Rodney Brooks (Robust.ai Co-founder) – Steps Toward Super Intelligence and the Search for a New Path (Sep 2019)


Chapters

00:00:00 Development of Robotics, AI, and Superintelligence
00:05:17 Steps Toward Superintelligence
00:09:12 The History and Evolution of Artificial Intelligence
00:17:29 History of Behavior-Based Robotics
00:21:47 AI Performance vs. Competence: Avoiding Misconceptions and Adversarial Examples
00:24:35 Pitfalls in Measuring Artificial Intelligence Progress
00:29:24 Future AGI Standards for Elder Care Work and Service Logistics
00:31:34 Hard Problems in Artificial Intelligence
00:36:59 Understanding Common Sense and Object Recognition
00:41:27 A Historical Overview of Computation: From Logarithms to Turing Machines
00:52:37 Exploring Computation's Role in Origin of Life and Human Intelligence
01:02:37 Exploring New Frontiers of Intelligence
01:06:23 Discerning Reality from Science Fiction: Exploring the Challenges of Achieving Artificial General Intelligence
01:12:44 Media Depiction of Robots and AI

Abstract

The Future of AI: Balancing Ambition and Realism

Harnessing the Past and Present to Navigate the Future of Artificial Intelligence

In a compelling exposition blending history, current trends, and future potential, Prof. Ken Goldberg, a renowned robotics expert, unravels the complex tapestry of artificial intelligence (AI). His insights, rooted in a rebellious spirit, challenge conventional wisdom and highlight the dichotomy between AI’s theoretical aspirations and practical achievements. From Turing’s foundational work to the advent of behavior trees in video games, and addressing the “Seven Deadly Sins of AI,” Goldberg’s narrative is a profound reminder of AI’s journey, its current limitations, and the pragmatic path forward.

In 1948, Turing authored a paper titled “Intelligent Machinery,” discussing the possibility of building intelligent machines using the human brain as a guiding principle. However, Sir Charles Darwin, his lab head and grandson of Charles Darwin, prohibited the paper’s publication until 1970 due to its perceived flakiness.

Marvin Minsky’s 1961 paper, “Steps Toward Artificial Intelligence,” classified AI into five areas: search, pattern recognition, learning, planning, and induction. Notably, three of these five areas were related to search, emphasizing the significance of search in AI at the time.

The Legacy of Turing and AI’s Evolution

Alan Turing’s groundbreaking work in the mid-20th century, particularly his “Uncomputable Numbers” (1936) and the Turing Test (1950), set the stage for AI’s development. Turing envisioned machines mirroring human intelligence, guided by the principles of the human brain. His successors, including Marvin Minsky, expanded AI into distinct domains such as search, pattern recognition, and learning. The evolution continued with diverse approaches: Symbolic AI, Neural Networks, Reinforcement Learning, and Conventional Robotics, each contributing uniquely to AI’s progression.

Gray Walter’s groundbreaking work on robots in the early 1950s significantly influenced robotics. His robots, called “tortoises,” utilized vacuum tubes as controllers and exhibited complex behaviors due to non-linear dynamics. Walter published his findings in Scientific American in 1950, showcasing the learning capabilities of these tortoises.

Eric Turner further advanced Walter’s work in the mid-1980s by creating a digital version of Walter’s tortoises. This concept became known as the subsumption architecture or behavior-based approach. Turner published his findings in Neural Computation, and the behavior-based approach gained prominence in robotics, leading to its implementation in the Mars rovers and Roomba vacuum cleaners.

Damien Isler refined the behavior-based approach further in 2000, resulting in the development of behavior trees. Behavior trees are widely adopted in video game engines like Unity and Unreal for controlling AI characters. They are credited with driving the largest number of robots with long-term existence and all AI characters in video games.

The Practical and Theoretical Aspects of AI

Goldberg critiques the hyperbolic perception of AI, emphasizing the “Generalization Gap” and the overestimation of AI capabilities. He notes the importance of distinguishing between AI performance and competence, underscoring the susceptibility of AI to adversarial examples and its struggles with reasoning and context. Meanwhile, DARPA’s $2 billion initiative aims to instill common sense in AI, targeting human developmental milestones as benchmarks.

Eric Turner presented a thought-provoking comparison of four main AI approaches: symbolic, robotics, behavior-based, and neural. He categorized symbolic and robotics approaches as deliberative, while behavior-based and neural approaches have both reactive and deliberative components. Turner evaluated these approaches across five areas: composition, grounding, spatial, sentience, and ambiguity, and concluded that symbolic approaches excel in composition, while neural approaches provide symbols and are grounded in the world but not in action.

Turner also compared the performance of these four AI approaches to human children and found that AI is still far from achieving human-level cognition. AI systems are not yet capable of handling tasks that require ambiguity, spatial reasoning, or sentience, as even squirrels outperform current AI approaches in food caching tasks.

In the quest to assess AI performance versus competence, Turner emphasizes that observing an AI’s performance does not reveal its true competence. This leads to incorrect assumptions about an AI’s capabilities. Additionally, AI systems are susceptible to adversarial examplesinputs designed to fool them into making incorrect predictions. These examples can be generated using techniques such as hill climbing and evolutionary algorithms, and natural adversarial examples occur in the real world.

Robotics and AI: A Reality Check

Despite significant advances, AI and robotics face substantial challenges in areas like manipulation, real perception, and service logistics planning. Goldberg points out that modern grippers still resemble their 1970s counterparts, and AI systems grapple with complexities that a two-year-old can easily navigate. This disparity underscores the need for a more nuanced understanding of AI’s capabilities and limitations.

Rodney Brooks, a renowned roboticist, has continuously challenged conventional wisdom and explored new frontiers in AI and robotics. He co-founded iRobot, which produced the world’s best-selling robot, the Roomba, based on his research and ideas. Brooks also played a pivotal role in promoting humanoid robotics, leading to the development of robots like ASIMO and Atlas.

In 1978, Eric designed a robotic hand, called the blue arm, with a parallel jaw gripper. 40 years later, in 2018, his company’s robotic hand had the same design and features, indicating a lack of progress in manipulation technology.

Rethinking Computation and Intelligence

Goldberg expands the discourse to the philosophical and conceptual underpinnings of AI. He reflects on Turing’s original motivations, the social constructions surrounding computation, and the limitations of our current computational models. He also introduces Katarina Fridisky’s work, which addresses AI’s data set limitations, and discusses the potential of abstract thinking and universal algebra in advancing AI.

Progress and Remaining Challenges in AI: AI has made significant strides in tasks like object recognition, language understanding, and manual dexterity. However, challenges remain in replicating more advanced human capabilities, such as common sense reasoning, complex sentence comprehension, and the ability to articulate one’s own beliefs and desires.

General Intelligence and Superintelligence: The pursuit of general intelligence and superintelligence presents a significant challenge. It is unclear whether humans have the intellectual capacity to create artificial intelligence that surpasses their own.

The Role of Computation in AI Progress: The availability of massive computational power has contributed to the rapid progress in AI. However, it is uncertain whether computation alone can drive further progress in AI, or if more fundamental breakthroughs are needed.

Turing’s Definition of Computation: Alan Turing’s 1936 paper “On Computable Numbers” provided a formal definition of computation. Turing’s definition was motivated by the need to determine whether mathematics is complete, consistent, and decidable.

The Media, Perception, and Future Trajectories

The media’s sensational portrayal of AI contrasts starkly with its actual achievements, leading to funding inconsistencies and skewed public perceptions. Goldberg calls for a balanced approach, focusing on achievable tasks and addressing real-world problems pragmatically. He cautions against overhyped claims and advocates for a deeper understanding of AI’s true potential and boundaries.

Computation and its Historical Evolution: The history of computation can be traced back to efforts to compute astronomical events. John Napier’s logarithmic tables, Johann Kepler’s use of logarithms, and Charles Babbage’s analytical engine were instrumental in the development of computation.

Hilbert’s Problems and Turing’s Contribution: David Hilbert proposed three problems related to the completeness, consistency, and decidability of mathematics. Turing’s work on computability was influenced by Hilbert’s problems, specifically the question of decidability.

Computation and its Limitations: Turing’s model of computation, with its emphasis on places and containers, may not be sufficient for capturing the nuances of human intelligence.

Rodney Brooks founded the International Journal of Robotics Research (IJRR) and was involved in the development of Lucid Lisp. He co-founded iRobot, which produced the world’s best-selling robot, the Roomba, based on his research and ideas. Brooks appeared in Errol Morris’s film “Fast, Cheap, and Out of Control,” where he shared his unique perspective on research and the importance of challenging assumptions. He played a pivotal role in promoting humanoid robotics, leading to the development of robots like ASIMO and Atlas.

Media Coverage of AI: The media’s coverage of AI has oscillated between hype and skepticism over the years. Outrageous claims about AI’s capabilities have led to funding crashes in the past. The Lighthouse Report in 1973 and the neural networks hype in the 80s serve as examples of such cycles.

AI, A Journey of Balanced Aspirations

As AI continues to evolve, Goldberg’s insights serve as a beacon for navigating its complex landscape. His emphasis on understanding AI’s historical roots, current limitations, and future possibilities highlights the importance of a balanced, informed approach. By reconciling ambition with realism, the AI community can forge a path that honors Turing’s legacy while remaining grounded in practical, achievable advancements.


Notes by: Alkaid