General, non-contextual, direct (one-step) Connections –

Considering the nature of and the role played by connections, connections are at the heart and soul of any intellectual or mental activity. So we need to define connections.

As a real, cognitive phenomenon, what is a connection? It relates to immediacy of thought-occurrence, in the mind, of something, from something. If this is present, there is a connection. Also, we are talking about direct one-step connections (hence the immediacy). (‘John – surgeon’ is a two-step connection since John’s mother is a surgeon. John is connected to his mother (one-step) and his mother to surgeon (one-step)).   

So we need to study the game of – if one thing is said, what other thing immediately comes to your mind? (Like the rapid-fire rounds in TV shows).

So we need to know about storage of knowledge in the mind, where these immediacies stem form. Lets look into this. 

We think we “know” a lot. So I would like to segregate all that we apparently “know” into 2 parts. Before coming to those 2 parts, lets us first see an example.

Suppose someone is teaching you about and to drive a car. The first thing he tells you is – A car cannot run without petrol. If there is no petrol, the car cannot move ahead. Later, on going to a petrol -pump, he tells you – here you fill petrol in the car. So I learn 2 things about cars – ‘a car needs petrol to run’ AND ‘you fill petrol in the car i.e. there is something like “filling petrol” into the car’. These are 2 things you have learned, and now you know them.

Suppose on some occasion later you happen to tell someone else, say a friend of yours – Hey, if you don’t fill petrol, how will the car run? Now, it appears from this sentence that you also “know” – if you don’t fill petrol into a car, the car cant run. But no one told you this, as it is; you have combined the 2 pieces of knowledge learned, mentioned above. This is actually thinking – (but) you have COMMONSENSICALLY, and hence, swiftly and easily, combined the “need of petrol for running” with the “filling (the concept of filling)” from the 2 pieces respectively to create this new statement which appears to be something you directly know. 

So, coming to the 2 kinds of knowledge,  the first kind of knowledge is what is impressed upon memory, as it is, in a certain given format, while learning about something (like the 2 pieces of knowledge learned, mentioned before). Hence, it is that which was perceived by us, as it is, while learning about something. And the second kind of knowledge is what happens to be easily and swiftly combined (since it is commonsensically so), from those of the 1st kind, via “thinking”. The latter is so easy and rapid that it creates the impression that we know it (and for all practical purposes, can also be said to be something we know.)

Lets call the first type as type-A knowledge-piece and the second type as type-B knowledge-piece.

Thus, the immediacy of thought, that we talked about in the very beginning, is the commonsensical combination of the literal memory-impressed data while learning.

So, coming back to the attempt of defining connections, 2 entities are connected if there exists a type-B knowledge-piece combining the 2 entities, stemming from the type-A knowledge-pieces about the 2 entities, (or, obviously, if the type-A knowledge pieces stemming from them directly coincide). 

(Otherwise, anything can be connected to anything. The above explains why a rat is not connected to the International space station (ISS), though some distant far-fetched connection can be drawn between the 2, just for the technical heck of it.)

Going from the Subjective Emotional to the Objective

We hate breaking subjective emotional things into their objective parts. We do so for the fear of losing the fun, the charm, the mystery and the beauty of the phenomenon. In fact, we don’t want to explain the beautiful and inexplicable in terms of its “constituent” parts. 

Consider, say, Love. We are enamoured and swept by love. But if someone asks a person – why exactly do you love so and so person? ENLIST the points. Lets enumerate it as a 15-point love or a 23-point love or ….. how much ever. The immediate response would be – Yuck! Don’t spoil the beauty, the fun. How can you break something beautiful and emotional like love into points and parts? 

Whenever something works, and we don’t know why and how it works, there is a sense of impish charm and beauty associated with it. For example, How come this batsman can hit this shot so elegantly and effortlessly? How could this Mathematician have come up with such a formula or a theorem? Why does this theorem work even though it’s not yet proved?…etc.

One possible explanation for the reaction in the above example (the enumeration of love points) is that if you break the “unknown” into the “known” then you know what exactly, and the ways in which, it can get damaged, and hence you might lose it! As long as its unknown and mysterious, it is somehow “protected” and “preserved”. But one principle does seem to work – if the size is very small or very large, then enumeration and enlisting does seem to make sense. If someone is a giant scientist, then we want to know the list of his achievements, however long. If something is a trivial emotion, we want to spell out the couple of components of it. 

Anyways, the point is something else. Consider Fear. Suppose there is some big, grand, cloudy phenomenon (say a person) which is the cause of your fear and is giving you sleepless nights. Here, if you break this fear into its constituent parts – by saying things like say, “OK, what am I afraid by? What exactly? Is it the face of the person? Is it because of that specific last month’s incident? Is it….. What can be done about the second and the third points? We can fix the second one easily with a ……. Here, this is indeed a positive exercise and this splitting of a subjective emotional phenomenon into constituent points and parts only helps us and solves our problem.

The process in case of Fear turns out to be exactly opposite to that in case of Love.

So, it appears as if it is not about breaking the subjective emotional into the objective at all. If things go pleasurably, there is joy, and it’s “on”! If otherwise, it’s “Yuck!”! 

A theory of Happiness

The ‘plus (+) theory’ –

Suppose hurting your brother makes you happy. So its a ‘+’. So you hurt him. You become happy. Then suppose someone teaches you how it is wrong to hurt him by giving you a philosophy / a theory, knowing and accepting which gives you a bigger happiness (a bigger ‘+’) than hurting the brother. Then the sign of ‘hurting brother’ would become ‘-‘ (negative) and it can become a ‘+’ only if I attach another ‘-‘ to it which would mean that I give it up / stop it. So I stop hurting him and move towards the bigger ‘+’ of following the philosophy. 

At any time, we are either only staying at / maintaining a ‘+’ (this includes converting minuses to pluses) or moving from one ‘+’ to a bigger ‘+’. 

Also this shows that ‘unless you get a ‘+’ for something you won’t do it’. We do anything because we think it will make us happy i.e. it maintains the ‘+’ (including converting minuses to pluses) or is a bigger ‘+’ than before. 

So, what is happiness? The reason for doing anything. Happiness is – the reason for doing anything. That happiness might just be a mental conceptualisation – pre-action and hence pre-result – and hence might be ‘wrong’ also as an error/mistake in thinking (it is thinking/a thought since its a mental conceptualisation).

The ‘happiness’ mentioned above could be a happiness (the usual usage) or joy or something which saves you from trouble or gives you peace or protects you or whatever……All these are pluses. 

The causal precursor to a plus, is a ‘want’.

Experiencing a plus, is ‘liking’. 

A commonsensical guide to commonsense reasoning in pronoun resolution problems

Consider this example – John banged his head on the rock. It started bleeding.

Here, I attempt to provide a commonsensical guide to the pertinent commonsense-processing (and of the data as a whole), as against a logical one. There is clearly a logical way of deciphering the pronoun – it is a pronoun; stands for noun; 2 nouns – head or rock; rocks cannot bleed; so head. Here is a more commonsensical account of the deciphering, taking into consideration that the mind is already impacted at the end of having consumed the first sentence and has a partial expectant idea of what is about to come, rather than just plain hear the 2 sentences, be absolutely clueless about the ‘it’ and then follow the logical path mentioned above, purely backwards. 

“Started” => wasn’t bleeding before => change of state

Change of state of whose? What’s been affected? If you bang your head on a rock, will your head be affected or the rock? Commonsense says – head. 

(Here, in fact the rock is also affected – it gets “blood-wetted”. So in a certain very loose sense, it can also be said that the rock started bleeding.

This possibility comes in, in the above guided reasoning-module at the stage of asking what is affected (head or rock?), to which one may also say that the rock is also “affected”). 

This module provides a commonsensical explanation for even those pronoun resolution cases which are not clear cut “black or white” cases like the purely commonsensical ones like the one above, but are like the one below – 

Bill pushed John. He fell down.

“fell down” => entity was upright before.

So there is a change of state.

Whose has changed? What’s been affected? The one who is pushed or the one who is pushing? One who is pushed is, obviously, mostly affected.


The reason I use the phrase ‘black or white cases’ is because in the head and rock case, rocks just can’t bleed – brute commonsense. But in the case of Bill having pushed John, there is the more unusual a (logical) possibility of the person pushing falling down (for various possible reasons).

If you insert a pin in a carrot there will be a hole in it. (Black and white case)

The elephant stepped on the dog. It died. (Grey case).

……..

A point – Commonsense Knowledge-base pieces

Commonsense Knowledge bases contain pieces of knowledge like – 

If you take a cab ride, you will have to pay fare at the end of it.

This is not actually a piece of knowledge; it is a thought. It is an observation. But all thought ultimately is, or rather becomes, knowledge. Hence it is knowledge – in that sense. The thought that nothing can travel faster than the speed of light first occurred to Einstein. Then, it became knowledge. All thought is first a thought and eventually knowledge. 

Once you have the basic definitive knowledge of cab, ride and fare, you need not know the above piece of “knowledge” separately again. It is clearly a thought, overlapping 3 concepts. The knowledge lies in the descriptive definitions of the terms in the piece; the particular linguistic arrangement (If you take a cab ride, you will have to pay fare at the end of it) of the contents from the 3 pools is the thinking – a connecting thought (perhaps tautological).

(Note : in this example, the basic definitive knowledge of a cab itself is actually sufficiently a source of the given “knowledge” piece. But in other cases of commonsense, like say, ‘Countries have leaders’, the basic definitive knowledges of the terms combine to yield the given piece.)

A strategy for spotting relevant commonsensical things – inferences, observations, general thoughts – in anything

  • Consider any change 
  • Mark the system
  • Mark the surroundings  
  • See the changes brought about in the surroundings (initial and final) of the system. This is where the relevant inferences/observations/…etc. will lie.

This is a general scheme for hovering around relevance, in any scene.

Examples : 

1) If I give you my book, it won’t remain with me. 

Change – book given.

System – Book 

The surrounding is your hand (palm) which is now empty. This (the underlined part) are the relevant inferences/observations/…etc. in this scene.

2) Heating a solution in a beaker changes its colour. Beaker turns warm. 

Change – solution changes colour

System – solution

The beaker is the surroundings, which is now warm. This (the underlined part) are the relevant inferences/observations/…etc. in this scene.

3) Cat jumps out of the bucket and there is a mouse nearby outside. Bucket turns empty / free for use, and the mouse gets eaten.

Change – cat jumps out

System – cat

Surroundings – The bucket is a surrounding which becomes empty / free for use, and the mouse is a surrounding which gets eaten. These (the underlined parts) are the relevant inferences/observations/…etc. in this scene.* NOTE : The changes in the system i.e. what you are considering, are very obvious – you see them directly in front of you in what you are considering. Hence they arent mentioned in the strategy. (Here, the systems are – book, solution and cat).

Addressing the robustness and flexibility aspect of Commonsense Knowledge Bases

Commonsense Knowledge Bases contain facts known to everyone, like say – Lemons are sour. But is this knowledge flexibly robust in the machine?

Here is a simple scheme (with its philosophy) to make them so :  

What are the Wh-words in language?

Who, When, Where, What, Which, How, Why, Whom, Whose.

  • Take a piece of commonsense entry in the KB. Say, ‘lemons are sour’. Now take various perspectives on it by asking all sorts of Wh-questions to it. (This is machine-generatable). This will generate 2 kinds of questions, one kind being very special (we shall soon see). 

So, firstly, this whole set of questions will be – 

  1. Who – INVALID.
  2. When do lemons taste sour?
  3. Where do lemons taste sour?
  4. What lemons taste sour?
  5. Which lemons taste sour?
  6. How do lemons taste sour?
  7. Why do lemons taste sour?
  8. Whom do lemons taste sour to?
  9. Whose lemons taste sour?

So we have 2 kinds of questions – the ones in blue and the ones in red – each primarily serving a purpose : 

  1. The blue questions – They help the machine understand a fact from different points of view – from the points of view of some basic allied aspects to the fact.
  2. The red ones – What is the nature of these questions? They are all grammatically perfect and hence technically sound. But commonsensical-ly they are weird and a joke (e.g. – whose lemons taste sour?). It seems like someone is unnecessarily trying to be logical and scientific about something very simple and obvious (commonsense). One would say – what do you mean by whose lemons taste sour? Irrespective of whose lemons they are they are produced in farms and they bear a sour taste. What do you mean by whom do lemons taste sour to? To everyone (every human) who tastes them. These seemingly tall questions can be countered and cracked by sheer commonsense if the commonsense is well in place.

Now consider this – Commonsense facts are mostly unchallengeable-y well-known and well-understood. Something like  lemons are sour. But think for a while – when can the validity of this sure and solid piece be challenged? When there are logically and scientifically “weird” questions attacking them. You doubt the basics; you become unsure of the very obvious via say some Cognitive Psychological state or phenomenon. It is these very weird questions which the red ones amongst those generated in the above exercise are! So, if the system can stand competent against these attacks, we can truly say that it is flexible and robust – that too, in a Cognitive sense.

  • Answer these questions either via a web search engine (blue) or manually (red). Note  that being a commonsense knowledge base, the blue ones would be answered at a very basic level. For example, a blue question like ‘why is the sky blue?’ in the set of questions to the fact – the sky is blue – would be answered as something like – something to do with light’s properties (and not the actual answer). 

Maybe the manual ones (red) will noticeably outnumber the blue ones. 


2 small ancillary purposes of the question sets: 

  1. The red questions test the flexibility and faith of commonsense thinking, considering the (unduly) twisted nature of theirs.
  2. A machine which is completely logical will face (and self-generate) such logical (grammatically, combinatorially possibilities/cases-like) questions like the ones in red. So  it better have answers to such questions also!

  • The exercise can be iterated over the answers of the questions also, for breadth and depth.

Cognitive Knowledge-Bases of Commonsense

Cognitive Knowledge-Bases of Commonsense. 

This essentially pertains to the case of commonsense knowledge acquired as kids, about the world around us. 

KBs contain statements like – A microwave is a kind of a kitchen appliance. Hearts are inside the body. You cannot throw an apple onto the moon …. etc. 

Consider the first one, about microwaves. The question is – Were we taught in this format? This piece of knowledge is introducing a microwave oven to someone. You don’t teach a kid in this format, that “listen John, a microwave is a kind of a kitchen appliance“. It is not a scholastic class that is going on. 

Yes, in case of something specific, exact and technical which you would want to tell/teach the kid, you would tell something like ‘when you want to convert litres to ml (millilitres), multiply it by 1000’. 

Let us now delve into the domain of microwave ovens (commonsense knowledge about them). 

We need “fragments” of commonsense knowledge, and not Synthetic “made” and “formal” sentences of the same, for commonsense thinking – for commonsense thinking to be integrated with commonsense knowledge. Commonsense knowledge pieces should be like the fuzzy and mixed information/data shown in movie trailers – short 2-3 second scenes, bearing the ‘content’. Commonsense knowledge should be in that ‘format’. 

While reasoning about microwaves or doing reasoning which involves a microwave, the commonsense knowledge comes through in that fashion. 

There are 2 points here – 

1. Sub-processes : Your knowledge about microwaves is in the form of random and patchy sub-processes involved in regard to using/experiencing a microwave. 

2. Analogue and Equivalence : Instead of the synthetic statement – ‘The container rotates for the duration of the entered time’, a more of a “cognitive” statement would be that one-second 

mental snapshot of ‘when times up, the rotation stops’. This is equivalent to the synthetic statement, but more “real”. Also, this more relates to the way in which that commonsense was experienced and procured – and hence ‘learned’. 

Hence, the “REAL” knowledge corresponding to ‘A Microwave is a kind of a kitchen-appliance’ is – ‘A microwave is in a kitchen’. (And further bits added in that way). 

Furthermore, the above discussion also just hints at how the commonsense knowledge pieces (of the domain of microwave ovens) are related to their Procurement, Experience and Memory. Lets dwell a bit on these 3 phenomena, with a one or two examples of commonsense “fragments” corresponding to each. . 1. Procurement fragments : 

These fragments are related to the procurement of the commonsense pieces. In the below examples, we see those which are related to the procurement of what a microwave oven is. 

E.g. – – mama’s cookies during Sunday gatherings of friends… – what uncle John had once called ‘your giant noisy box’ ……. 

2. Experience fragments : 

These fragments are related to our experience as regards to the commonsense pieces. In the below example, we see those which are related to our experience of the key features of the usage of the oven. 

E.g. – the timing adjustment – (the pressing of buttons repeatedly by you) – in multiples of 30 seconds….. 

the eagerness when the countdown is getting to 4 3 2 1 ….yeah. 

3. Memory fragments : 

These fragments are related to the storage of the commonsense pieces in our memory. In the below example we see those which are related to the storage of the procedural logics about a microwave oven. That is, the memory of a piece of logic concerning an oven. 

E.g. – oh ya, mummy had (first) opened the door, to take out the dish …. 

the more the time, the hotter the dish. 

Another concern is that these entries being mere facts, they digress from and lack the aspect of cognition. Even though in the human system too, there are present these facts, they are “integrated” with the mental processes, which make them solid and grounded. Lets see the discussion below. 

A KB will have an entry like – ‘Hearts are inside the body’. (A typical entry would be like A-X-B; A and B being the entities/concepts and X being the relationship between them). 

But, there are 3 components to any piece of such commonsense knowledge – 

1. What is a heart? what is a body? 

2. The prevalent entry – hearts are inside the body. 

3. The meaning of the relation – ‘inside’. Which is, understanding ‘insideness’. And this involves the grasp of simpler experiential phenomena as regards to ‘insideness’ like say – “covering”, “obstruction in seeing”, “(the act of) open and see”, “something coming out of something (say you pierce a ball, and a fluid comes OUT OF it” etc. These exemplary aspects of “insideness” can be perceived by a kid (one or more in each instance of something being inside something), amalgamating into a holistic understanding of insideness. 

Each of these 3 components would get even more solid if we apply Minsky’s point to them – you dont understand anything unless you understand it in 3-4 ways. 

The harder knowledge (to understand) or loosely speaking, the knowledge to understand is the 1st and the 3rd components; the middle component is just a FACT. You understand the middle component only when you have an understanding of the 1st and the 3rd. We have integrated knowledge, not just discrete facts. Otherwise it would make no harm in making a statement like say – “OK, they (hearts) could have been inside the pillows we sleep on or inside our mothers’ bodies or……but it so happens that they are inside our own bodies”. That would rob the human element in the very phenomenon of the possession of that knowledge. 

Commonsense Facts (not facts in general) are integrated with cognition. You may just memorize the capitals of 100 countries without really having an idea of what a capital is; thats not the case with ‘Hearts are inside the body’. 

We don’t really make much mistakes in, or, forget commonsense knowledge. If someone says ‘I forgot whether my nails are inside my lungs or on my fingers’, the (cognitive) rot would have had spread deep and far within – and not just restricted to that isolated discrete fact! 

Commonsense Associations – an issue

There is a problem with associations. It is that anything can be associated with anything. Donald Trump can be associated with Sydney; the association being ‘city at 150 km distance northeast of the airport he visited last year’. This is, technically speaking, an association. But this is meaningless and hence invalid. 

But there are some valid commonsense-associations with an entity –  like Donald Trump is associated with his wife, his bank balance, nationality etc. The difference between these associations – wife, bank balance etc. – and the earlier weird example is that ‘wife’ is an association associated with Trump (with a man in general; Trump in this case). There is thus a “double layer” of association in certain cases (like wife, family, bank balance, nationality etc.) which is that the kind of association (or simply the association) is also associated with the entity. This is by itself and generally. Whereas, the association ‘city at a certain distance in a certain direction relative to his place of visit’ is not an association associated with a man (Donald Trump, in this case), by itself, generally. 

Simply speaking, there are some associations, which are associated with an entity, making them inseparably bound to it (like say, a man’s nationality and family). 

(This is so Linguistically because when we say ‘Nationality is an association associated with a man’, the first word ‘association’ is what we accept to be one and thereby it follows that the second word – associated – becomes acceptable (it being a derivative of the first word ‘association’). Hence it (the first word) becomes legitimately associated with the man.)

Commonsense Information v/s Commonsense Knowledge

         Commonsense Information v/s Commonsense Knowledge 

Definition of ‘Commonsense data’ : things which everyone knows. 

Let us divide this commonsense data into 2 sets – Set (1) and Set (2).

Let’s look at the sets.

SET (1) – 

It would contain commonsense data like – 

The sun is hot

Hearts are inside the body

etc.

Here, these data pieces link 2 entities/concepts with a conceptual relationship

In the first example the 2 entities/concepts are the ‘sun’ and ‘heat’, and the conceptual relationship between them (not explicitly stated in the sentence) is ‘bearing’. Whereas, in the second example the conceptual relationship (‘insideness’) links the 2 entities – ‘heart’ and ‘body’.

Let’s delve more and see some properties of the data pieces in this set – 

  • These are not the first things one comes to know about the subjects – here, sun and heart respectively. Our first exposure/introduction to the heart is its beating that we feel when we put our palm on our chest. The first experience of the sun is that it’s yellow and in the sky.
  • There is something to understand in such data. (This will get clearer later).

Hence it should be called ‘Knowledge’; it isn’t mere ‘Information’. And since it is something everyone knows, it should be called ‘Commonsense Knowledge’.

SET (2) – 

It would contain commonsense data like – 

Sun rises and sets

The sun is yellow/orange-ish, light comes from it, if you look at it straight your eyes cant bear it. Hearts beat, hearts are parts of bodies.

A hypothesis statement about such data – 

These wont play to be relevant/noticeable deduction-grounds of anything intelligent that one thinks/speaks. These are so basic that rarely will a thoughtful question arise and will it be answered by factoring in one of these items. Say, suppose one is thinking about the sun or the heart, there will never be a situation wherein you say to yourself – why is ‘X’ happening? Oh right, that’s because of ‘Y’ (where the Y is an element in the Set (2)). Questions won’t arise in the first place since they would be nullified / nipped in the bud, by one of these extremely basic facts at the “very pre-inception”, so to speak, and things will move ahead in the thought process without it being noticed that these data pieces factored into the thinking. The next link in thought would arise “automatically” i.e. without it being so that — ‘you stop somewhere, then a fact like this occurs in your mind, and the next link is created’ — the whole as an observable, noticeable process. 

In principle, yes, there would be deductions which would be based on these items, but they won’t be worth noticing. Hence such data should be called (mere) ‘information’ (something like ‘My uncle’s name is Peter’). And hence, in this context, ‘Commonsense Information’

Other comments on and differences between the two sets – 

  • If you know an element in Set (1), it means you know elements from Set (2). For e.g. – It is not possible and is pointless to know that the sun is hot without knowing the colour of the sun. But knowing only (2) and not (1) would seem like an odd unexposed mind. 
  • Set (2) is more preliminary a kind of a data-set than Set (1). But the items in set (2) are still commonsense-data. (But “ultra-basic”).
  • Set (2) data doesnt occur in AI Commonsense Knowledge Bases (e.g. CYC). A KB has entries like ‘a man is bigger than a fruit’ or ‘hearts are inside the body’ which belong to Set (1). In these  knowledge pieces (i.e. of Set (1)), you are creating “new” links  between 2 concepts/entities (e.g. the comparative sizes of man and fruit), thus pulling together a relationship. Hence there is a sort of a thinking/understanding element associated with these set (1) commonsense data pieces.

Also, the Set (1) pieces will never be proactively taught to a kid by anyone. (No mother, without context, will tell a kid – a computer mouse doesnt have a heart). 

  • The first things that come to mind when you come across a subject (say, sun or heart), without any context, are the items in set (2); when you come across the subject as a part of a context, the first thing that comes to the mind are the items in set (1).
  • Suppose someone asks a kid – Do you see the sun all day? and the kid says – ‘it rises and sets. So how can one see it all day?’ The basis of this answer is an item from set (2) – ‘sun rises and sets’. The basic visualization of the Set (2) item (rising and setting) is inherently binded with the question statement itself (‘…seeing the sun all day or not’). ‘Rising and setting’ is too clearly and obviously connected to the appearance or disappearance of the sun. So, it hardly comes across as some intelligent thinking was done using that piece to answer the question.                                                                  But when an item from set (1) is used, it does come across that the kid has done some genuine thinking (though tiny in representable amount). For e.g. – Suppose someone asks a kid – What if I take you near the sun? And the kid says – the sun is hot, so we will burn. Here the kid uses ‘the sun is hot’ (set (1)).