diff --git a/lecture10.md b/lecture10.md
index 571add16..24aa1904 100644
--- a/lecture10.md
+++ b/lecture10.md
@@ -4,6 +4,10 @@ class: middle, center, title-slide
Lecture 10: Artificial General Intelligence
+???
+
+https://futureoflife.org/wp-content/uploads/2017/01/Jurgen-Schmidhuber.pdf?x56934
+
---
# From technological breakthroughs...
@@ -17,14 +21,14 @@ Lecture 10: Artificial General Intelligence
.grid[
.col-1-4[]
.col-1-4[]
-.col-1-4[]
-.col-1-4[]
+.col-1-4[]
+.col-1-4[]
]
.grid[
.col-1-4[]
-.col-1-4[]
+.col-1-4[]
.col-1-4[]
-.col-1-4[]
+.col-1-4[]
]
.grid[
.col-1-4[]
@@ -35,29 +39,18 @@ Lecture 10: Artificial General Intelligence
---
-# Singularity
-
-Irving John Good (1965):
-
-- Let an **ultraintelligent** machine be defined as a machine that can far surpass all the intellectual activities of any man however clever.
-- Since the design of machines is one of these intellectual activities, an ultraintelligent machine could *design even better machines*.
-- There would then unquestionably be an **'intelligence explosion'**, and the intelligence of man would be left far behind.
-- Thus the first ultraintelligent machine is the *last invention* that man need ever make, provided that the machine is docile enough to tell us how to keep it under control.
-
----
-
# Artificial narrow intelligence
- Artificial intelligence today is still very **narrow**.
- Modern AI systems often reach super-human level performance.
- - ... but only at very specific problems!
- - They do not generalize to the real world nor to arbitrary tasks.
+ - ... but only at *very specific problems*!
+ - They **do not generalize** to the real world nor to arbitrary tasks.
---
# AlphaGo
-Convenient properties of AlphaGo:
+**Convenient** properties of AlphaGo:
- *Deterministic* (no noise in the game).
- *Fully observed* (each player has complete information)
- *Discrete action space* (finite number of actions possible)
@@ -70,6 +63,8 @@ Convenient properties of AlphaGo:
---
+class: smaller
+
# Picking challenge
.center.width-100[]
@@ -78,7 +73,12 @@ Convenient properties of AlphaGo:
---
-class: middle
+class: smaller
+
+
+# Picking challenge
+
+.center.width-100[]
- *Deterministic*: OK
- *Fully observed*: **OKish**
@@ -90,41 +90,76 @@ class: middle
---
-# Where could AGI come from?
+# Artificial general intelligence
+
+XXX
+
+---
+
+# Singularity
+
+Irving John Good (1965):
+
+- Let an **ultraintelligent** machine be defined as a machine that can far surpass all the intellectual activities of any man however clever.
+- Since the design of machines is one of these intellectual activities, an ultraintelligent machine could *design even better machines*.
+- There would then unquestionably be an **'intelligence explosion'**, and the intelligence of man would be left far behind.
+- Thus the first ultraintelligent machine is the *last invention* that man need ever make, provided that the machine is docile enough to tell us how to keep it under control.
+
+---
+
+# When computers get smarter than we are
+
+.center[
+
+]
+
+---
+
+# How to build AGI?
+Several working **hypothesis**:
- *Supervised learning*: "It works, just scale up!"
- *Unsupervised learning*: "It will work, if we only scale up!"
-- *AIXI*: "Guys, I can write down optimal AI."
- *Brain simulation*: "This will work one day, right?"
+- *AIXI*: "Guys, I can write down optimal AI."
- *Artificial life*: "Let just do what Nature did."
+Or maybe something else?
+
---
-# Supervised learning
+class: middle, center
----
+# AIXI
-# Unsupervised learning
+---
---
-# AIXI
+class: middle, center
+
+# Artificial life
---
-# Brain simulation
+???
+
+https://en.wikipedia.org/wiki/Artificial_life
---
-# Artificial Life
+class: middle, center
+
+# Conclusions
---
-# Something else?
+# Don't fear intelligent machines
----
+.center[
+
+]
-# Further readings
+---
-- Add ref to Karpathy's slides
-- Bostrom book
+# Going further