Development of artificial intelligence by AI: AI self-development?

Governance of Pandemic Response by Artificial Intelligence (Part #4)

---

[Parts: First | Prev | Next | Last | All] [Links: To-K | Refs ]

---

Those working with AI express considerable enthusiasm for the manner in which artificial intelligence might develop the capacity for improving itself. (Dom Galeon, Google's Artificial Intelligence Built an AI That Outperforms Any Made by Humans, Futurism, 12. February 2017). For Edd Gent:

Artificial intelligence (AI) is evolving -- literally. Researchers have created software that borrows concepts from Darwinian evolution, including "survival of the fittest", to build AI programs that improve generation after generation without human input. The program replicated decades of AI research in a matter of days, and its designers think that one day, it could discover new approaches to AI (Artificial intelligence is evolving all by itself,  Science, 13 April 2020).

As clarified by Pierre-Yves Oudeyer:

Autonomous lifelong development and learning is a fundamental capability of humans, differentiating them from current deep learning systems. However, other branches of artificial intelligence have designed crucial ingredients towards autonomous learning: curiosity and intrinsic motivation, social learning and natural interaction with peers, and embodiment. These mechanisms guide exploration and autonomous choice of goals, and integrating them with deep learning opens stimulating perspectives. (Autonomous development and learning in artificial intelligence and robotics: Scaling up deep learning to human--like learning, arxiv.org, 5 Dec 2017)

Further distinctions are however necessary, as noted by Jolene Creighton (The Unavoidable Problem of Self-Improvement in AI, Future of Life, 19 March 2019; The Problem of Self-Referential Reasoning in Self-Improving AI, Future of Life, 21 March 2019):

Today's AI systems may seem like intellectual powerhouses that are able to defeat their human counterparts at a wide variety of tasks. However, the intellectual capacity of today's most advanced AI agents is, in truth, narrow and limited. Take, for example, AlphaGo. Although it may be the world champion of the board game Go, this is essentially the only task that the system excels at. Of course, there's also AlphaZero. This algorithm has mastered a host of different games, from Japanese and American chess to Go. Consequently, it is far more capable and dynamic than many contemporary AI agents; however, AlphaZero doesn't have the ability to easily apply its intelligence to any problem. It can't move unfettered from one task to another the way that a human can.

The same thing can be said about all other current AI systems -- their cognitive abilities are limited and don't extend far beyond the specific task they were created for. That's why Artificial General Intelligence (AGI) is the long-term goal of many researchers. Widely regarded as the "holy grail" of AI research, AGI systems are artificially intelligent agents that have a broad range of problem-solving capabilities, allowing them to tackle challenges that weren't considered during their design phase. Unlike traditional AI systems, which focus on one specific skill, AGI systems would be able to efficiently tackle virtually any problem that they encounter, completing a wide range of tasks.

It would be naive to neglect the possibility that AI had not been developed beyond that highlighted by its use in board games. Given the success of AI against Go masters, it is approprioate to note that the complexity of Go exceeds that of chess by almost 240 orders of magnitude (J. Burmeister, The challenge of Go as a domain for AI research: a comparison between Go and chess, Intelligent Information Systems, 1995; C. Koch, How the Computer Beat the Go Master, Scientific American, 19 March 2016). For Andrew Ilachinski:

The military is on the cusp of a major technological revolution, in which warfare is conducted by unmanned and increasingly autonomous weapon systems. This exploratory study considers the state-of-the-art of artificial intelligence (AI), machine learning, and robot technologies, and their potential future military implications for autonomous (and semiautonomous) weapon systems. Although no one can predict how AI will evolve or how it will affect the development of military autonomous systems, we can anticipate many of the conceptual, technical, and operational challenges that DOD will face as it increasingly turns to AI-based technologies. We identified four key gaps facing DOD as the military evolves toward an â-"e;autonomy eraâ-"... (Artificial Intelligence & Autonomy Opportunities and Challenges, Center for Naval Analyses, October 2017)

---

[Parts: First | Prev | Next | Last | All] [Links: To-K | Refs ]