Workshop on Architectures for Generality & Autonomy (AGA 2017)

Why can't (current) machines reason like Euclid
or even human toddlers?
(And many other intelligent animals)

This is part of the Meta-Morphogenesis (M-M) project.

Recorded video talk presented at the event: ai-cogsci-bio-sloman.webm
(56MB -- approximately 42min)
The web page, covers a lot more material than the recording,
and was revised and extended after the recording had been made. (and PDF)
Later, links referring to parts of the video (listed below) were inserted on the web page.

Aaron Sloman
School of Computer Science, University of Birmingham

Last updated: 13 Sep 2017; 20 Sep 2017

(Added: 13 Sep 2017)
The recording was unscripted, unrehearsed, badly edited, and is generally very amateurish, but at least it has no stupid background music. I may later replace this with a more carefully planned version that avoids all the scrolling down the web site to find the next point to be explained.
Extracts linked separately from the web page
  1. Video of infant exploring 3D topology (4min38s) (With comments)
    (with pencil and hole in a sheet of paper); The video conjectures that rich internal languages are used by pre-verbal toddlers e.g. to encode percepts, intentions, conjectures, plans, explanations etc.

  2. Explaining why Euclid's axioms were discoveries, not arbitrary starting points(19min32s)
    And why discoveries in logical systems are much easier for computers than discoveries in geometry and topology.

  3. Brief introduction to the Turing-inspired Meta-Morphogenesis (M-M) Project(4mins 31s)
    For more details on the M-M project see the project's web site:
    For the theory of construction-kits produced and used by evolution see (long),
    An earlier version of this paper, frozen in 2016, is an invited chapter in The Incomputable: Journeys Beyond the Turing Barrier,
    2017, Ed. S. B. Cooper and Mariya I. Soskova,, Springer,

  4. A brief introduction to the Chappell/Sloman theory of layers of epigenesis(4min25s)
    Contrasted with the Waddington's 'Epigenetic landscape' among others.
    Our ideas overlap with those of Annette Karmiloff-Smith in Beyond Modularity (1992)

1. In order to meet the time constraints I had to post-edit the recording quite viciously -- removing pauses and some of the less important explanatory points. This led to some anomalies where the transitions are not very natural.

2. While doing the recording I switched between virtual desktops, with the text file displayed in a web browser. Unfortunately I forgot to move the top of the browser window out of the recording area, thereby wasting precious screen space.

3. After the event I broke the video into shorter extracts, listed above, now also referenced from the main site.

At some later date I may produce a new, more carefully planned, recording, or collection of recordings on this topic, not constrained by the 40 minute workshop presentation limit.

The full video includes these three files concatenated:

The video presentation includes extracts from a BBC youtube video on weaver birds, presented by David Attenborough. The full video is here:

I may later expand the recording with extracts from this video on Nicaraguan deaf children:
which I think supports the claims in the lecture about epigenesis and prior evolution of internal languages.
See also:
Children create new sign language
By Julianna Kettlewell
BBC News Online science staff 2004.

This is

Installed: 12 Aug 2017
Last updated:
6 Sep 2017 (Further reorganisation -- parts linked from talk web page.)
21 Aug 2017 (reorganised);
13 Aug 2017 (corrected link to long file); 14 Aug 2017; 18 Aug 2017 (new toddler insert);

Maintained by Aaron Sloman
School of Computer Science
The University of Birmingham