1
2	Dartmouth 1956: Challenges Machine methods of forming abstractions from sensory and other data Carrying out activities which may best be described as self-improvement Manipulating words according to rules of reasoning and rules of conjecture Developing a theory of the complexity for various aspects of intelligence
3	"Hard problems to be solved..." Hard problems to be solved by drawing methods from: Logical reasoning Statistics Cybernetics Information theory Automata theory Pattern recognition
4	"Hard problems to be solved..." Hard problems to be solved by drawing methods from: Logical reasoning Statistics Cybernetics Information theory Automata theory Pattern recognition
5	Perspective: Intelligence amidst Inescapable Incompleteness Limited agents immersed in large universes Limited representations Limited time and memory
6	Perspective: Intelligence amidst Inescapable Incompleteness
7	…and the Bottleneck Economics of computation and memory Flexible computational strategies Phase transitions and hardness Bounded optimality
8	Economics of Flexible Computation
9	Economics of Flexible Computation
10	Economics of Flexible Computation
11	Expected Value of Computation
12	In Pursuit of Bounded Optimality
13	Bounded Optimality not the Goal à Guide research, build insights Comparative analysis Contributions of innovations Ideal situation-action reflexes Partition of resources to different phases of reasoning Ideal continual computation
14	e.g., Continual Computation Policies for continuing to use all available resources all of the time
15	Stream of Problem Instances
16	Trading Off Present for Future
17
18	Directions and Challenges Integrative intelligence Situated long-lived autonomous systems Learning amidst growing data resources Artificial intelligence in supportive & collaborative roles Learning from neurobiology Our evolving relationship with machines
19	Challenge: Integrative Intelligence Sensing & symbols Goals & preferences Learning Real-time reasoning, metareasoning, and compilation Action execution
20	Coordinating an Elegant Dance
21
22
23	"Flexible adaptation to varying tasks" Flexible adaptation to varying tasks, situations, environments
24
25
26	Automated Framing & Execution of Local, High-Precision Decision Problems
27	Automated Framing & Execution of Local Decision Problems
28	Multi- and Varying Timescales
29	"What initial endowment is good..." What initial endowment is good enough for survival but that can be most efficiently adapted with experience?
30	"What initial endowment is good..." What initial endowment is good enough for survival but that can be most efficiently adapted with experience?
31	Opportunity: Machine Learning in a Data-Rich World Large quantities of data available through new sensors and processes + Advances in tractable machine learning _____________________________________ à Opportunities for insights …..and new applications and services
32	Advances in Learning and Adaptation
33	Identification of Hidden Variables
34	Opportunity: Building Intention Machines Example: Web search
35	Opportunity: Building Intention Machines Beyond web search
36
37	Learning about Reasoning Opportunity for making reasoning more efficient Identify dead ends Understand and guide heuristic procedures
38	Learning about Reasoning Machine learning about ideal restarts of a theorem prover (see Horvitz, et al., Kautz, et al., Ruan, et al.) Grappling with long tail of satisfiability search
39	Learning to Work with the Web
40	Learning to Work with the Web
41	Learning to Guide Question-Answering System
42	Learning to Guide Question-Answering System
43	Learning to Guide Question-Answering System
44
45	Mixed-Initiative Interaction
46	Mixed-Initiative Interaction
47	"20^th Century cognitive psychology" 20^th Century cognitive psychology: Characterizable limitations & bottlenecks
48
49
50	Application: Triaging Communications
51	Vigilant Systems Ongoing support for safety Smart surveillance Transportation Healthcare … Reasoning about things that surprise people
52	Reasoning about Surprises Systems designs to predict when a user would be surprised by events
53
54	Neurobiology and Artificial Intelligence: From Existence Proofs…to Insights
55	Neurobiology and Artificial Intelligence: From Existence Proofs…to Insights
56