Stuart Armstrong – Predicting the Timeline of Artificial Intelligence!

Stuart talks about how prediction, how we predict for AI, and how we keep failing – also how to make better predictions. Strong AI will probably be here this century.

Also see his talk at Winter Intelligence:

Stuart Armstrong’s research at the Future of Humanity Institute centres on formal decision theory, general existential risk, the risks and possibilities of Artificial Intelligence (AI), assessing expertise and predictions, and anthropic (self-locating) probability.

How We’re Predicting AI—or Failing To:
This paper will look at the various predictions that have been made about AI and pro-pose decomposition schemas for analyzing them. It will propose a variety of theoretical tools for analyzing, judging, and improving these predictions. Focusing specifically on timeline predictions (dates given by which we should expect the creation of AI), it will show that there are strong theoretical grounds to expect predictions to be quite poor in this area. Using a database of 95 AI timeline predictions, it will show that these expectations are borne out in practice: expert predictions contradict each other considerably,
and are indistinguishable from non-expert predictions and past failed predictions. Predictions that AI lie 15 to 25 years in the future are the most common, from experts and non-experts alike.

Many thanks for watching!

Consider supporting me by:
a) Subscribing to my YouTube channel:
b) Donating via Patreon: and/or
c) Sharing the media I create

Kind regards,
Adam Ford
– Science, Technology & the Future:


  1. Awakened2Truth - Disciple of Jesus the Christ says:

    AI is a crackpot pipedream because it’s based on a nature of reality that
    really doesn’t exist. A computer is a computation device, a machine,
    machines will be alive. Those who actually believe “AI” will be something
    like in science-FICTION movies, prove they never understood the reality of
    nature they live in or they are the most the subtle deceivers fooling those
    who don’t know the nature of reality we live in and making much revenue off
    of them. Even the this guy at MOTHERBOARD knows at least enough to know why
    such claims are more fantastical then an eventual reality;


  2. Stoyan Atanasov says:

    Hi, all. I felt some more things could be said on the topic. I would say
    the timeline for AI is highly correlated with available computing power
    because it enables the research and experimentation leading to AI. After
    all now AI comes into play because the computing power (and data) is
    sufficient to test 30 years old theories. All near human AI capabilities
    like speech recognition, picture description, etc. are based on that. After
    all if we could experiment modelling worlds (rich enough like ours that
    promote intelligence) and leave evolution to do it’s work – this should
    lead to AI. So I would say enough floating point operations to model
    neurons in a human sized brain and enough to model a representation of
    sufficiently rich world around it should do it. And yes the brain emulation
    put’s a cap on the timeline, but I would say that the AI we have now when
    applied to the world if all goes well will accelerate things immensely so
    “it feels more like 30 years in stead of 100”. I mean in a world 10-20
    years from now with intelligent systems handling all data, self driving
    cars and trucks and all robotic manufacturing – things look different. And
    that seems much easier to predict. I would love your comments.

Comments are closed.