Science currently has three frontiers: the very big, the very small, and the very complex.
Science currently has three frontiers: the very big, the very small, and the very complex.
My research currently focuses on disruptive technologies or major paradigm shifts brought about by new technology, such as von Neumann (programmable) computers, computer trading and superintelligence.
Disruption is taken to be an interdisciplinary area of research combining computing, philosophy, artificial intelligence, engineering, social sciences, and future studies, as inspired by the work of John von Neumann, Alan Turing, and Alvin Toffler. In particular I am interested in that transformative effects of increasingly intelligent AI.
If we knew what it was we were doing, it would not be called research, would it?
I read computer science in Tel Aviv University, writing my dissertations on software design and machine learning, but as student in the Interdisciplinary Program I was also given the opportunity to maintained active interest in other disciplines, including brain sciences, cognitive psychology, ontology, and semantics.
It has long been my personal view that the separation of practical and theoretical work is artificial and injurious. Much of the practical work done in computing, both in software and in hardware design, is unsound and clumsy because the people who do it have not any clear understanding of the fundamental design principles of their work. Most of the abstract mathematical and theoretical work is sterile because it has no point of contact with real computing.
Theory and practice of software design, architecture, modelling, visualisation, and evolution;
Formal methods: quantifying and formalising design traits; theory and practice of formal verification
Popular software modelling notations visualize implementation minutiae but fail to scale, to capture design abstractions, and to deliver effective tool support.
The first version of the modelling language was named LePUS by its inventor, Dr Eden. The LePUS notation only modelled design patterns. LePUS3, the language of Codecharts, was developed by Dr Eden with his research students Dr Notis Gasparis and Dr Jonathan Nicholson.
The formal design description language of Codecharts was tailored to—
LePUS3, the language of Codecharts, also satisfies the following requirements:
Design patterns are not programs but blueprints of an implementation, templates which describe the correlations and collaborations between generic participants in a program. Modelling them therefore requires an appropriately generic language.
We define Codecharts, a visual language for modelling the static (decidable) aspects of design patterns, and demonstrate how to verify whether a given Java implementation satisfies the specification of the pattern.
Visualising object-oriented design (OOD) poses several challenges, for example:
This line of work demonstrates the answers given to these questions.
Everyone knows that software needs to be flexible. We also know that design patterns and architectural styles improve the ability of programs to evolve. But how can flexibility can be measured?
Our work suggests an objective measure for flexibility, called “evolution complexity”. We use the measure to compare the flexibility of design alternatives, thereby providing a robust definition of flexibility.
For example, we know that adding a new ‘Visitor’ class to a program that implements the Visitor pattern is easy, but now we can prove that the change has low evolution complexity.
We also show that flexibility is relative to a particular class of changes. For example, while it is easy to add a ‘Visitor’, it’s hard to add an operation. We demonstrate that this change is hard by showing that adding an operation to an implementation of the Visitor pattern has high evolution complexity.
What is the difference between architecture and design? This distinction is important since the choice of architecture has strategic and far-reaching implications over the implementation and therefore must be made early in the development process, whereas design decisions have localized effect and must be deferred to a latter stage in the process.
We distinguish between three abstraction strata in software design statements:
We define criteria of distinction in mathematical logic and present the Intension/Locality Hypothesis:
This distinction allows us to prove the Architectural Mismatch Theorem, which formalizes the notion of architectural mismatch (Garlan, Allen & Ockerbloom 1995).
Inconsistency between design and implem entation is a major problem for most programmers. Conflicts are common because there are no effective tools for detecting them. Original design documents are a hassle to update (and nobody bothers with that). Absence of current and accurate design documentation results in software that is unpredictable and poorly understood.
Round-trip engineering tools support an iterative process of detecting conflicts and resolving them by changing either the design or the implementation.
A Toolkit we developed, the Two-Tier Programming Toolkit, supports a round-trip engineering of native Java programs without interfering with any existing practices, tools or development environments, thereby posing a minimal barrier on adoption. The Toolkit includes a user-guided software visualization and design recovery tool, which generates Codecharts from source code. A ‘round-trip’ process is possible because Codecharts visualizing source code can be edited to reflect the intended design, and the Verifier can detect conflicts between the intended and as-implemented design.
Our work has demonstrated how the Toolkit effectively helps to close the gap between design and implementation, recreate design documentation, and maintaining consistency between design and implementation.
Suppose that [special and general relatively, automobiles, airplanes, quantum mechanics, large rockets and space travel, fission power, fusion bombs, lasers, and large digital computers] had all been presented to mankind in a single year! This is the magnitude of “future shock” that we can expect from our AI expanded scientific community.
Machine learning; disruptive impact; superintelligence
Humans learn to recognize birds and chairs without explicit definitions. ‘Natural concepts’ are noisy categories which have no crisp boundaries, yet humans learn identify them quickly and easily. Can artificial intelligence perform the same task?
Creating an agent capable of identifying natural concepts effectively can become possible by building upon lessons learned from the human ability to perform this task.
Our line of investigation borrows heavily from the interdisciplinary literature on concept formation and models of representation, in particular cognitive psychology and the philosophy of mind corpora.
Consider intelligence as the general ability to solve problems and increase utility that animals and machines have, not only present humans. It would be naive to assume Homo Sapiens Sapiens has the highest intelligence possible or anything remotely close to it. It is much more likely that much more intelligent beings could exist.
An agent whose intelligence exceeds that of the brightest person is said to be superintelligent. It is hypothesized that intelligence amplification (eg cognitive enhancement) may lead to superintelligent posthumans. Others have argued that self-improving artificial intelligence can also become superintelligent, possibly in a process of intelligence explosion.
Given the advances both in intelligence amplification (IA) and artificial intelligence (AI), the question many ask is which one will arrive first?
It was not made for those who sell oil or sardines
Ontology, epistemology, and paradigms of computer science
Computer science is a broad discipline. Theoretical computer science is very mathematical, whereas software development has all the hallmarks of an engineering activity. Further facets merge with psychology and cognitive science and applications are found throughout the physical sciences and engineering.
The philosophy of computer science is concerned with conceptual issues that arise by reflection upon the nature of computer science. The topics covered reflect the latter’s wide range: What kind of discipline is computer science? What are the roles of design and experimentation? What role is played by mathematics? What do correctness proofs establish? These are only a few of the questions that form the body of the philosophy of computer science.
Our entry on the philosophy of computer science in Stanford Encyclopedia of Philosophy attempts to cast these questions in the context of existing philosophical issues and shed light on the answers that the various branches of computing may offer.
Is computer science a branch of mathematics, an engineering discipline, or a natural science? Should knowledge about the behaviour of programs proceed deductively or empirically? Are computer programs on a par with mathematical objects, with mere data, or with mental processes?
Public debates on the answers to these questions reveal three very different perspectives over methodological, ontological, and epistemological issues:
We’re all so busy being practical that we don’t have time to be intelligent