Week 1 - article 1

Question 1

Progress Since the 1980s

Answer 1

The text maintains that computers have not fully replicated the multifaceted cognitive abilities of humans. Notably, successes in chess and computational vision are highlighted, but the author contends that these achievements often result from human programming rather than machines exhibiting true cognitive intelligence.

Question 2

Example: Chess and Advancements

Answer 2

The text mentions Deep Fritz, a computer chess system that defeated the world champion Vladimir Kramnik in 2006. This example showcases the advancement of computers in specific tasks.

Question 3

Example: Vision and Advancements:

Answer 3

Computational approaches have made substantial gains in vision. The text cites a neural network architecture that matches human performance in animal/non-animal categorization using unsupervised learning.

Question 4

Example playing bridge:

Answer 4

Example from bridge playing to illustrate a fundamental aspect of human cognition:
the capacity to incorporate external information into decision-making.

Humans can consider information from outside a specific domain to reason effectively, a skill not yet mastered by AI systems.

Question 5

Example playing bridge: limitations of AI (3)

Answer 5

Despite progress, machines lack the broad cognitive abilities exhibited by humans, such as:

open-ended thinking; and

adaptability.

Computers often lack awareness of external factors influencing a situation, limiting their ability to make contextually appropriate decisions.

Question 6

Example of playing bridge:
Reason behind limitations AI

Answer 6

Narrow Focus: One major limitation is the narrow focus of existing AI systems. They excel in specific domains but struggle with incorporating diverse sources of information.

Question 7

Example of playing bridge:
Prospects for overcoming limitations AI (2)

Answer 7

Incremental* Progress: The text expresses optimism about incremental progress in natural cognitive tasks, including language processing and memory retrieval.

Addressing Essential Shortfalls: There is a recognition of essential shortfalls in machine intelligence, suggesting a need for continued research and innovation.

Question 8

Limitations of Computational Approaches (2)

Answer 8

Lack of Open-ended Characteristics:

Machines lack the open-ended characteristics of human cognition, which involves thinking beyond predefined rules.

Dependence on Human Programmers:

Current AI systems heavily rely on human programmers, raising questions about their autonomy and independence. It suggests that human programmers remain central to AI functionality, positioning machines more as sophisticated tools than autonomous, independent entities capable of self-directed thinking.

Question 9

Reasons Behind Limitations (5)

Answer 9

The text suggests that machines still lack:
- fluidity
- adaptability
- creativity
- purposefulness
- insightfulness

seen in human cognitive abilities.

Question 10

Prospects for Overcoming Limitations of computational approaches (2)

Answer 10

Exploration of Computational Power:

While acknowledging the potential role of computational power, the text implies that it alone may not be sufficient to overcome the identified limitations.

Future Progress:

The need for progress in computational theory, algorithms, representations, and architecture is highlighted as essential for advancing machine intelligence.

Question 11

Marr’s Taxonomy*:

Answer 11

The text introduces Marr’s three-level taxonomy, emphasizing the significance of the computational level. It stresses the ongoing need to understand the information available in the environment and how it can be optimally utilized, particularly in natural task domains such as vision and speech perception.

Question 12

Challenges in Computational Theory

Answer 12

The author discusses the intricacies* of framing computational problems, especially in terms of learning from observations.

Two contrasting views on the learner’s goal are presented:

1. one emphasizing structured statistical models
2. the other favoring optimal prediction.

The text calls for further exploration of these ideas.

Question 13

Example of Natural scene statistics

Answer 13

The text mentions Geissler and Perry’s analysis of natural scene statistics, demonstrating a shift from intuitive heuristics to data-driven approaches.

Question 14

Difficulty in Framing Computational Problems

Answer 14

The text acknowledges the challenge of precisely framing computational problems, especially in understanding the relationship between stimulus variables and underlying reality.

Question 15

Prospects for Overcoming Difficulty in Framing Computational Problems (2)

Answer 15

Continuing Focus:

There is a call for continued focus on understanding what information is available in the environment and how it can be optimally used.

Exploration of Learning Goals:

The ongoing debate about the learner’s goal, whether structured statistical models or optimal prediction, indicates the need for further exploration.

Question 16

Debate about structured vs less structured approaches

Answer 16

The debate between structured approaches (Markov-Chain Monte Carlo) and less structured approaches (neural network models) is highlighted.

Question 17

Limitation in the Debate about structured vs less structured approaches (1)

Answer 17

Computational Intensity:

Both approaches mentioned are computationally intensive, raising concerns about efficiency.

Question 18

Prospects for Overcoming Limitations in the debate about structured vs less structured approaches (1)

Answer 18

Frontier in Brain Representations:

The exploration of brain representations offers a promising frontier, potentially leading to more efficient and biologically inspired algorithms.

Question 19

Cognitive Architecture

Answer 19

with mention of hybrid models combining symbolic and connectionist components.

The text envisions a future architecture that is fundamentally sub-symbolic but can produce symbolic processes as emergent outcomes.

Question 20

Future Computing Architectures

Answer 20

The discussion introduces the possibility of transitioning from the conventional von Neumann architecture to more parallel and brain-like computational systems. Neuromorphic engineering is noted as a potential avenue for achieving truly parallel computing, suggesting a transformative shift in the coming decade.

Question 21

Current Architectures: (2)

Answer 21

Hybrid Architectures:

The text discusses hybrid architectures combining symbolic and connectionist components

Question 22

Limitation of the von Neumann architecture

Answer 22

The reliance on the von Neumann architecture is noted, and the need for a paradigm shift is suggested.

Question 23

Prospects for Overcoming Limitations of the von Neumann architecture (1)

Answer 23

Emergence of Parallel Computing:

The text anticipates the emergence of truly parallel computing, potentially marking a significant shift in cognitive architecture.

Question 24

Influences on Human Cognitive Abilities: (3)

Answer 24

1. Social;
2. Cultural; and
3. Educational Influences

The text emphasizes the profound impact of nurturance, culture, and education on shaping human cognitive abilities.

Question 25

Prospects for Overcoming Limitation of influences of nurturance, culture and education on the human cognitive abilities
(2)

Answer 25

Understanding Human Experience:

Future progress in AI may require a deep understanding of human experience, including how social, cultural, and educational influences shape cognitive abilities.

Replicating Influences:

The creation of AI systems that can exploit these influences is suggested as a potential avenue for achieving human-like intelligence.

Question 26

Incremental

Answer 26

Incremental means progressing or advancing gradually in small, incremental steps rather than making significant changes or developments in one large leap.

Question 27

Intricacies

Answer 27

Intricacies are the complex and detailed components within a system or situation that require careful examination due to their interconnected and subtle nature.