The word was left for you

we

Chapter 4 Intelligence


Deep Thought does not know the ultimate question to Life, the Universe and Everything, but offers to design an even more powerful computer, Earth, to calculate it. After ten million years of calculation, the Earth is destroyed by Vogons five minutes before the computation is complete.
From "The Hitch Hikers Guide to the Galaxy" by Douglas Adams

This chapter contains two halves. The first half is about the "Statistics of words" and why that gives humans an enormous advantage. The second half is "How to develop intelligence" from using language and neural nets.


For "Artificial Intelligence" as Deep Thought would tell you if it was a real computer, that what you need is a lot of radical individualist thinkers all with free will listening to whomever they choose to.

This is because, with many simple connections, the intelligence goes up in the order of squares of the number of thinkers not linearly. So having a planet full of trillions of beetles should be a million times better at thinking than the billions of humans. Assuming the beetles communicate globally. However, Words make all the difference.

Statistics on Words


It may be that we do not understand Beetle, but with humans, it is a lot more obvious that we communicate globally. Human individuals are parts of families; families are parts of tribes; tribes became cultures, city-states, countries; Renowned trading families struck trading agreements. These then formed alliances and grew into trading blocks, and religions and global corporations transcend all these borders.


All these groupings have rulers and governments and committees, they create minds within minds, within minds as each culture Chooses mechanisms for making decisions as does every sub-culture.


Those minds have been slowly getting bigger and bigger as the groups have moved towards a global economy. It seems that all humans do is communicate.


However, there is a big difference between humans and beetles; we appear to have many more words than beetles; words expand the power of communication way more than astronomically.


When we use words, these connections can have so many more subtleties. The formula for the number of meanings is the number of letters in the alphabet raised to the power of the number of characters in the word. The growth is astounding.



Characters in phrase
Letters
1
2
3
4
5
6
7
8
9
10
1
1
1
1
1
1
1
1
1
1
1
2
2
4
8
16
32
64
128
256
512
1024
3
3
9
27
81
243
729
2187
6561
19683
59049
4
4
16
64
256
1024
4096
16384
65536
262144
1048576
5
5
25
125
625
3125
15625
78125
390625
1953125
9765625
6
6
36
216
1296
7776
46656
279936
1679616
10077696
60466176
7
7
49
343
2401
16807
117649
823543
5764801
40353607
282475249
8
8
64
512
4096
32768
262144
2097152
16777216
134217728
1073741824
9
9
81
729
6561
59049
531441
4782969
43046721
387420489
3486784401
10
10
100
1000
10000
100000
1000000
10000000
100000000
1000000000
10000000000

With an alphabet of just ten characters and a ten letter statement, you can create 10 billion words approximately 3,000 times as many as there are in the English language.


The extra length to the words in an actual language is mostly due to the grammatical restrictions in human languages of mixing consonants and vowels, using prefixes and suffixes in standard ways; which means we do not use the full flexibility of letters.



With just an alphabet of 17 characters and 17 letters, you can create over 827 x 1018, many more than we could ever need in a lifetime.

An aside on random passwords
This is why using truly random long passwords including upper case, lower case, numbers (some accept punctuation as well), that are different on all your internet and computer accounts is so important.


Just the lower case letters is an alphabet of 26 letters mix that with upper case gets you to an "alphabet" of 52 letters and adding ion the digits gets you to 62 characters. This increase in the number of characters increases the options a lot but not as much as by the length of your password. Each character you add makes it 62 times harder to crack; as opposed to just lower case, which is only 26 times as complicated.


Lower Case. Lower case & upper case. Lower case; upper case & digits.
Characters 26 52 62
1 26 52 62
8 2 x 1011
5 x 1013
2 x 1014
16 4 x 1022 3 x 1027 5 x 1028
24 9 x 1033 2 x 1041 1 x 1043
32 2 x 1045 8 x 1054 2 x 1057
40 4 x 1056 4 x 1068 5 x 1071
48 8 x 1067 2 x 1082 1 x 1086
56 2 x 1079 1 x 1096 2 x 10100
64 4 x 1090 7 x 10109 5 x 10114

Punctuation complicates the issue again. However, long memorable passwords are always the best. You can use any long quote you want, possibly from your favourite song or poem or a sentence or two from a manual or notice that is always visible. Deliberately mis-spelling them adds to the complexity of hacking them enormously, preferably in a different way every time you use a particular quote as a different password.

Back to astronomical numbers...


However, any sentence can be as long and as complicated as you want. The actual number of logical statements is way beyond astronomical.

With our 26 letter alphabet with punctuation, one could in just 16 letters uniquely name every single star in the visible Universe. Although, at the moment we cannot see them and I suspect we would choose a far greater number of letters to avoid names with excessively repeated consonants or vowels or names that sound the same.


With our language, it is possible to describe far more than the Universe, within one's own lifetime. Our community of billions with all its hundreds of thousands of words far outstrips any animal which does not use words or only uses very few words.


It is the use of words that make humans so much more intelligent than beetles. The word is far more crucial discovery than fire. If someone discovered fire without the word to pass it on to their children, it would have soon gone out.

How to develop intelligence

 

American and English difference

 

Neurone and neuron mean the same thing. On the BBC we use neurone. So, please excuse the British spelling.


Getting back to intelligence

 

Returning to the subject of many small thinkers and how to develop artificial or even collective intelligence.


We find in America and indeed most "Westernised" democracies free-thinking humans tend to listen almost exclusively to their leaders, whether that is political leaders, thought leaders in science; religion; fashion or popular culture. They give up their freedom voluntarily to fit in with society and gain cooperation and acceptance from their peers.


Let us look at a technically more straight forward example.


Neural nets are the most economical and practical form of artificial intelligence, otherwise known as AI. They mimic our understanding of the structure of the brain, hence the name. The network consists of many nodes, each with many inputs and one output.


The brain is made up of many neurones. Each neurone has many inputs called dendrites, and one output called an axon, the same pattern as the neural net.


Neural Networking has many layers of these very simple nodes; each node listens to whichever node it can on the previous layer and has one output to the next layer. Almost always the most crucial input node is what is known as the feedback loop. The connection which tells the network how well the whole system did last time it made a decision the axon of the whole of the net.


Initially, each node chooses in some predetermined or random way which input it is going to deem important. Each input is another node's output. All it can do is adjust how important it thinks the various sources are, how much it "loves or hates" that input, and adjust its output accordingly. So it multiplies each by a value which modifies its importance and so comes up with a different result.


In its simplest form, love is a measure of how much we accept someone and hate is how much we reject someone or something. It is the weighting the lover gives as to how important the loved one is. The adored at one end, the ignored in the middle and the detested at the far end of the negative end of the love scale.

That is a very simple but profoundly important definition of love. It is a part of the simplest intelligence. It keeps track of how much it appreciates the relevance of each input and has a memory of it. The axon the output of the network feeds back to the first layer. That feedback gives the neural net the ability to learn. As it provides a measure of success of each combination of weightings and the network learns the best way to get the desired result.


The only input from the "Programmer" is a definition of the desired result.

Like a child, it learns from experience, when it makes mistakes it corrects them and usually within about 36 generations of evaluation the neural network is quite capable of knowing how many cars are waiting at a set of traffic lights or on a road faster and more accurately than a human can. There are some things they are good at and some they are bad at.


For example, one military experiment went badly wrong, and instead of being able to count tanks on the battlefield, the neural net learnt to recognise whether it was a cloudy day! On cloudy days there were no angular shadows or shapes. On
sunny days there were lots often many more than there were tanks.


On many web sites, the owners want to protect themselves from computers accessing their pages.
They do this using a variety of means. These are generally known collectively as CAPTCHA - "Completely Automated Public Turing test to tell Computers and Humans Apart".

They can be as simple as adding a checkbox you have to check. This option blocks any attackers that do not check the scripting on the page. It is far from perfect. More complicated ones such as Google's offer you a series of photographs, and you have to identify some repeated features on them. The simplest one explicitly aimed at defeating AI uses distorted images of letters and numbers. We recognise these much better than neural nets and so we can stop computers using AI that pretend to be humans from accessing the web site.

Most people would not call the weighting of an input love in one node appreciation nor intelligence but put only a few hundred nodes together in a layered network, and they can recognise how many people are on a busy platform on an underground platform to within 1 person something very few humans can do!


The advantage of Neural nets is that groups of neural networks can form more complex neural networks. For example, at each road junction in London is a closed circuit television camera and associated neural network can tell how much traffic there is at each junction. A central neural network takes the input from all these junctions and can then keep the traffic flowing through London. It can take in to account the concerts, the football matches and it manages London's traffic lights. It does so much better than any human could. The central neural network coordinates all the lights in the junctions centrally and then sends out the timings back to each set of lights individually.

Some AI lift systems in Japanese skyscrapers know the habits of their inhabitants better than the inhabitants themselves. The systems ensure the elevator is already present and open lift doors, as that person approaches the lift, not through a camera seeing them but by knowing their daily and indeed weekly habits and annual habits.

Now that is what I call service. I think you can see where this very simple form of intelligence is providing a service we could recognise as loving.


However, we must recognise that this "service" is programmed in as the desired goal.

These neural nets are making real-world choices. They are a part of our culture and performing useful services for humans, both individually and collectively.

We listen to our thought leaders is for the same reasons. We follow our leaders for a similar purpose to find out how well we did and how we can improve.

Lastly, let us look at the downside of AI. Various "bots" and websites, e.g. Google and Facebook, are doing the same on the internet. They are using AI to show you the ads that you are likely to want to see because you share the same habits as other people, and those people used those ads.

They are also showing the news articles they think you will find interesting. So they cause auto-censorship based on your habits. These techniques increase the likelihood you will only read news articles that support your point of view, thus eliminating the need for analytical thinking and educated discussion. You do not see any contrary opinions to discuss. So you do not learn to defend your views, and it is polarising the populace through their own habits.

The programmed goal of these AI platforms is to increase the profits of the companies that use them. We need a contrary set of AI whose purpose is purely for the benefit of the global community, including humans and the welfare of every living thing on the planet.