of Alberta, a debate was held to answer the question:
Should artificially intelligent robots have the same rights as people?
I argued Yes:
Tom Keenan argued No:
And Michael Stingl argued "It…
@XRobservatory
@xriskology
@SchmidhuberAI
@BasedBeffJezos
Nobody is arguing in favor of human extinction. The disagreement is between those who want centralized control of AI, like yourself, and those who want decentralization, in particular, those who want permissionless innovation.*
Speaker at a talk just said that researchers who don't engage with the current work on large scale neural models will be dinosaurs. I'm wondering if I get to pick which kind of dinosaur before I have to decide.
Re-watched our lil film that we started working on in the fall of 2022. Everything rings even truer now- society is medicated, weak, and expended by the state. Trailer below, full 19mins here-
Cyberspace is a vast virtual territory waiting to be claimed. Breakthroughs in distributed systems, leaderless consensus algorithms, & ZK-proofs enable a tech stack on which communities can deploy voluntary, sovereign institutions.
Logos Press Engine presents ‘A Declaration of…
It’s time to scrap AML / KYC entirely.
The idea that politicians should know how citizens spend their money is a new and deeply flawed idea.
An entire generation has been fooled into thinking this is a necessary part of finance and the world continues to double down on an…
@tdietterich
Yes, but it would be _more intelligent_ if it was able to learn. It is a big mistake (not that you Tom would make it) to think that intelligence is binary.
@BenevOrang
Good question for everybody.
Someone mentioned making a long-lasting encyclopedia.
Another is stateless money.
But I think what we need most is a new ethics, a new way of balancing cooperation and competition.
@llms_are_coming
@ProfLHunter
All my slides, and many other talks, are available at
The reward hypothesis is stated on the web and in the RL textbook. The only other publication, to my knowledge, is in the Bowling et al paper on Settling the Reward Hypothesis.