|
An Attempt to Generalize AI - Part 16: Speculation on Autism
By Paul Almond |
24 July 2010 |
|
This is the sixteenth in a series of articles attempting an overview of how minds may
work and how similar systems could be implemented in computers. The cognitive model and
approach to artificial intelligence uses a probabilistic hierarchy based on patterns. A
pattern has a specification describing a set, or population, of pattern instances,
distributed throughout a hierarchy containing the pattern instances of all the patterns.
The hierarchy needs processing to provide relevance, with useful pattern instances being
featured in the hierarchy, and used as a basis for exploratory extension of the
hierarchy, while less useful pattern instances are removed, so that the hierarchy
"grows" into high-relevancy regions, and this is achieved by a back-propagation
relevance measurement process and a process to assign relevance values to pattern
instances. Restrictive and repetitive behavior in autistic spectrum conditions could be
explained in terms of a problem with the processing to provide relevance in the
hierarchy. If the rate of turnover of pattern instances - the rate at which
low-relevance pattern instances are removed and new ones are added - is reduced, the
ability of the hierarchy to adapt its structure in response to occurrence of external
inputs/outputs will be affected; however, the availability of "old" pattern instances
in the hierarchy could also make otherwise inaccessible regions of the hierarchy
accessible. Other characteristics of autism might be explained by other causes, though
there could be an overall, ultimate cause, but they might also be explained in terms of
the general limitation in modeling described here. The cognitive model treats the
"self" as just part of the model, so any limitation in modeling ability which affects
representation of other people might affect representation of the "self".
|
| Read this article in: Microsoft Word Format PDF Format |
|
|
On God, Evil and Simulation
By Paul Almond |
20 July 2010 |
|
The occurrence of evil is often used as an argument against the existence of a
benevolent god. Against this, theists typically make excuses for the occurrence of evil.
One of these is that humans need to be given freedom to act. It is argued that this is
flawed, because God should be able to provide each of us with a personal world that
allows freedom to act, while insulating each of us from everyone else’s actions. Any
being who cannot do this is not worthy of being called “God”, and rather than believe
in a god that could do this, but chooses not to, we should doubt that a benevolent god,
as typically believed in by theists, exists.
|
| Read this article in: Microsoft Word Format PDF Format |
|
|
rationalskepticism.org Podcast - Episode 0: The Possibility of Extraterrestial Life
By hackenslash, jerome, natselrox, Paul Almond |
14 July 2010 |
|
Along with hackenslash, jerome and natselrox, I was invited to take part in the first podcast made for the http://wwww.rationalskepticism.org
forum. The subject being discussed was the possibility of extraterrestial life. The recording was made on 3 July 2010, and after editing by hackenslash, this is the result. More podcasts like this are intended on a regular basis.
In the end, it was decided that this was not up to the standards needed to be an "official" podcast: More organization will be needed for future
podcasts. However, it was decided that, rather than waste it, it would be placed online anyway as "Episode 0".
|
| Listen to the podcast at: http://www.rationalskepticism.org/podcast/ |
|
|
An Attempt to Generalize AI - Part 15: A Complete Description
By Paul Almond |
10 July 2010 |
|
This is the fifteenth in a series of articles attempting an overview of how minds may
work and how similar systems could be implemented in computers. For readers wanting an
understanding of what is being suggested, this article is an ideal starting point, as it
gives a complete description of the cognitive model as it stands at the time of writing.
This article brings everything together so that it can be understood by reading just
this article, instead of having to read fourteen articles and having to read about ideas
that are changed later, or read through a long discussion in which the concepts are
developed. The cognitive model and approach to artificial intelligence uses a
probabilistic hierarchy based on patterns. A pattern has a specification describing a
set, or population, of pattern instances, distributed throughout a hierarchy containing
the pattern instances of all the patterns. Each pattern's set of pattern instances is
used to obtain statistical information for probabilistic predictions. Each pattern's
population of pattern instances is to be described in a very general way, to provide a
very general ontology. The hierarchical model is actually used to plan the system's
actions, and this implies that what people regard as the "self" is really an object in
the hierarchical model. The hierarchy needs to be relevant, with pattern instances that
are useful being featured in the hierarchy, and used as a basis for exploratory
extension of the hierarchy, while less useful pattern instances are removed, so that
the hierarchy "grows" into high-relevancy regions, and this is achieved by a
back-propagation relevance measurement process which assigns relevance values to
pattern instances. Relevance can also be provided by reflexive outputs: special outputs
that are made in the same way as normal outputs, but which alter the hierarchical model.
|
| Read this article in: Microsoft Word Format PDF Format |
|
|
An Attempt to Generalize AI - Part 14: Mind Control Speculation
By Paul Almond |
24 June 2010 |
|
This is the fourteenth in a series of articles attempting an overview of how minds may
work and how similar systems could be implemented in computers. Previous articles
described a probabilistic hierarchy based on patterns. A pattern has a specification
describing a set, or population, of pattern instances, distributed throughout a
hierarchy containing the pattern instances of all the patterns. Each pattern's set of
pattern instances is used to obtain statistical information for probabilistic
predictions. Each pattern's population of pattern instances is to be described in a
very general way, to provide a very general ontology. The hierarchy's structure is
based on the history of previous inputs/outputs, so "faking" outputs at the point of
entry into the hierarchy as bottom-level pattern instances could be used to control the
representation of the world in the hierarchy. The behavior of a human brain may be
altered by such "faking" of the outputs that are stored in bottom-level pattern
instances. The "self" has no qualitatively special status, but is merely an object in
the hierarchy's representation of the world, and would therefore be altered in such a
process. The extent to which such a process would affect a human would depend on the
extent to which the human brain is like the hierarchical system described in this series
of articles. A non-invasive version of the process, involving manipulation of the
environment, is also discussed, as is combining this with the process of manipulating
outputs. The more extreme idea of copying an entire mind, or some other computer system,
into an artificial hierarchy or a human brain is discussed.
|
| Read this article in: Microsoft Word Format PDF Format |
|
|
An Attempt to Generalize AI - Part 13: Reflexive Outputs
By Paul Almond |
17 June 2010 |
|
This is the thirteenth in a series of articles attempting an overview of how minds may
work and how similar systems could be implemented in computers. Previous articles
described a probabilistic hierarchy based on patterns. A pattern has a specification
describing a set, or population, of pattern instances, distributed throughout a
hierarchy containing the pattern instances of all the patterns. Each pattern's set of
pattern instances is used to obtain statistical information for probabilistic
predictions. Each pattern's population of pattern instances is to be described in a
very general way, to provide a very general ontology. An exploratory relevance process
has been described, to restrict the hierarchy to representing relevant features of the
world. This article introduces a further way of providing relevance: reflexive outputs.
Reflexive outputs are outputs that, instead of controlling devices in the
"outside world", control the modeling system itself - making adjustments to the
hierarchy to direct its modeling. An AI system could learn to make reflexive outputs in
the same way that it learns to make other outputs, because the distinction between the
system itself and the "outside world" is really artificial: Outputs affect the world and,
if a system is able to learn how to make appropriate ones, it should not matter if part
of the world being affected happens to be the system itself. Reflexive outputs are not
being proposed as an alternative to the exploratory relevance process proposed
previously, which is still the main process for providing relevance, but are intended
to augment it. A philosophical consideration of reflexive outputs in relation to the
"self" is also given.
|
| Read this article in: Microsoft Word Format PDF Format |
|
|
An Attempt to Generalize AI - Part 12: Pattern Relevance
By Paul Almond |
15 June 2010 |
|
This is the twelfth in a series of articles attempting an overview of how minds may work
and how similar systems could be implemented in computers. Previous articles described a
probabilistic hierarchy based on patterns. A pattern has a specification describing a
set, or population, of pattern instances, distributed throughout a hierarchy containing
the pattern instances of all the patterns. Each pattern’s set of pattern instances is
used to obtain statistical information for probabilistic predictions. Each pattern’s
population of pattern instances is to be described in a very general way, to provide a
very general ontology. An exploratory relevance process has been previously described,
the purpose of which is to ensure that the pattern instances included in the hierarchy
are those which are relevant to the predictions of future evaluation function score
values required in the system’s action selection process. This is done by assigning
relevance values to pattern instances in a back-propagation process. The issue of
selecting the patterns themselves that will be involved in this has had little
discussion, and this article remedies this by discussing the inclusion of pattern
selection into an exploratory relevance process. A general view of the exploratory
relevance process is also given: Although the exploratory relevance process involves
trial and error, modifications to the hierarchy build on what already exists, and the
hierarchy also directs its own future development.
|
| Read this article in: Microsoft Word Format PDF Format |
|
|
An Attempt to Generalize AI - Part 11: Explaining Dreaming
By Paul Almond |
30 May 2010 |
|
This is the eleventh in a series of articles attempting an overview of how minds may
work and how similar systems could be implemented in computers. Previous articles
described a probabilistic hierarchy based on patterns. A pattern has a specification
describing a set, or population, of pattern instances, distributed throughout a
hierarchy containing the pattern instances of all the patterns. Each pattern's set of
pattern instances is used to obtain statistical information for probabilistic
predictions. Each pattern's population of pattern instances is to be described in a very
general way, to provide a very general ontology. If the proposed view of the mind is
generally correct, phenomena in human brains should make sense within this context. This
article discusses dreaming within the context of the hierarchy of pattern instances. It
is suggested that dreaming occurs in a partially functioning hierarchy. Functioning
pattern instances that would normally receive pattern inputs from non-functional ones
instead receive noise. The wide scope of the hierarchy, in terms of explaining the "self"
as part of the model, similarly gives any explanation of dreaming involving the
hierarchy being compromised a wide scope, economically explaining the general effects of
dreaming on perception of reality, memory and behavior. If this is correct, dreaming is
not profoundly different from waking experience, but only differs by a matter of degree.
Looking at this another way, waking experience could be considered to be a particularly
detailed dream that is strongly influenced by reality, with the "self" being "dreamed"
by the hierarchical model with everything else.
|
| Read this article in: Microsoft Word Format PDF Format |
|
|
An Attempt to Generalize AI - Part 10: Pattern Instance Construction Alternatives
By Paul Almond |
21 May 2010 |
|
This is the tenth in a series of articles attempting an overview of how minds may work
and how similar systems could be implemented in computers. Previous articles described a
probabilistic hierarchy based on patterns. A pattern has a specification describing a
set, or population, of pattern instances, distributed throughout a hierarchy containing
the pattern instances of all the patterns. Each pattern’s set of pattern instances is
used to obtain statistical information for probabilistic predictions. Each pattern’s
population of pattern instances is to be described in a very general way, to provide a
very general ontology. An exploratory relevance process has been described, which
achieves the relevance of the hierarchy by removing low-relevance pattern instances
while the hierarchy “grows”, so that it will tend to retreat from low-relevance regions
and grow into high-relevance ones, and forgetting is part of this process. The way in
which patterns relate to pattern instances has been described as being a “constructive”
one, in which each pattern is considered as a machine that “builds” its pattern
instances. It is not essential for such an approach to be used, however. The only
requirements are that pattern instances belong to patterns and that the relationship
between a pattern and its pattern instances can be expressed very generally. This
article discusses possible alternatives for the relationship between patterns and
pattern instances. These are pattern instances being made randomly and selected by
patterns, pattern instances being made by some exploratory process and selected by
patterns and the connection of pattern instances being influenced by patterns.
|
| Read this article in: Microsoft Word Format PDF Format |
|
|
An Attempt to Generalize AI - Part 9: Improving the Exploratory Relevance Process
By Paul Almond |
13 May 2010 |
|
This is the ninth in a series of articles attempting an overview of how minds may work
and how similar systems could be implemented in computers. Previous articles described
a probabilistic hierarchy based on patterns. A pattern has a specification describing a
set, or population, of pattern instances, distributed throughout a hierarchy containing
the pattern instances of all the patterns. Each pattern's set of pattern instances is
used to obtain statistical information for probabilistic predictions. Each pattern's
population of pattern instances is to be specified in a very general way, to provide a
very general ontology. A basic, exploratory relevance process has been described, which
achieves relevance of the hierarchy by removing low-relevance pattern instances while
the hierarchy "grows", so that it will tend to retreat from low-relevance regions and
grow into high-relevance ones. The basic, exploratory relevance process uses a relevance
measurement process, which back-propagates relevance through the hierarchy. The basic,
exploratory relevance process also performs forgetting - the removal of obsolete pattern
instances from the hierarchy. This is done in the relevance measurement process, by
taking account of obsolescence. The basic, exploratory relevance process is only a
simple exploratory relevance process, with about the minimum of processing needed to
function. This article describes possible improvements to the basic, exploratory
relevance process, to produce a more sophisticated exploratory relevance process.
Possible improvements are varying update frequency, making pattern instance addition
explicitly dependent on relevance and examining relevance for regions.
|
| Read this article in: Microsoft Word Format PDF Format |
|
|
An Attempt to Generalize AI - Part 8: Forgetting as Part of the Exploratory Relevance Process
By Paul Almond |
07 May 2010 |
|
This is the eighth in a series of articles attempting an overview of how minds may work
and how similar systems could be implemented in computers. Previous articles have
described a probabilistic hierarchy based on patterns. A pattern has a specification
describing a set, or population, of pattern instances, distributed throughout a
hierarchy containing the pattern instances of all the patterns. Each pattern's set of
pattern instances is used to obtain statistical information for probabilistic
predictions. Each pattern's population of pattern instances is to be specified in a
very general way, to provide a very general ontology. An exploratory relevance process
has been described, which achieves relevance of the hierarchy by removing low-relevance
pattern instances while the hierarchy "grows", so that it will tend to retreat from
low-relevance regions and grow into high-relevance ones. The exploratory relevance
process uses a relevance measurement process, which back-propagates relevance through
the hierarchy. The exploratory relevance process, as previously described, does not
provide forgetting - the removal of obsolete pattern instances which are no longer
having any useful effect on the hierarchy. A forgetting process was described earlier,
but that description now needs revising after the change to a completely probabilistic
hierarchy. In this article, the relevance measurement process is modified to take
account of obsolescence of pattern instances, so that they will tend to be assigned
low relevance and the exploratory relevance process will remove them. Forgetting is
reduced to being merely a special case of the exploratory relevance process.
|
| Read this article in: Microsoft Word Format PDF Format |
|
|
An Attempt to Generalize AI - Part 7: A Basic, Exploratory Relevance Process
By Paul Almond |
18 April 2010 |
|
This is the seventh in a series of articles attempting an overview of how minds may work
and how similar systems could be implemented in computers. Previous articles have
described a probabilistic hierarchy based on patterns. A pattern has a specification
describing a set, or population, of pattern instances, distributed throughout a
hierarchy containing the pattern instances of all the patterns. Each pattern’s set of
pattern instances is used to obtain statistical information for probabilistic
predictions. Each pattern’s population of pattern instances is to be described in a very
general way, to provide a very general ontology. A way is needed of ensuring that the
hierarchy only represents relevant features of the world. This requires a way of
measuring the relevance of individual pattern instances in the hierarchy, or of parts
of the hierarchy. The sixth article described such a measurement process. Such a
measurement process is made possible by the way that the action selection process,
described in the second article, requires probabilistic predictions of specific pattern
instances corresponding to future evaluation function score values. This enables a
measurement process in which these pattern instances are regarded as relevant,
relevance then being back-propagated from them through the rest of the hierarchy. In
this article, this measurement process is used in an exploratory process that reduces
the hierarchy, by removing pattern instances, where it appears insufficiently relevant
to justify its existence, while continually extending the hierarchy so that it grows
into relevant regions.
|
| Read this article in: Microsoft Word Format PDF Format |
|
|
An Attempt to Generalize AI - Part 6: Measuring Relevance
By Paul Almond |
08 April 2010 |
|
This is the sixth in a series of articles attempting an overview of how minds may work
and how similar systems could be implemented in computers. Previous articles have
described a probabilistic hierarchy based on patterns. A pattern has a specification
describing a set, or population, of pattern instances, distributed throughout a
hierarchy containing the pattern instances of all the patterns. Each pattern’s set of
pattern instances is used to obtain statistical information for probabilistic
predictions. Each pattern’s population of pattern instances is to be described in a very
general way, to provide a very general ontology. The fourth article discussed the need
to focus the hierarchy on what is relevant, and how this requires the ability to remove
pattern instances from the hierarchy, and the fifth article modified the description of
the hierarchy, making it completely probabilistic – without reliance on any special case
of certainty – to enable this. The hierarchy will be made relevant by an exploratory
process that extends it where it is likely to be relevant, and prunes it where it is
less relevant. Such a process will need a way of measuring the relevance of individual
pattern instances in the hierarchy, or of parts of the hierarchy. This article describes
such a measurement process. The measurement process is made possible by the way that the
action selection process, described in the second article, requires probabilistic
predictions of specific pattern instances corresponding to future evaluation function
score values. This enables a measurement process in which these pattern instances are
regarded as relevant, relevance then being back-propagated from them through the rest of
the hierarchy.
|
| Read this article in: Microsoft Word Format PDF Format |
|
|
On Macros
By Paul Almond |
03 April 2010 |
|
Software users often have requirements that are not met by the software. One solution to
this is macros – external scripts that can interact with programs to provide extra
functionality – but such a solution has shortcomings. In some environments, macros do
not have logical access to the user interface of programs. Even when this is available,
macros typically provide extra functionality that is not integrated with the user
interfaces of programs. An approach is proposed which deals with these issues. Programs
have two distinct user interfaces: a human user interface and a logical user interface.
Macros can interact with the logical user interface of the program. Elements can be
added to the human user interface of a program without any modification of its internal
processing, the added elements interacting only with external macros, and this can make
the functionality of a program appear to be extended. The extra functionality required
in programs may be provided by adapting computer language software so that any programs
produced with it automatically have the capability of working with macros in this way.
|
| Read this article in: Microsoft Word Format PDF Format |
|
|
An Attempt to Generalize AI - Part 5: A Completely Probabilistic Hierarchy
By Paul Almond |
27 March 2010 |
|
This is the fifth in a series of articles attempting an overview of how minds may work
and how similar systems could be implemented in computers. The first article described
a hierarchy based on patterns. A pattern has a specification describing a set, or
population, of pattern instances, distributed throughout a hierarchy containing the
pattern instances of all the patterns. Each pattern's set of pattern instances is used
to obtain statistical information for predictions. Each pattern's population of pattern
instances is to be described in a very general way, to provide a very general ontology.
The fourth article discussed the need to focus the hierarchy on what is relevant, and
how this requires the ability to remove pattern instances from the hierarchy. This is
likely to cause some pattern instances to have missing pattern inputs, so that their
states, and the states of any pattern instances which depend, directly or indirectly on
them, cannot be determined with certainty. This would make the process of "fixing", an
important aspect of the system up until now, impractical. The hierarchy needs to be
modified so that it does not rely on any special-case "fixing" process, instead allowing
pattern instances to be dealt with by purely probabilistic processes. This issue is
dealt with in this article. This "fault tolerant" hierarchy will prepare for later
articles to discuss how the system can be focused only on the relevant, and made more
efficient, by preventing explicit representation of all pattern instances.
|
| Read this article in: Microsoft Word Format PDF Format |
|
|
An Attempt to Generalize AI - Part 4: Modeling Efficiency
By Paul Almond |
20 March 2010 |
|
This is the fourth in a series of articles attempting to give an overview of how minds
may work and how similar systems could be implemented in computers. The first article
described a probabilistic, hierarchical modeling system, based on patterns, which are
sets of pattern instances, intended to provide a general ontology. The second article
described the use of this for planning actions. A serious issue with AI systems is
ensuring that the computation that is done is useful. A system like this finds
patterns, then patterns based on the patterns, and so on. Only a small amount of this
computation will be relevant. Unless something is done to prevent it, there will be a
huge proliferation of pattern instances that are of little use in the making of
predictions. This article explores this issue, and starts to consider, very broadly,
what may be involved in reducing the number of pattern instances in the model. This
will be built on in later articles, which will discuss specific methods of achieving
what is discussed here.
|
| Read this article in: Microsoft Word Format PDF Format |
|
|
How to Prevent Phishing
By Paul Almond |
15 March 2010 |
|
A method is described for preventing phishing. In a secure login process, user details
(such as the username and password) are combined with some identifying code associated
with the website, and automatically obtained from it by the web browser, and a
cryptographic hash function applied to this information, generating a hash code that is
sent to the website. An obvious example of such a unique identifying code is the
website's address, although a code associated with a digital signature, identifying the
owner of the website, could be used. The hash code must match the value expected from a
user for a successful login. The information sent to the server in a secure login
depends on what is being logged into, causing the information stolen by a fake website,
in a secure login, to differ from what the real website expects. The method involves
ensuring that the user makes secure logins, in which the hashing process as described
above occurs. If this happens, the user is safe, even if he/she logs into a fake
website. Secure logins are made hard to avoid. They are associated with specific
signals and actions that the user receives and performs, that are unique to a secure
login. Various methods of achieving this are described.
|
| Read this article in: Microsoft Word Format PDF Format |
|
|
An Attempt to Generalize AI - Part 3: Forgetting
By Paul Almond |
13 March 2010 |
|
This is the third in a series of articles attempting to give an overview of how minds
may work and how similar systems could be implemented in computers. The first article
described a probabilistic, hierarchical modeling system intended to provide a general
ontology. The second article described the use of this for planning actions. The
hierarchy as described so far is idealized. One way in which it is impractical is that
all pattern instances exist in the hierarchy permanently. It is implausible that the
human brain works in this way. It would mean, for example, that every single input to a
light sensitive cell on your retina was stored permanently, so that your brain would
contain a detailed "video recording" of your entire life. A system working like this
would waste computing resources on dealing with pattern instances that are no longer
relevant. A basic forgetting procedure is described, in which a pattern instance is
erased when it has a known state and no longer has any effect on the hierarchy: This is
when it is no longer being used as a pattern input by any other pattern instances which
do not yet have known states. The possible desirability of increasing or reducing the
amount of forgetting is discussed, as well as ways of doing this. The basic forgetting
procedure is also related to human experience of memory and forgetting. The two kinds
of forgetting are not exactly the same, but we should expend some correspondence
between them.
|
| Read this article in: Microsoft Word Format PDF Format |
|
|
An Attempt to Generalize AI - Part 2: Planning and Actions
By Paul Almond |
06 March 2010 |
|
This is the second in a series of articles that will attempt to give an overview of how
minds may work and how similar systems could be implemented in computers. The first
article described a probabilistic, hierarchical modeling system intended to provide a
general ontology. This article describes the use of this for planning actions. The
approach to planning of actions makes planning a process occurring mainly in the
hierarchical model itself. An evaluation function score is continually computed and
encoded as input data for the hierarchy, just as conventional, external inputs are, the
scores corresponding to bottom-level pattern instances. When an output is imminent, the
different values for the output are tried, the hierarchy being updated appropriately,
and predictions of a future input of the evaluation function score are obtained. It is
shown that an approach like this allows learning. There is no explicit attempt to
represent "consciousness" or a "self". Instead, in humans, these are objects within the
modeling system, constructed to explain previous behavior, and are only different from
other objects in that they relate outputs with later outputs, and therefore relate to
intentionality.
|
| Read this article in: Microsoft Word Format PDF Format |
|
|
An Attempt to Generalize AI - Part 1: The Modeling System
By Paul Almond |
13 February 2010 |
|
This is the first in a series of articles that will attempt to give an overview of how
minds may work and how similar systems could be implemented in computers. The approach
is based on patterns. A pattern has a specification describing a set, or
population, of pattern instances, distributed throughout a hierarchy containing
the pattern instances of all the patterns. Each pattern’s set of pattern instances is
used to obtain statistical information for predictions. Each pattern’s population of
pattern instances is to be described in a very general way, to provide a very general
ontology. Later articles will discuss how the system can be made more efficient, by
introducing an analogy of “forgetting” and preventing explicit representation of all
pattern instances, and how the system can be used to provide intelligent behavior,
rather than just modeling.
|
| Read this article in: Microsoft Word Format PDF Format |
|
|
On Information Scarcity
By Paul Almond |
08 February 2010 |
|
Computing and network technologies mean that some products can be considered as
information products and can be distributed to large numbers of people for not much more
than it would cost to supply them to small numbers of people. This does not happen: The
need for sellers to make money causes a scarcity of information products, which is shown
to be artificial. Distribution of information products by the market is inefficient. A
method of dealing with this scarcity problem is described that does not rely on state
interference and which abstracts agreements involving buyers and sellers. The economics
behind such an idea are already understood. What is proposed here is a business model
that can make them work on a large scale, without any centralized system of assigning
value to products.
|
| Read this article in: Microsoft Word Format PDF Format |
|
|
On Double Standards About Causes in Religious Apologetics
By Paul Almond |
08 January 2010 |
|
Some arguments for the existence of God, such as William Lane Craig's Kalam cosmological
argument, assume, or try to show, that the universe has a start, before arguing that
anything which begins to exist has a cause. The cause is claimed to be God. There are
problems with justifying the assertion of a universal rule that everything that begins
has a cause. Asserting such a rule would need to be justified on the grounds that some
philosophical view requires it or that our experience of the world shows it to be the
case. Such justifications, when explored in more detail, would eliminate God as easily
as they eliminate an uncaused thing or event - if the reasoning behind them is even
valid. Claiming that everything that begins has a cause in a proof of God therefore
involves a double standard.
|
| Read this article in: Microsoft Word Format PDF Format |
|
|
On Drugs Policy
By Paul Almond |
29 November 2009 |
|
The war on drugs has not solved the problem of illegal drugs. That leaves us with a
choice between legalizing drugs or finding another way of fighting the illegal drugs
industry. This article proposes an approach for fighting the illegal drugs industry by
waging economic warfare against it. This combines taking away its established customers
to deprive of it money with ensuring that drug dealing has severe penalties and high
risks, so that dealers have a high risk of severe penalties for little money. The idea
is to reduce the rationality of working in the illegal drugs industry and force it into
a decline, reducing its tendency to create more habitual drug users. The progression
from non-user to dependent user is therefore interrupted. Although the approach would
initially be expensive, after it has damaged the illegal drugs industry it would be
cheaper to maintain.
|
| Read this article in: Microsoft Word Format PDF Format |
|
|
Launching anything is good: How Governments Could Promote Space Development
By Paul Almond |
01 November 2009 |
|
Government organizations have failed to develop space technology to the point where
economies of scale apply and space activity becomes self-supporting. Private businesses
may have the problem that in the early stages, before space is economically developed,
there will be a limited market for space travel. When space is developed there could be
significant economic returns, but development of the technology to achieve that could
be hindered by the limited market in the early stages. This article suggests that a
government can help with this by providing a guaranteed market for space travel and
development.
|
| Read this article in: Microsoft Word Format PDF Format |
|
|
There is no god, and yes I can say that.
By Paul Almond |
07 June 2009 |
|
Theists and agnostics often say that nobody can say that there is no god. This is
not merely disagreement with the statement that there is no god - for example because
they think the claim is reasonable or is supported by evidence. They mean that nobody can validly say
that there is no god even in principle, because this would mean proving with certainty that God
does not exist, and total knowledge would be needed to rule out the idea completely. This article argues
against this, showing that it is semantically consistent to say that there is no god, if we consider God
to be sufficiently implausible, even in the absence of an absolute proof of God's non-existence.
Any claim made for an object's existence implies many claims of the non-existence of "invalidators" -
implausible objects that would invalidate the claim, or at least make it unsafe, if they existed. If God
is like an invalidator, asserting God's non-existence is entirely consistent, regardless of what position
is taken on invalidators, with how semantics normally deals with invalidators.
|
| Read this article in: Microsoft Word Format PDF Format |
|
|
A Proposal for General AI Modeling
By Paul Almond |
10 April 2009 |
|
A probabilistic, hierarchical modeling system is proposed which generates a hierarchy of
patterns, each with a probability. Lower-level pattern (or object) probabilities are
generated from the data to be processed and used to generate still higher-level
probabilities, and so on. High-level pattern probabilities also influence low-level
ones, ultimately causing low-level, unknown data to be assigned probabilities.
Meta-patterns are also used, providing a more general ontology by allowing description
of the occurrence of the same kinds of patterns in different contexts and at different
levels of the hierarchy. The modeling approach is intended for use in planning as
modeling, an approach to AI planning which involves almost all planning being performed
by a modeling system. When it is used in this way, predictions of future evaluation
function scores could be obtained from the system to evaluate any given output. To make
development of the modeling system more feasible, this article describes a modeling
system intended to fill in missing parts of images. Such a modeling system could be
adapted for use in general AI, in planning as modeling, later.
|
| Read this article in: Microsoft Word Format PDF Format |
|
|
Machines Like Us Interviews: John Searle
Paul Almond interviews John Searle. Edited by Norm Nason |
15 March 2009 |
|
John Searle is Slusser Professor of Philosophy at the University of California, Berkeley,
and has made notable contributions to the philosophy of language and the philosophy of mind. He was
awarded the Jean Nicod Prize in 2000 and the National Humanities Medal in 2004. Professor Searle is
well-known for his criticism of the idea that artificial intelligence research will lead to conscious
machines, and in particular for his famous Chinese Room Argument.
I interviewed Professor Searle for the website Machines Like Us. The link given here is
an external link to the interview on the Machines Like Us website at http://machineslikeus.com/interviews/machines-us-interviews-john-searle.
I would like to thank Professor Searle for giving his time for this interview and Machines Like Us for inviting me to conduct it.
|
| Read this interview at Machines Like Us |
|
|
Is it fair for God to reward belief and punish disbelief?
By Paul Almond |
10 February 2009 |
|
Many Christians claim that believers will be rewarded in the afterlife, while
disbelievers will be punished. You cannot rationally choose to believe what you find
unbelievable. God seems unfairly to reward people for irrationality. Many claims might
be at least as likely to be correct as Christianity, and it seems that non-believers
will be rewarded for choosing the right claim by chance, when nothing suggests that it
is right to a non-believer. God seems unfairly to reward good luck. Christians could
reply that you are not being asked just to believe - you can test Christianity - however
this demands a special investment of time and resources that other claims of equal or
greater probability may not receive. This also rewards irrationality and punishes
rationality. God as imagined by many Christians is unfair. God is supposedly perfect,
and unfairness is imperfect, so such a god does not exist.
|
| Read this article in: Microsoft Word Format PDF Format HTML Format |
|
|
Searle's Argument Against AI and Emergent Properties
By Paul Almond |
28 December 2008 |
|
John Searle argues against the idea that an appropriately programmed computer would be
conscious by saying that consciousness is an emergent property of a physical system,
caused by a particular kind of physical process, and that you should regard
consciousness as likely when the kinds of physical processes causing your own
consciousness are occurring. According to this argument, there is no reason to think
that a computer is likely to be conscious, irrespective of its behaviour, because that
behaviour is not being caused by physical processes known to cause consciousness.
This is a flawed position. While it is reasonable to regard consciousness as an
emergent property of a physical system there is no profound sense in which it can be
said that different people's brains work according to the same kinds of processes and
an appropriately programmed computer and a human brain would work according to different
processes. Any difference between these situations is just a matter of degree and any
argument that we should presume other people conscious because their brains work in
basically the same sort of way could also be used to justify presuming an appropriately
programmed computer conscious.
|
| Read this article in: Microsoft Word Format PDF Format HTML Format |
|
|
Against the Supernatural as a Profound Idea
By Paul Almond |
01 November 2008 |
|
Shows that the term "supernatural", and similar terms, cannot have any of the profound
meanings that people normally think they imply. This leaves a choice of discarding the
word as incoherent or accepting its use but only with less profound meanings. This has
implications for the frequent theistic claim that a "supernatural" god exists who is
profoundly different to anything else.
|
| Read this article in: Microsoft Word Format PDF Format HTML Format |
|
|
The Principle of Modal Realism Equivalence
By Paul Almond |
13 August 2008 |
|
The principle of modal realism equivalence states that the methodology we apply for
making statistical predictions about the world and the results obtained should be the
same regardless of whether or not modal realism is true. If modal realism is true then
probability calculations would be about where we are likely to be in the set of different,
actual worlds. If modal realism is false then they would be about which is likely to be the actual world in
the set of possible worlds. Both approaches differ only in the semantics of "possible" and "actual".
|
| Read this article in: Microsoft Word Format PDF Format HTML Format |
|
|
The Theistic Apologist's Worst Nightmare: A Reality In Which Time Is Unimportant
By Paul Almond |
03 August 2008 |
|
Some views of reality feature time as a constructed phenomenon instead of part of the
fundamental framework of reality. These kinds of views have serious consequences for some
arguments used to support the existence of God by claiming that God is needed to "cause"
the universe, because they relegate time and causality to an unimportant, and possibly
parochial, position. Further, if most of reality is atemporal, as some views now suggest,
then the entire concept of God is questionable because it appears to rely on concepts such
as intention which only make sense as temporal ideas.
|
| Read this article in: Microsoft Word Format PDF Format HTML Format |
|
|
Civilization-Level Quantum Suicide
By Paul Almond |
02 June 2008 |
|
If you think quantum suicide is valid then you should expect an advanced civilization to
think that quantum suicide at the level of an entire civilization is valid, as you should
expect it to know anything that you know. If you also think the technological singularity
idea is correct then you should expect a civilization to start performing
civilization-level quantum suicide around the time it undergoes a technological
singularity. Motives for quantum suicide could be quantum suicide reality
editing - using quantum suicide to enter desirable situations - and quantum suicide
computing - using quantum suicide to gain huge computing capability. This provides
possible answers for the Fermi paradox, the Doomsday argument and the simulation
hypothesis. This article is not arguing for or against the validity of quantum suicide,
but merely considering implications of quantum suicide being valid.
|
| Read this article in: Microsoft Word Format PDF Format HTML Format |
|
|
Minds, Substrate, Measure and Value - Part 3: The Problem of Arbitrariness of Interpretation
By Paul Almond |
11 May 2008 |
|
Previous articles in this series argued that probability in some thought experiments in
which you are unsure about your status is based on the substrates on which algorithms are
running and that the nature of the substrate is statistically relevant to the measure
with which an algorithm runs and affects the measure of any minds associated with such
algorithms. This article provides a deeper explanation by discussing arbitrariness of
interpretation. Previously called "multiple realizability" by John Searle, this is the
problem caused by the need to apply an interpretation to a physical system to say that
it is running an algorithm and the possibility of applying any interpretation to obtain
any algorithm, leading to an apparent observer subjectivity in the algorithms that a
physical system is running. Searle argues that multiple realizability makes the strong AI
hypothesis incoherent. This conclusion is unnecessary, although the strong AI hypothesis
needs some clarification. Instead, the many-interpretations position is proposed
to deal with the issue of observer subjectivity.
|
| Read this article in: Microsoft Word Format PDF Format HTML Format |
|
|
Minds, Substrate, Measure and Value - Part 2: Extra Information About Substrate Dependence
By Paul Almond |
03 November 2007 |
|
A continuation of the series of articles exploring the relationship between minds and
physical systems (substrates) on which they are based. The previous
article Minds, Substrate, Measure and Value, Part 1: Substrate Dependence used
a thought experiment to show that substrate matters, but not in the way that John Searle
thinks. It was shown that the substrate influences the probabilities that you are in
various situations in some thought experiments in which there is uncertainty about the
substrate on which you currently exist. The substrate is statistically important
and influences the measure of minds associated with computing done on it. This article
will strengthen the argument made in the previous article.
|
| Read this article in: Microsoft Word Format PDF Format HTML Format |
|
|
Minds, Substrate, Measure and Value - Part 1: Substrate Dependence
By Paul Almond |
12 September 2007 |
|
The first in a series of articles exploring the relationship between minds and substrates
on which they are based. Strong AI advocates typically maintain that the substrate is
irrelevant, provided that the required computation can be performed on it, and only the
computation matters. John Searle, an opponent of strong AI, argues that the substrate
does matter and that a mind is not just computation on a substrate but is caused by
specific physical processes. Searle states that there is no reason to assume that all
substrates that allow general computation can support minds. A thought experiment will show that
the substrate matters, but not in the way that Searle thinks. The substrate matters
statistically and affects the measure of a mind.
|
| Read this article in: Microsoft Word Format PDF Format HTML Format |
|
|
Planning As Modelling: A New Version
By Paul Almond |
29 July 2007 |
|
Describes a revised version of the planning as modelling approach to planning in
artificial intelligence (AI). Planning As Modelling is based on the idea that planning
is prediction and uses an AI system's modelling system to produce probabilistic
predictions of future behaviour that are equivalent to planning of future
behaviour. The tree search approach described in earlier articles is no longer necessary,
the processing that it did being absorbed into the modelling system as part of its
prediction.
|
| Read this article in: Microsoft Word Format PDF Format HTML Format |
|
|
Planning As Modelling: A Revised Description
By Paul Almond |
27 April 2007 |
|
Describes the planning as modelling approach to planning in artificial
intelligence (AI). This uses an AI system's modelling system to produce probabilistic
predictions of future behaviour that are equivalent to planning of future
behaviour. The article combines concepts from previous articles about planning as
modelling.
|
| Read this article in: Microsoft Word Format PDF Format HTML Format |
|
|
Resolving the Horizon Problem in Planning As Modelling
By Paul Almond |
30 March 2007 |
|
Previous articles such as Planning as Modelling in AI
and Programming of Planning as Modelling in AI suggested the planning as modelling approach to planning
in artificial intelligence.
Planning as modelling provides AI planning by using the AI's modelling
system to produce probabilistic predictions of future behaviour equivalent to planning
of future behaviour. The approach as described previously has a "horizon" problem.
Planning as modelling is supposed to limit the searching required for optimum behaviour
according to the likelihood of different possible futures. A course of action could be
found desirable after being considered part of an unlikely future, the previous
consideration that the behaviour is unlikely meaning its selection was based on a
shallow, and therefore possibly unreliable, search. This article will modify planning
as modelling to deal with this problem.
|
| Read this article in: Microsoft Word Format PDF Format HTML Format |
|
|
A Refutation of Plantinga's Modal Ontological Argument - and why it even suggests a disproof of God
By Paul Almond |
28 February 2007 |
|
Alvin Plantinga's modal ontological argument is intended to prove God's existence
without reference to empirical observation, but instead showing that the existence of God
logically follows from the definition of God. The argument suggests that the
actual existence of God follows from the possible existence of God as a necessary entity.
This article will show that the modal ontological argument is invalid and, worse, that
the assumptions in the modal ontological argument are more useful in disproving
God's existence.
|
| Read this article in: Microsoft Word Format PDF Format HTML Format |
|
|
Why Is Space 3D?
By Paul Almond |
21 January 2007 |
|
Our perception of the world is that it has 3D space. This highly speculative article will
suggest an anthropic explanation of this. The article supports an ensemble view of
reality in which many space-times exist and considers the type of space-time in which
most observers will find themselves. It is suggested that such a space-time is the type
in which we should most likely expect to find ourselves.
|
| Read this article in: Microsoft Word Format PDF Format HTML Format |
|
|
Can God Exist Outside Space-Time?
By Paul Almond |
20 January 2007 |
|
Many theists claim that God exists outside of space-time and created it. This article
will show that this is incoherent without significant qualification.
An attempt is made to interpret the concept of "creation" in a tenseless,
or atemporal (timeless), way and it is then shown that it still does not work without significant qualification on account of it
being ontologically meaningless to claim any difference between the "creator" and the
"created" in an atemporal situation. Although there are ways in which the concept could be qualified to make it coherent,
this may reduce God's resemblance to the entity claimed to exist by theists.
|
| Read this article in: Microsoft Word Format PDF Format HTML Format |
|
|
Programming of Planning As Modelling in AI
By Paul Almond |
28 December 2006 |
|
Previous articles such as Planning as Modelling in AI suggested the planning as modelling approach to planning
in artificial intelligence. This provides planning in an AI system by using the AI's
modelling system to produce probabilistic predictions of future behaviour that are
equivalent to planning of future behaviour. This article will provide example computer
programs in Pascal for planning as modelling.
|
| Read this article in: Microsoft Word Format PDF Format HTML Format |
|
|
Planning as Modelling in AI
By Paul Almond |
26 November 2006 |
|
Previous articles, How AI Would Work and Occam’s Razor Part 9: Representation and Planning of Actions in Artificial Intelligence, described planning as modelling, an approach
to planning in AI which breaks the partition between planning and modelling which has traditionally been a
feature of AI. In planning as modelling the AI system's predictions of its own future behaviour are used
to plan its future behaviour. Planning therefore occurs almost completely within the modelling system as a
trivial special case of modelling. In previous articles planning as modelling was discussed in the context
of the specific hierarchical, probabilistic modelling system which has been proposed. The idea is more
general than this and can be used with other probabilistic modelling systems. This article will describe
planning as modelling in a more general context, separate from the specific details of the probabilistic
hierarchy previously proposed, so that planning as modelling is easier to understand and not seen as
dependent on other features of the AI system. It will also describe a change to the concept described
in previous articles about planning which will improve the AI system’s planning ability by including
inputs in its search for optimum behaviour.
|
| Read this article in: Microsoft Word Format PDF Format HTML Format |
|
|
Downward Transfer of Probabilities in AI
By Paul Almond |
15 October 2006 |
|
My previous article How AI Would Work and
other articles proposed an approach to AI in which probabilities of meanings obtained
from probabilistic interpretation of partial models are stored in a hierarchy.
Information on the bottom level of the hierarchy, where input and output events occur,
is abstracted on higher levels. Downward transfer of probabilities is an important
process in the AI system, but has not yet been described in any detail. This article
will give an idea of how downward transfer of probabilities could work.
|
| Read this article in: Microsoft Word Format PDF Format HTML Format |
|
|
AI as a Boundary System
By Paul Almond |
17 September 2006 |
|
Proposes a paradigm for dealing with the AI (artificial intelligence) system suggested
in my previous article How AI Would Work and other
articles. The paradigm is that of an AI system as a boundary system – a system with limited
capabilities, confined to a simple, basic “layer” of computation between two “worlds” - outside reality
and the AI system’s own functional hierarchy.
|
| Read this article in: Microsoft Word Format PDF Format HTML Format |
|
|
How AI Would Work
By Paul Almond |
04 September 2006 |
|
Outlines how artificial intelligence (AI) should work. In previous articles I have
proposed an approach for making AI, but I have not yet outlined it in a single
article. This article will give an overview of the proposed AI system. This article is
not a substitute for previous articles, but it should make my approach more
accessible. It will also provide a discussion of how the carpet texture problem can be
resolved in this system.
|
| Read this article in: Microsoft Word Format PDF Format HTML Format |
|
|
Funding of Ambitious Projects
By Paul Almond |
05 August 2006 |
|
Very ambitious projects have a track record of going nowhere. Why should anyone invest in a dangerous, ambitious project like this when it is well known that they tend to fail?
This is a self-fulfilling prophecy. Anyone considering investing has to consider the problem that ambitious projects tend to fail by not achieving adequate investment and this is likely to dissuade many investors. This expectation of failure, however, can be the very cause of the failure! This article will describe how the routine availability of conditional investment could be a solution for this problem.
|
| Read this article in: Microsoft Word Format PDF Format HTML Format |
|
|
Occam's Razor Part 9: Representation and Planning of Actions in Artificial Intelligence
By Paul Almond |
29 July 2006 |
|
Describes use of the probabilistic, hierarchical ontology from the previous article in planning. With the previous article this gives an overview of how an intelligent machine would work, from processing of sensory inputs to acting. There is no separate, hierarchical planning system, "closely-coupled" to the hierarchy of meaning extraction from sensory inputs, for transmitting actions down to the bottom level of the hierarchy, as may be expected. The hierarchical model makes predictions about reality and the planning of actions is really just the system predicting its own behaviour, so planning is modelling and can be performed by the modelling system.
|
| Read this article in: Microsoft Word Format PDF Format HTML Format |
|
|
Occam's Razor Part 8: Modelling in Artificial Intelligence
By Paul Almond |
09 June 2006 |
|
Builds on previous articles in the series to show how a computer can use sensory inputs obtained from reality to assemble a probabilistically and hierarchically expressed worldview - a model of how reality works. This is explored in some detail and the emphasis is on making intelligent machines rather than philosophy.
|
| Read this article in: Microsoft Word Format PDF Format HTML Format |
|
|
How to Rig an Internet Election
By Paul Almond |
04 May 2006 |
|
This article will describe a weakness in the security of Internet voting processes. Although an electoral process may be well defended against direct electoral attacks it will be much harder to defend it against indirect electoral attacks. I will explain what I mean by direct and indirect attacks and how an indirect attack could be made: I will be showing how to rig an Internet election.
|
| Click here to read this article. |
|
|
Occam's Razor Part 7: Hierarchy and Ontology
By Paul Almond |
30 April 2006 |
|
This is the latest article in the series on Occam's razor. The previous article
proposed that partial models are best described by meaning extraction algorithms,
but did not include any concept of hierarchy. This article will show how the previous
concept of partial models as meaning extraction algorithms can be extended to allow
construction of hierarchical models. There will also be an ontological discussion
relating this to the concept of "existence".
|
| Click here to read this article. |
|
|
Many-Worlds Assisted Mind Uploading: A Thought Experiment
By Paul Almond |
06 April 2006 |
|
This article proposes and discusses a thought experiment relating to mind uploading,
the many-worlds interpretation of quantum mechanics, quantum suicide and quantum
immortality. This is a speculative article and I do not take a final position on all
of the issues that it raises, but it facilitates philosophical discussion about what we
are and what continuity of self means.
|
| Click here to read this article. |
|
|
Occam's Razor Part 6: Partial Models as "Envelopes"
By Paul Almond |
26 February 2006 |
|
To use Occam's razor in artificial intelligence, or to test real claims, we need a
better way of representing models than what we have been discussing so far. It must be
general enough to express any human models of reality and must somehow allow concepts
within a model to relate to concepts within a human equivalent of the same model.
This is the subject of this article. It will not deal directly with Occam's razor,
but rather with how we can describe partial models differently to enable a complete
worldview to be described by a number of partial models being used together.
|
| Click here to read this article. |
|
|
When Are People Responsible?
By Paul Almond |
09 February 2006 |
|
There is controversy about the extent to which we should hold people "responsible"
for actions. In some situations we say that a person is "responsible" for an action,
but in others, despite the act being committed, that he/she is not "responsible".
What is this "responsibility"? If we can say when people have it and when they do not
then we would be answering this question. Common ideas about assigning responsibility
are too vague or inconsistent to be philosophically useful. This article will suggest
an approach that is consistent and could be formalized.
|
| Click here to read this article. |
|
|
Occam's Razor Part 5: How Mapping Can Work
By Paul Almond |
14 January 2006 |
|
The previous articles in this series have set out assumptions needed for Occam's
razor to work and started a justification of it. The concept of "mapping" of partial
models onto complete models has been used, but no explanation has been given of what
it means to say that a partial model "maps onto" a complete model. This article will
show how a partial model can be said to "map onto" a complete model and why partial
models favoured by Occam's razor will tend to map onto large numbers of complete
models, making them more likely to be "correct".
|
| Click here to read this article. |
|
|
Occam's Razor Part 4: An Overview of How Occam's Razor Works
By Paul Almond |
24 December 2005 |
|
This is the fourth article in the series about Occam's razor. The last article
suggested that a single algorithmic description of reality is correct and that no
particular algorithmic description should be given preferred status. This means that
any algorithmic description of reality should be as likely as any other to be the
correct one. This may seem to contradict Occam's razor which suggests that certain
descriptions of reality - the "simpler" ones - are preferred. This article will
resolve this and describe how Occam's razor can actually work.
|
| Click here to read this article. |
|
|
Against the Idea that Religious Belief is Needed to Have Ethics
By Paul Almond |
04 December 2005 |
|
There is a type of question which challenges an atheist to show how society could
have any ethics if there were no religion. The question is clearly intended to make
the point that society could not be ethical in the absence of religious belief to
provide ethical values by the presumed inability of an atheist to answer the question
satisfactorily. The purpose of this article is to refute the argument implicit in the
asking of this question.
|
| Click here to read this article. |
|
|
Occam's Razor Part 3: Assumptions About Reality
By Paul Almond |
13 November 2005 |
|
This is the third article in the series about Occam's razor. There needs to be a kind
of way of looking at reality which allows us to determine which model is best for
describing it. This article will deal with this. It will present the minimal
meta-assumptions regarding reality that are needed to allow Occam's razor to be
obtained.
|
| Click here to read this article. |
|
|
Against the Idea that God is Superior to Logic
By Paul Almond |
20 October 2005 |
|
Some theists claim that God is superior to logic, which is to say that God is unrestrained
by logic and that God is 'above logic'. This article will demonstrate that such claims
cause all proofs of God's existence to fail and that anyone making the claim that God is
above logic is depriving him/herself of any good arguments supporting God's existence.
|
| Click here to read this article. |
|
|
Occam's Razor Part 2: Principles of Language
By Paul Almond |
9 October 2005 |
|
This is the second article in the series about Occam's razor.
The previous article Occam's
Razor Part 1: What is Occam's Razor? introduced the concept of Occam's razor and
discussed it in general terms. This article will deal with principles about how theories
should be expressed. The next article will deal with assumptions about reality, and how
theories relate to it, that are necessary to allow Occam's razor to be derived.
|
| Click here to read this article. |
|
|
The Diminished God Refutation: Why unlikely sequences of events do not prove a god
By Paul Almond |
24 September 2005 |
|
One type of attempted proof of God refers to a sequence of events known to have
occurred, but considered to be outrageously unlikely. These events usually have a combined probability which is supposed to be very low and the proof suggests
that a God must have therefore been the real cause.
This article will refute this proof and weaken the case for the existence of God.
|
| Click here to read this article. |
|
|
Occam's Razor Part 1: What Is Occam's Razor?
By Paul Almond |
22 August 2005 |
|
Occam's razor is a philosophical idea with importance in a number of areas, including scientific theory selection, human psychology
and artifical intelligence. This is the first of a series of articles which will attempt to describe Occam's razor
in a formal way and present a justification for it. In this article the idea of Occam's razor is described and its importance discussed.
|
| Click here to read this article. |
|
|
What is a Low Level Language
By Paul Almond |
17 July 2005 |
|
The idea of 'low level' languages is one with which most computer programmers can deal intuitively, yet the term 'low level' is harder to define than is commonly thought.
What constitutes a low level language is of some philosophical importance. This article is intended to address this problem. It will discuss the idea of 'low level' language, consider the need for a workable concept of 'low level language' in philosophy and propose a way of defining it.
This article will precede a series that I intend to write about Occam's razor.
|
| Click here to read this article. |
|
|
Improving Computer Security
By Paul Almond |
06 July 2005 |
|
This article discusses the threat posed to computer systems from programs that have been intentionally installed
on them by their legitimate users. Programs have enormous power on computers and the consequences of malicious intent or
carelessness by programmers can be serious. This issue will ultimately be more serious than the
issue of perimeter security in areas ranging from home computing to large scale government software contracts. This article proposes a security paradigm which could help to control this problem.
|
| Click here to read this article. |
|
|
Taking the Virtual out Of Virtual Reality
By Paul Almond |
17 March 2005 |
|
An idea for placing virtual reality into reality to form a shared-augmented reality, occupying the existing reality with which we are familiar.
The shared-augmented reality could be altered in real-time by its users and the emphasis would be on users generating content rather than on it being
provided by some central source. This would be equivalent to projecting a three-dimensional version of the existing internet into physical reality.
This article discusses the implications of such a system and gives some consideration to the issue of how it could be built.
|
| Click here to read this article. |
|
|
Getting Darwinian Evolution to Work - Part 2
By Paul Almond |
18 November 2004 |
|
In a previous article it was suggested that Darwinian evolution has higher levels of abstraction and that explicitly exploiting these in attempts to evolve software on computers could allow programs of greater sophistication. A multi-layered system was proposed which used the concept of 'evolving evolvability'. This article proposes an
improvement to this method which involves using a 'compilative' architecture.
|
| Click here to read this article. |
|
|
A Refutation of Penrose's Godel-Turing Proof that Computational Artificial Intelligence is Impossible
By Paul Almond |
24 October 2004 |
|
In his 1994 book Shadows of the Mind Roger Penrose gave a proof attempting to show that computational artificial intelligence is impossible. The proof is often known as the Godelian proof or the Godel-Turing proof and is intended to show that there are certain facts that can be known to humans yet can never be known to algorithms. This article provides a refutation of Penrose's proof.
|
| Click here to read this article. |
|
|
Did 'I' Write This?
By Paul Almond |
08 October 2004 |
|
Is it possible that certain technologies could appear to extend your life,
yet are so extreme that they actually fail because you are not the same person afterwards?
'Continuity of self' appears to be a simple and self-evident concept to many people who may have opinions
about whether various processes that humans can undergo would or would not preserve this continuity.
This article will suggest that this is a fallacy and that 'continuity of self' is an incoherent and unnecessary concept.
|
| Click here to read this article. |
|
|
Game Theory with Yourself
By Paul Almond |
30 December 2003 |
|
What would it be like to play a game involving a duplicate of yourself? This article discusses various scenarios involving mind uploading and 'free will' within the context of games in which your reward depends on the actions of your duplicate.
|
| Click here to read this article. |
|
|
Could Computers be Religious?
By Paul Almond |
11 August 2003 |
|
Discusses the possibility of computers acquiring religious beliefs in the future, within the context of a possible explanation for the occurrence of religion in human societies based on the social modelling capabilities of the brain.
|
| Click here to read this article. |
|
|
John Searle's Position within an Evolutionary Context
By Paul Almond |
09 August 2003 |
|
An argument against John Searle's reasoning about consciousness that uses Darwin's theory of evolution to attempt to weaken his case. It is suggested
that statements made by Searle suggest that human consciousness is a very high specificity feature of
some biological systems that is not easily explained by Darwin's theory of evolution.
|
| Click here to read this article. |
|
|
Indirect Mind Uploading: Using AI to Avoid Staying Dead
By Paul Almond |
09 August 2003 |
|
A consideration of life extension from an artificial intelligence
perspective, discussing the idea of indirect mind uploading - the idea that a
computer model could be made of a human mind without the use of sophisticated scanning technnologies.
The intention of this article is to present a mind uploading method that could be useful to a person who is alive today.
|
| Click here to read this article. |
|
|
Getting Darwinian Evolution to Work
By Paul Almond |
12 February 2003 |
|
Suggests that Darwinian evolution has higher levels of abstraction and that explicitly exploiting these in attempts to evolve software on computers could allow programs of greater sophistication. A multi-layered system is proposed which uses the concept of 'evolving evolvability'.
This article was originally published at to http://ai-depot.com/Articles/54/Evolution.html on 12 February 2003 and a copy published here on 26 October 2004.
|
| Click here to read this article. |
