3rd Jan 1998
Could use a triplet S R F F=feeling or S S F where F gives strength of relationship / association between them.
So when S1 S2 Fx recognized do we get Sx Fx and then another one S2 S4 Fw Sz Fw recognized so Sx Fx Sz Fw do we have equivalent of Sx Sz Fz where Fz is “sum” of Fx + Fw such that feelings propagate such that most recent feeling applies.
In Pavlovian conditioning a response is elicited by an earlier stimulus – so response is triggered earlier in time that it would otherwise be. S1 S2 Ra S1 Ra does not wait for S2 but S1 to do this needs a feeling of “good” to start it executing and miss S2.
If we use a degree of similarity instead of exactly same then S1 S2 may be recognized as Sx and then have Sx Ra and then when just S1 part of Sx is received degree of similarity is high enough to trigger Sx and thus execute the Ra.
*! – Earlier thought (days ago) – Every response – no matter how trivial a muscle movement is involved has an associated – consistent stimulus from kinaesthetic senses. Thus allowing sequences of actions to be built up such as result of one response can trigger next response in any particular habit / behaviour sequence.
Do we still need a recalled stimulus as a new kind of input now that we have abstract stimuli (habits) to any level? Yes, in order to have thinking – model actual stimulus or abstract stimuli.
6th Jan 1998 Tuesday
Thought from Dec 97. If don’t want it to have good and bad feelings associated with survival related stimuli then feeling of new / different versus similar / familiar could be used from resulting stimuli to reinforce the S-R that preceded it. Thus only drive of robot is to explore – find new info.
14th Jan 1998 Wednesday
What are we conscious of when we learn with reward / punishment.
Stimulus – context / trigger
Feeling of need if not habit / reflex
Response done for unspecific reason – desire to escape
Stimulus result Desired outcome (implies expected)
Good feeling produces feeling of success – reward – stimulus result is innately good.
Stimulus – context / trigger
Expected stimulus result (idea)
Expected feeling of success (idea)
Stimulus result- matches expected or executed stimulus result
Good feeling produced from stimulus >= expected gives success < expected gives failure
Feeling of success
It is possible that the difference between the expected feeling and actual feeling will provide either a success or failure feeling for the sequence.
11th Feb 1998
Learning depends on difference between expected feeling and actual feeling – If difference is less then you have disincentive if more than an incentive.
The expected feeling is tightly coupled with the expected stimulus and this forms the goal. When resulting stimulus experienced it is compared with expected stimulus and feeling. But learning can take place just by comparing expected stimulus and actual resulting stimulus without feelings compared. The difference produces a feeling.
19th Feb 1998
- Make sure every response has an associated body generated resulting stimulus.
- Make feeling of Novel / New / Unfamiliar / different and
Same / similar / familiar explicitly stored.
- Difference between expected feeling and actual feeling is stored on new sequence.
- Perform if resulting feeling is novelty
- Don’t do if resulting feeling is familiar
- Make feelings into stimulus so can be recognized / recalled
When executing 1st sequence while performing 2nd one the feeling from 1st sequence is used to decide whether to execute it or not. But the expectation need not be stored unless we are modelling cognition. Adaptron just used to make decision go / no go. Then in Adaptron while executing 1st sequence when resulting stimulus received the associated feeling received is compared with that stored / being executed. The difference produces the resulting feeling.
1st Time 2nd Time Resulting
Feeling expected Feeling received Feeling
New New New store Trigger + result
New Same New collapse – recognized pair
Same Same New collapse – recognized pair
1 2 3 4 5 6 7 8 9 10
Sa Fn Sb Fn Sz Fn Sa Fs Sb Fs (the Sa Fn Sb Fs) S1 Fn Sd Fn Sb Fs Se Fn
There are # of feelings involved – Any particular stimulus is either
Familiar – I have experienced it before – I have a memory of it. Or
Unfamiliar – Never experienced it before.
Any particular pair of stimuli is either
Same - 1st is familiar (must be) + 2nd is same as expected (success)
New - 1st is familiar (must be) + 2nd is unfamiliar
Different – 1st is familiar (must be) + 2nd is familiar but not expected
Feeling Action taken
Familiar Ff Rs Matches memory – obtain list of memory locations
Don’t store - obtain associated S/R – obtain pair feeling
Based on paired feeling – go / no go concentration level set
Unfamiliar Fn Rn No memory match – interest – flush STM to store and store this
New stimulus stops concentration level zero
Do nothing (attend)
Same Fs Rc Success – matches expected (being executed) – this feeling
occurs 1st before familiar feeling above – Action new compound
New Fn Re Empty STM into LTM as is.
Different Fd Rd Put St – trigger stimulus in LTM and now use the resulting stimulus that was familiar to start executing again.
Can actually - consciously experience the unfamiliar feeling only if concentrating / doing past experience and stimulus received not found.
The unfamiliar feeling triggers reflexive action if not concentrating.
Every sequence is made up of St Fe R Sr Fr R
St – Stimulus Trigger – that which is familiar
Fe – Feeling Expected start or stop execution
R – Response – maybe nothing / empty = Rn
Sr – Stimulus resulting
Fr – Feeling resulting of success / neutral / failure?
When unfamiliar stimulus occurs
Sa Fu Rn Sb Fu Rn Sc Fu Rn
Now if Sa received – its familiar followed by Sb – same – response compound / collapsed Fs = same = success
Sa Ff Rn Sb Fs} Rc s1 Fu Rn
Sd Fu Rn
And Sb familiar feeling – start Rs executing (provided next feeling neutral or positive)
Sb Ff Rs Se Fn Re Se is not expected and unfamiliar do new feeling so B+E put in LTM
Then Sf a filler Sf Fu Rn
Then Sd familiar – start executing expecting a B but get C
Sd Ff Rs Sc Fd Rd Sc is not expected but is familiar so different feeling so store the Sd and keep the Sc and start executing it.
Two stage process if executing –
Expected feeling – Fs collapse
Unexpected feeling – Fn or Fd stop concentrating + stop executing put St in LTM
Then based on Fu or Ff of Sr do Rn or Rs
26th Feb 1998
Habits with actions become successful in repeating past experiences get expected result but are boring and not repeated because will continue to get expected result which is familiar and expected. Now have learnt to get expected result and only repeat it if desire to obtain that result. I.E. goal oriented behaviour. Expected result must be desired. Implies some degree of reasoning given goal stimulus and current situation (stimulus) can identify behaviour sequence that produces that goal.
Past experience Trigger Stimulus St
Feeling bored or familiar Ft
Response done R
Result stimulus Sr
Result feeling Fr
What are possible new feeling resulting from repeating this experience when concentrating.
Past Result Feeling Actual Result Result! If doing Habit
= Expected feeling Feeling (+ve, neutral, -ve) Feelings not important
Unfamiliar Unfamiliar +ve Success/collapse Validates Habit
Familiar -ve Still worth trying again
Expected -ve Environment repeats / predictable
Familiar Unfamiliar +ve Success/collapse
Familiar neutral still try again
Expected -ve Environment repeats / predictable
Expected Not possible (did not execute) Habit is successfully completed
Would not do this unless part of achieving a goal already made into a habit!
If success based on unfamiliarity, an action only becomes successful if it produces unfamiliarity a second time. Then is made into a habit. Chances of it continuing to do this very unlikely in a static environment. We make actions into habits without success based on feelings of unfamiliarity – desire of curiosity.
What does table look like if we have Good/bad feelings.
Past St Ft R Sr Fr
Fr Actual result feeling Result If doing a habit
Good Good validated
Good Unfamiliar new sequence – stop / record
Good Familiar new sequence – stop / record
Good Bad stop /record
Unfamiliar Good Success /collapse
Familiar Good Success/collapse
Exp Good Not possible / 1st time should have been good
3rd March 1998
To be successful + make a habit out of S-R-S the resulting stimulus must be expected (gives predictability to habit) and either resulting feeling = good or resulting feeling improves? I.e. goes from bad to neutral or neutral to good?
May be all is required is to repeat the expected stimulus. We have already decided to do it because it achieved a goal the last time. Now it repeats so it’s worth habitualizing it so can be considered a unit to build on.
When you decide to repeat past experience given trigger stimulus you look at expected resulting stimulus and resulting expected feeling. Don’t need / want any good/bad feeling involved in successful recognition or successful habit doing – only expected used to habitualize and perform.
Feeling used to decide if I do – not used by submind to continue doing habits.
[This is the start of the blue note book]
3rd May 1998 Tests for Consciousness
1. Verbal report of consciousness
Awareness / Unaware
2. Introspection 1st person
Inner-sense – Inner monitoring (internal, self)
Mental states observable behaviour
Consciousness == luminosity – a phenomenon
3. Model others mental states in oneself
Purpose of consciousness?
Properties of consciousness?
Representation (at states) – expectations
Control of action
Testing for learning
Need observable / measurable change in action as a result of experiences. If does it right 1st time every time – no learning. Need definition of right – what is the goal? Does action sequence get shorter and still reach the goal? – Does it get faster? – More parallelism?
Expectations of stimulus / actions – a sequence of them. It can be performed to get expected goals. Thinking rehearses the sequence – when actually done it is being compared with the remembered sequence of expectations. Thinking is mental performance of stimulus / response (or stimulus) pairs. It matches the same expectation if had / remembered before. A thought sequence that has never been done in that order can be done by performing the remembered expectations. So thoughts can be actually performed.
4th May 1998 Thinking
One thought (expectation) can be associated with another thought (expectation). A stimulus can be associated with an expectation. A recalled stimulus is an expectation.
5th May 1998
Intelligence implies intentionality – a purpose trying to be achieved, maybe survival, and maybe pursuit of interesting / new information.
7th May 1998
Thinking is like doing (a habit) just another way of repeating past experiences. But it is never subconscious. All the thoughts go through attention. Can not have a subconscious thought.
9th May 1998
Thinking about hanging a chair upside down from a lampshade / chandelier. Sequence of words results in composite image in short term memory of goal to be achieved. Then think about series of steps / actions to obtain goal. Think about it to a level of detail where know can perform sub-steps known have done befores. Then decide to start doing – so execute steps – each one decomposed and done. Concentrates while doing on parts that are new – might require some detailed problem solving/ thinking about as come across unexpected circumstances in performance of a step. Therefore, need a reverse association capability. If have a goal stimulus – do I know a triggering stimulus which when associated action done results in goal stimulus. Maybe this can be achieved by imagining the goal then thinking of one possible triggering stimulus and recalling its result stimulus – if that does not match goal then think of next possible triggering stimulus. Until goal match caused feeling of desire. How to generate possible triggering stimuli – use current sensory stimuli to obtain associated expected stimuli (thoughts) which are the triggering stimuli.
7th June 1998 Thoughts from last 2 weeks
Only need to work out learning and thinking using simple stimuli / ideas because the complexity of stimuli handled by hierarchical recognition / performance. So work out S-R-Feeling straightforward learning and it will handle all degrees of complexity.
Should have a good / bad feeling which learning is based on – and only based on. Then if new / familiar is to be used as learning criteria then novelty and boredom should produce these good / bad feelings as well as stimuli themselves.
Iris maintenance behaviour must have feedback that is interpreted / measured as bad – too bright & not enough light and good – just right level. I.e. negative feedback but also change in light level might be interpreted as reward not absolute level.
Learning contains an expectation of resulting stimulus and then expectation of resulting feeling that is used to decide to repeat or not the past experience. – Provided these can attract attention – concentration level not too high. Maybe if concentrating such thoughts don’t interrupt attention.
10th June 1998
Response – action
Idea of resulting stimulus (expected) ------------------]
Idea of resulting feeling (expected) ]> Achieve a goal
No/Go decision ]>
Go -> response – action delegated to submind ------]
Goal idea – it’s good
Recalled habit to achieve it – do it
Recall & do habit
11th June 1998
Feeling New -> Good
Feeling New -> Good
[ Stimulus A
[ Feeling familiar – expected Stimulus B – Good idea
[ Expected feeling – New -> good idea
[ Do habit after A
[ Stimulus B
[ Feeling familiar – feeling ‘was expected’ – matches executable – success habit – good
--- [ bracketed piece -> Habit 1
Feeling New -> good
Feeling familiar – expected stimulus B – good idea
Expected feeling New -> good idea
If goal / result stimulus not equal to expected then habit interrupted
Feeling New -> good
Feeling same as STM => bored -> BAD
Do response X random
Result stimulus B
Feeling NEW -> good
[ Stimulus A
[ Resulting stimulus B – Goal idea
[ Feeling familiar – resulting feeling NEW -> good idea
[ Do habit after A
[ Response X
[ Result stimulus B
[ Feeling familiar – feeling ‘was expected’ – matches executable – successful habit – good
[ - so collapse into habit – identify it
----[ bracketed part -> Habit 1
Feeling New -> good
Feel New -> Like
Feel New -> Like
Stimulus A Familiar
Expected feel New -> Like idea or
Feel New idea -> like (No because no need to learn what to do when this idea stimulus occurs)
Trigger stimulus A
Idea of habit done
Idea of resulting stimulus
Idea of resulting feeling These three ‘idea of’s = mental model
Get expected stimulus = resulting stimulus C
Get expected feeling = resulting feeling
Commit it to habit ABC
Later on, idea of resulting stimulus can trigger this habit without trigger stimulus.
11th June 1998
Thinking is sequence of ideas – expected stimuli in order that corresponds to experienced stimuli – so can simulate the reality before doing it. So the simulation should repeat the actual until an appropriate good feeling occurs which then makes you decide to start doing it.
Timing of coincidental events A then B recognized as AB new habit but if close enough in time B will trigger expectation of A and thus instigate habit. So A is trigger stimulus and B is expected Goal – Just the expected Goal might trigger habit even if not in context of A happening.
15th June 1998
Thinking is a series of recalled stimuli that are processed just like any other stimuli – recognized and combined to form higher level ideas. When performed the ideas are broken down into components just like doing response (habit) except not done by sub mind – all ideas are conscious.
22nd June 1998
Learning from shock treatment of mouse – jumps wall to safety – learns to jump before shock if preceded by light each time. Goal is to seek positive change in feelings. Needs an internal state such that the difference from last feeling to this feeling is what is recognized as good / bad
2nd July 1998
Learning is driven by the feeling of worthiness of being learnt – Familiar and novelty can be used for this purpose to learn to recognize and to learn to do but also good / bad feelings can be used for this purpose.
Drive is a phenomena – really is avoidance of bad stimulus e.g. hunger stimulus.
2nd July 1998 Short Term Memory
STM used to recognize and collapse familiar sequences is STM of subconscious and it only notifies / interrupts consciousness when novelty occurs. Then only if subconscious runs out of memory (off end) does it give familiar stimulus to consciousness which then is interpreted as boring. Boredom is conscious familiarity.
5th July 1998 Expectations -> Ideas
Same sequence with thoughts added = T
At 5 do we really have success or are we not doing because @ 4 B was new? If so @ 5 the B should be familiar and generate an expectation of an A.
5th July 1998 Punish / Reward
Need a mood indicator that can be compared with feelings from latest stimulus to obtain good/bad feeling. Mood is either happy, neutral, sad. X Stimulus is interpreted as bad. B. Y Stimulus is interpreted as good. G. All other stimuli are neutral.
Maybe we have moods of happy H, interested I, bored B, sad S. No it seems more like
There is the stimulus, Its feeling (Novel, familiar, good, neither, bad) and Its reflexive response (if any).
8th July 1998 Punishment – Avoidance Learning
Cat in box tries to escape. Being in the box (context) combined with desirable goal (out of box) results in attempts to escape – tries all sorts of things – one random response results in freedom. Is it same as: Mouse is in box, hears tone (context) combined with shocks desired to avoid them? It attempts to escape shocks tries all sorts of things – one response results in no more shocks. Intensity of undesirable stimulus is greater in shocks and probably triggers more reflexive primitive responses. Sequence 1st time is:
Context (inbox, tone)
Bad stimulus (shock, no freedom)
Random responses (reflexive)
Repeats above if no change.
Non Bad stimulus (out of box, no shocks)
2nd time is: (ideal)
Context (inbox, tone)
Repeat response (last one done)
No bad stimulus.
The bad stimulus – although it generates pain it is not associated with any response because it simply follows stimuli of context. However if loop repeats because random response is not successful the random response is followed by a repeat of bad stimulus. But does the second bad stimulus make you sadder? No already sad. Bad stimulus continues to trigger random responses but has no effect on mood.
2nd time alternative is:
Context (in box, tone)
Bad Stimulus (shock, no freedom)
Repeat response (last one done)
Non bad stimulus.
3rd time is:
Non bad stimulus
Punishment / Reward
At 7 it gets another A, B=1 which is familiar @3, 4 but expected results are either F (boredom at 4) (If Listen) or B (Sad at 5)(If do an a). So subconscious interprets conscious with the 1 and feeling of familiar which results in reflex action.
9th July 1998 Expectation of Feeling – Avoid Punishment revisited
1st Alternative to bottom of Page 11
When B at 7 received and we were doing recognition of A, B at 1, 2 the B is not only found to be familiar but also has an expected bad feeling from the resulting action so perform a reflexive response ‘c’ in order to avoid the X from occurring. If the sequence was slightly different and the A at 6 had not been received we would not be doing.
We would not start doing at 2 because B expected next stimulus generates bad feeling. Do we just not do anything? I.e. Listen or do we do something other than listen? This is all based on the expected bad feeling.
No expectation of punishment
If have no expected bad then will go ahead and repeat the listen operation of previous B as in bottom of page 11.
Without any other response than listening when 5 recognized we continue to repeat 5 X -> 8 until 8 repeats and we are bored and do something reflexively as a result. Either we need a random reflexive response whenever a bad feeling occurs which means don’t reuse experienced successful responses to bad stimulus. Or we need to respond as on page 13 to the expectation of the bad feeling. All animals have fear! Simpler example of above scenario.
1 2 3 4 5
Stimulus A X A X A
Feel N B F B ?
Resp L a L a
Recognition & Responses
From Neural net perspective:
Novel / Familiar produces a neuron that recognizes a pair of stimuli
When repeated again bored
Response hierarchy should look like (possibly)
Responses – feedback
Every response has a consistent kinaesthetic feedback stimulus so hierarchy’s of responses can be built.
When doing we are repeating a sequence we have done before trying to validate that it is repeatable predictable -–The sequence is
A response (if any)
Using feelings to decide if we should repeat the doing! We try the no response 1st before trying the something response because no response does not use any energy. If no response results in no novel stimulus then we try a something response. The sequence should be
A response (if any)
A Stimulus }
A response (if any) } A high level stimulus
A stimulus }
Maybe only create a higher level stimulus if the response is no response. If a response is something then maybe can’t collapse it into higher level stimulus. (See July 2nd 99)
Maybe something response, listen, result stimulus gets collapsed into high level response if stimulus is familiar. High level response has no context stimulus. If result stimulus is familiar we are verifying the response produces this stimulus.
Don’t habitually repeat (as part of recognition) something responses which last time produced a familiar resulting stimulus. Only do it again (repeat it) if consciously decide to do it due to desire to or it produced a novel resulting stimulus.
14th July 1998 Orienting Response
Orienting responses are learned with the aid of novel / familiar feedback – no good / bad feeling are involved.
21st July 1998 Words / ideas
Words are habit responses associated with ideas. They create a sound that is the resulting stimulus and the expectation of this sound is what we use when we can hear our inner voice.
22nd July 1998 Habits
When a familiar (no associated good / bad) stimulus occurs it should trigger a habit performance. The habit then performs the response. Unless there is no resulting stimulus because the place in memory where executing is end of memory or the resulting stimulus is a new stimulus or the resulting stimulus is familiar but has never occurred after the triggering stimulus and performed response. If it has occurred before in these circumstances then habit shifts to this familiar experience and continues. Habit also terminates if good or bad stimulus is received or it comes across a good or bad (or expected bad) feeling in a familiar experience it is about to switch to.
28th August 1998 Generalization / Specialization (discrimination)
In Adaptron each stimulus could be a number of letters each of which is a feature experienced at the same time rather than one after another.
A sensor’s first level of innate processing of stimulus determines the maximum degree of generalization possible. All learnt stimuli from then on are more discriminated. Generalization is only achievable via thoughts through identification of common features and associating responses to features. We learn to specialize based on one feature e.g. colour but also learn to generalize (realize the irrelevancy of) other features e.g. size. (Also based on expected features i.e. expected feedback stimulus)
We remain executing a habit as long as the context stimuli remain familiar i.e. continue to drive a car as long as context (the road & car) remain. But while doing habit we trigger a variety of sub-habits based on events that happen – brake lights come on. So habits need context stimuli & triggering stimuli to perform themselves. They also require feedback stimuli to continue executing. Why are feedback stimuli not confused with triggering stimuli and / or context stimuli?
Similarity / difference
A degree of similarity is important for learning such that transfer of learning can take place. Provides a list of possible responses that can be tried in a new situation based on its similarity to experienced situations. Else every situation is new and has to be learnt separately.
Expected feedback stimulus of the pressure on your bum is the feature that determines if any object belongs to (generalized into) the class of “chair”.
Context stimuli provide continuous unchanging features that must be present to be doing habit. Trigger stimulus is any stimulus that changes – could be expected and causes appropriate sub habit to be done or could be unexpected in this context providing learning opportunity.
Timing is a stimulus that beats regularly and incoming stimuli synchronize with it causing the time between stimuli to be a stimulus itself. So time between stimuli is a stimulus.
2nd time dimension feature used to determine distinguish (similarity) between 2 stimuli is the direction and distance of the values of the stimuli. I.e. 1st note is 266 Hz and the 2nd note is 276 Hz. The difference is a 10 Hz increase – a 10 Hz increase is a similar (feature) which is similar to the 2 stimuli 344 Hz and 354 Hz.
Use a tone environment – map it onto the numbers 0 to 9 and use a similarity finding algorithm based on timing and / or pitch differences between 2 notes to generate new stimuli and have transfer learning.
1st Sept. 1998
Need two or more sensors in Adaptron so it can switch between sensors so that when it listens it is either doing a listen or it is reflexively listening. An orienting response is an overt paying of attention and it should be possible to learn it.
Need a response marker that says at this point the habit that corresponds to the triggering stimulus in past memory was done. Thus the delegation to sub-mind is marked and the mind can carry on with a different “train of thought”.
27th Sept 1998
Timing between stimuli is also a stimulus as documented Aug 28th. I could implement it as a counter that starts counting up, starting at zero when an external stimulus occurs. The rate of counting would be at the smallest determinable / distinguishable interval between ticks / counts. When the next stimulus occurs, the value of the counter is also a stimulus. Repeated values of the counters are recognized. Since this stimulus is quantified direction and distance between any 2 counting stimulus can be a stimulus itself. This counter may slow down as time elapses such that 1 minute and 1 minute and 1 second between stimuli are not distinguishable but half second and 1 second between stimuli is a significant difference.
2nd Oct 1998
Stages in learning Habits / Recognition
-New stimulus – never experienced before.
- Familiar – but has not been habituated – has not been confirmed so concentrate on and confirm – if confirmed record as habit – higher level stimulus.
- Familiar – Has been habitualized so sub-mind performs the habit subconsciously – as long as remains familiar and habitualized stay subconscious.
Just A’s received and sub-mind:
Fc = Familiar – concentrate on, U = Unfamiliar
E = Expected – successful recognition
- collapse to new stimulus
Fh = Familiar and have habit so sub-mind recognition takes over
Note: at 2 an A was recognized at 1 as Fc even though A at 2 had just been input.
So A at 2 put in STM and 2 in LTM before recall so can do at 1.
At 3 an A is recognized and an Fh goes subconscious. A not stored at 3 subconscious starts executing at 1 expecting a 1 – as long as it gets an A next all is well – it disregards any number of A’s.
But the conscious mind is going to get bored – no stimulus coming in. How do we define boredom now?
It keeps on expecting at 1 and expecting what is at 2 – a 1 – expectation does not change - Bored.
At 4 sub-mind interrupts mind with Familiar 1, Fc stimulus but not habitualized so start doing at 3 expecting a 1 as in 4. Another AB produces a 1 which was as expected in 4. So 4&5 collapses into:
1 2 3 4
I A B 1 3
F U U U U
O L L L L
Now if A, B continues we have same situation as previous page. Boredom occurs, sub-mind detects repetition?? Or as on page 7 (2nd July 98), it runs off of end of memory and returns a familiar stimulus to mind which is then bored. Pg 7 says bored occurs when you have conscious familiarity!!!
Generalizing A as input and it is familiar and it has been habitualized. A is familiar at location 6, 23, 49. 6 and 49 have been validated, repeated and 6 and 49 exist as higher level stimuli so sub-mind adds to its list of places where it is performing habits 6 and 49. But mind uses 23 as familiar location which has yet to validated / confirmed. So does it – starts executing at 23? Next stimulus may either 1/ be as expected to produce a 6 or 49 so sub-mind will deal with it or 2/ it validates at 23 and that produces a new stimulus 23 or 3/ is a new stimulus and is recorded in LTM. If 1/ the sub-mind matches one of 6 or 49 then recalls a 6 or 49 to see if they are familiar and habitualized. If a 6 or 49 is habitualized it carries on at the 6 or 49 location. If the 6 or 49 is not habitualized then it notifies mind of a familiarity that needs conscious validation.
From pg16 (9th July 98 Responses - Feedback) – feedback from all outputs and pg23 (1st Sept 98) a 2nd sense – the feedback stimuli from responses should be coming in on a 2nd sense – then orienting listening responses can be included in the model and learnt.
5th Oct 1998 Monday
It might be simpler to have the sub-mind recognize all familiar things and only have it interrupt mind when it either receives an unfamiliar stimulus or an unfamiliar pair of stimuli. This means the mind would not concentrate on a once only familiar stimulus in order to validate / verify it before the sub-mind will disregard it.
So if A is familiar at 6, 23 and 49 then the sub-mind starts doing at all 3 locations. Next stimulus – it may match 24 and cause a new 23 stimulus to interrupt mind or it may match at 6 and 49 and sub-mind carries on. Or it may be a completely new stimulus which interrupts mind just like 23/24 did.
Let us say next stimulus is a B. If it matches at 6 but not 23 or 49 but also at 35 then we have a habit at 6/7 that sub-mind can follow which has a context and also an out of context B at 35. Sub-mind should be doing both but execute the one with greatest context (depth, history) before the most recent. If context depth is equal amount recalls then do the most recent one first.
Let us say sub-mind is deep inside a context, 4 layers deep and next stimulus is unexpected. But part of last stimulus series matches another experience. Sub-mind should automatically shift over to it. Only if no other series would sub-mind interrupt mind with new experience and context. E.g. has habit for ABCDEFGHI and has habit for EFGJK and it has received ABCDEFG and then receives J. Sub-mind should make smooth transition to second experience. Or should mind be interrupted with new ABCDEFGJ.
What do they look like if collapsed?
It is important to do / execute at 28 rather than 35 because responses might be involved.
15th Oct 1998 Portugal
The execution loop for Adaptron is like:
- Obtain all stimuli from all senses and features.
- Recall all of them as to familiar / unfamiliar / good / bad and where found in memory.
- For all habits if any stimuli are expected then process appropriate stimuli and produce appropriate response. Some habits may collapse and high level stimulus needs recall.
- Failed /succeeded produces additional stimuli or flag existing stimuli as unfamiliar / familiar.
- Conscious attention – which stimulus attracts attention – interrupts concentration.
- Store this stimulus.
- Perform any response – maybe reflex, maybe doing past experience, results in kinaesthetic stimulus.
- Go to top of loop.
Eating with knife and fork is a habit. It includes the sub-habits / paths / experiences to pick up cutlery, stab food, put in mouth and chew / swallow. Each path is done by sub-mind without any mind intervention. Whole habit instigated at the beginning. Sub-mind decides if it needs to include path in performance if triggering stimulus present and paths include orienting responses to locate such triggering stimuli. Habit only stops and interrupts consciousness when it has a stimulus which it does not recognize in this context (an unexpected (triggering) stimulus). It is an unfamiliar stimulus to habit even though it may be familiar to mind. Mind can handle it and make decision of what to do and then continue habit – how?
A habit is complete when its STM collapses to the stimulus that is the expected resulting stimulus of the triggering stimulus that caused habit to be given to sub-mind. E.g. have experience of
S1 S2 S1 is the triggering stimulus.
R1 R1 Response done
F U S2 expected resulting stimulus
When S1 received it triggers start of sub-mind performing S2 recognition (after R1 done) When sub-mind succeeds in recognizing S2 then habit ends and S2 is available for conscious attention with familiar feeling and expected. It could become the trigger stimulus for the next habit. If mind is still waiting for sub-mind to recognize S2 then S1 S2 is collapsed into aa new habit. Mind would have to be concentrating on this S1 S2 sequence. If the mind is not concentrating, S2 could come back in the middle of some other conscious sequence such as a thinking sequence. If S1 is still in mind’s STM maybe S1S2 can be collapsed into a habit.
If mind started sub-mind on S2 but not at a concentration level (just something to do) and S2 is recognized by sub-mind then this does not form new collapsed habit, just a familiar resulting stimulus. If S3 comes back instead and S3 is unfamiliar then it attracts attention away (concentration interrupted). If S3 is familiar but unexpected then the pair S1 S3 is unfamiliar.
Do we need to have these feeling combinations? Expected and Familiar (collapsed if concentrating), Unexpected and Familiar = Novel, Unexpected and Unfamiliar?
Idea! Expected and Familiar collapse if did listen at any concentration level. But it only collapses if did response if at level>= 1 concentration. Novel is pair of stimuli is new even though resulting stimulus is familiar. Want to do it again at concentration level 1. Unfamiliar is always desired to repeat at level 1. With Expected and Familiar may end up with S1 S2 that did not collapse. When sub-mind is doing these it will try to collapse (submind only given one stimulus to do) and find no higher level stimuli, so maybe it says if S2 is familiar why not perform at location S2 found to be similar with the S1 S2 location most likely because it agrees with context (S1). Now how does a series of A’s get interpreted??
# 1 2 3
I A A A
F U Fc ?
O L L ?
31st Oct 1998 On way to Taipei
Algorithm for sub-mind. Have S1 S2 as on pg 31(15th Oct) and S1 received – conscious – decide to do. Sub-mind issues response and then does S2. If S2 is external stimulus it is waiting for external input cycle to see if stimulus received. If S2 is high level stimulus then sub mind is responsible for recognizing it. There is only one possible tree of execution to recognize it. So it needs to call itself recursively to
– get left stimulus
- output response if any
- get right stimulus.
If left or right stimulus is high level it needs to call itself on that stimulus. If left or right stimulus is an external one then it waits for input cycle. If sub-mind checks input and it is not expected it fails. The entire ‘stack’ of recognized stimuli and responses is then available as another stimulus for conscious attention. Whether it becomes conscious depends on concentration level.
7th Nov 1998
Maybe should store feeling of novelty and expected feeling of novelty just like expected good feeling. This would propagate backwards in the sequence chain and trigger behaviour. Recalled expected feeling of novelty is still expected feeling of novelty. What concentration level? Recalled feeling of familiarity just drops concentration level a notch. Recalled feeling of bad drops concentration level to 0 and stops the execution of the sequence because want to avoid repeat of past.
Simultaneous versus Sequential
When three stimuli occur too close together in time and stimulus 1 and 2 have been recognized sequentially as a higher level stimulus it is possible the 1,2,3 is new compound stimulus? – Let’s make the question simpler. 1 and 2 have been habitualized when occurring in sequence. Now they occur simultaneously without the ability to tell if it was 2 then 1 or 1 then 2 (so close – inside resolution of senses). Do we need the ability to recognize the pair with no order implied and it should have a partial match with a single 1 or a single 2 – giving us generalization so that a 1 or a 2 triggers the habit that the 1-2 triggered?
23rd Nov 1998
Trying to sort out difference between habits being done that include responses which have to complete by successfully receiving all resulting feedback stimuli and the automatic recognition habit of stimuli where no responses involved. It appears recognition habit is continuous bottom up process of given any stimulus immediately asking is it expected by any habit and these habits are not initiated like the ones involving responses. It’s an automatic background pattern matching finding all matches and places where it is expected. Collapsing all expected ones and taking these high level stimuli and doing the same again. However a habit involving a response must be first consciously initiated and then it executes to completion (or failure). While it is executing it is not just one exact sequence / hierarchy of stimuli / response /stimuli. It may recognize a variety of triggering stimuli and perform sub-habits appropriately without conscious decision / desire to initiate.
Thoughts on how to program this include: I get an S10. First was it expected, i.e. the resulting stimulus of a habit – if so collapse the habit to higher level and now have another S# so can repeat this loop. Second each place it matches in memory can be checked if it has been habitualized and if so then add them to the list of things being recognized. Or each time an S is received first it is used to collapse any recognized habits to higher levels and then second it is used as a trigger stimulus of other habits that should be done based on matches that have been habitualized successfully. This would mean successful habits that include responses would have to meet stringent success criteria so can be initiated subconsciously. Recognition only habits need not meet these criteria.
24th Nov 1998 S & R-Habits
So have two types of habits – S-habits and R-Habits. S-habits are stimulus only with no interspersed responses. These are being done automatically without the need to be initiated. R-habits include responses and only are done when consciously started (because there is the expectation of success) or when they exist and have proven successful in the past.