Understanding the Brain p. 5 Josephson
Suppose, in the present scheme, there is a certain type of system corresponding to the abstraction classes of
objects . One instance of this type would consist of our example, the class of cats, which we may for the sake
of argument suppose comes into existence by some built-in mechanism through exposure to a particular cat.
We can build up a more complex system by creating, whenever an instance of a cat is encountered, a system
combining and linking together appropriately, in a way goverened by the various applications that there may be
of the class abstraction, a system for the particular cat with the system for the cat class. Such specific
applications might include recognising a cat as a being a member of the class of cats (and activating the system
corresponding to the class), or the process searching for a cat. There are a number of generic uses of
abstractions, demanding a collection of different types of functional circuitry. In accord with the concept of
observational emergence, complexes of a given type are formed only when the appropriate target condition is
satisfied.
What has been discussed above involves the use and integration into activity of isolated abstractions (e.g. just
cats, not relationships between cats and anything else) only. Once connections have been established between
higher levels and the level of activity, combinations of higher level constructs with useful translations of these
constructs into the level of perception and action can be created, such as assertions as to the existence of
relationships between elements of a class (an abstract connection with concrete correlates). There may be a
general mechanism to create such combinations, but one can anticipate also the existence of specific circuits for
handling a range of particularly useful forms, such as anticipation and making models. Planning involves
processes where activity involving abstractions and concrete activity are completely separated in time, which we
hypothesise is built up step by step in a process analogous to going from walking with considerable support to
walking with partial support to walking with no support (an example of the moving target concept), the
equivalent to support in the planning case being confirmation of the validity of a plan by testing it against the
reality. Information and rules can be tested initially in cases where the planning operations are carried out
during an action and compared with activity without affecting it. Next, planning processes can be carried out
separately from action, but in a situation where the outcome of planning can be tested immediately. The next
stage again is where a plan is prepared in a situation that is not immediate, using signs to represent the situation
concerned, further filtering being possible favouring the processes that are most reliable. Language permits yet
another stage of sophistication, involving communal knowledge and practices (which may be an important
factor structuring the life-scheme system postulated above).
A interesting issue arises here in connection with processes such as these. The success and reliability of, for
example, a planning process is dependent on the validity of particular laws in the given context (as applies even
in the context of simple processes such as taking a step). In other words, a given system applies specific rules
in a given context, and the system has to be adjusted to give rise to the desired outcome. The point is that it is
a good strategy in testing a generic mechanism to try out simple cases where success or failure can be quickly
determined and improvements based on feedback made and tested quickly. Thus we envisage that much of
intellectual development is based on systems that try out general types of rules, and quickly learn to discard the
majority, leaving a residuum that can be characterised as rational.
One point needs amplification in regard to the above, namely the relationship between activity and knowledge
(equivalently cognitive structure), the former being a process and the latter a state related to the process.
Cognitive structures are built up, already noted, by linking together systems concerned with processes, and
these links subsequently allow the process concerned to be recreated. Applying this concept to the case where
the activity concerned involves abstractions rather than overt activity, we see that activity at the abstract level,
which can be characterised as thought can, subsequent to a learning process related to some target being
achieved, become a structure capable of regenerating the same thoughts. Such a process is the equivalent in the
present scheme of the creation of semantic structures representing relationships in the discipline of artificial
intelligence. Language can now be seen as a special process that connects with such structures in a very
systematic way, allowing them to be manipulated at will to achieve particular ends and transferred between
individuals.