Between the hours of 8am and 10am on a Sunday, chances are I’ll be sitting with cinnamon porrdige and a coffee at my dining room table. This ordinary for me, and I hope many others, in the fact a repetitive task as monotonous as me telling you my breakfast routine doesn’t make you bat an eye, if anything you will be wondering why just am I writing about it?

It’s human nature that, we do something once and our brains feel already comfortable with the task; we create a chain of thought, logic of events and their entailments, chronological momenents that we can refer back to if we feel we ever feel like we are appraoching the same situation. This is the human mind, do once and remember, do again and compare routine. Tailor our own unique experience to what makes us happy and prepare ourselves for the day, in my case of reference, the Sunday long run.

Yes, that's a universal experience

Now following is a very niche reference, I don’t have comments on my blog posts yet so I’ll hope this reference lands. In the disk world series Terry Pratchet’s, ‘Wen the eternally suprised’ fictional character states, and in my opinion this is a hell of an opening:

Wen considered the nature of time and understood that the universe is, instant by instant, recreated anew. Therefore, he understood, there is in truth no past, only a memory of the past. Blink our eyes, and the world you see next did not exist when you closed them. Therefore, the only appropriate state of mind is surprise. The only state of the heart is joy. The sky you see now, you have never seen before

The case is, Wen is eternally suprised (Not my joke the books), Wen sees’ what is - his concious mind is present always. He has no subconcious that he delegates tasks to, he sees all and everything at this very moment, quite the opposite to my muscle memory of breakfast and coffee.

We get it, you are not a monk

No, I’m not. But the fact is not that I’m not a cool monk, the fact is the autonomic process of repetive action being learnt and monouved involuntary and instantaneously by our brains. It likes my warm drink and food, and therefore my concious self - designed for terrain navigation, survival and many other things to keep the human race going… Doesn’t have to think very hard. The task is delegated to the iceberg of the sub-concious. Now this to conclude my point, is a double edged sword for humanity - and the underlying principle - technology.

Alright, that was a weirdly long intro...Continue

I hope these titles are quoting you - the readers - inner monolouge, and hope even further you said continue… So I will. I finish my breakfast and pick up my coffee to take it upstairs.

“Don’t drop it James”

I tell myself, yet I have taken a hot drink upstairs hundreds of times - yet why on earth do I remind myself not to dribble luke warm coffee all on the strairs?

Inevitbly I do, and that’s the point. My mind has taken hot drinks upstairs hundreds of times, spilling none; It knows exactly the finite motor control to balance a mug with drink - not saying I’ve got gyroscopic hands but I’m far enough down the chain of evolution to not spill any. Yet I do, when I get involved with myself.

Nice story punk, you spill your drink

I do! And do you also not question that fact? That when we tell ourselves to do something, we often end up doing the complete oppostite. Like we tie our own shoe laces together and then take a step, or we put clingfilm at our own head height and walk into it…. (I’m running out of 2000’s pranks) It’s this meta-awareness, the self-involvement and inter-communication practise which is the theme of todays post.

The book ‘The inner game of Tennis’ by Timothy Galleway encapsulates this perfectly, He calls it the ‘Self 1’ and ‘Self 2’ The ‘Self 1’ being the concious, decision making command generating general. The ‘Self 2’ the do-er, the one with much better control over the nervous and musculoskeletal system and the one which always watches and obeys. Timothy explores the idea that we have lost the intuitive capabilites of the mind, to utilize both left and right hemispheres to execute with efficiency, without the scololding of the ‘Self 2’ from the ‘Self 1’ through negative self talk and getting in our own way.

“Right, now there is coffee all over the floor”

Good thing I haven't split any on the dog yet.

Timothy’s notion of the two selves, working in unison for eachother, guides atheletes to becoming the pinnicle of their respetive sports. To utilize the dormant process of intuitive learning and subconcious activation, a term popularly coined is ‘flow state’ and is the locking of ‘Self 1’ out of the room and letting ‘Self 2’ figure it all out.

If I don’t tell myself not to spill the drink (Double negative in da houseee!) then I won’t; If I tell myself to not spill the drink I do, and If I tell myself to spill the drink… Although I have free will my moral compass overrules any form of the eventuality of coffee on the floor, and for that parents Im’m most thankful.

Although this is very consequential, this form of reinforcement self talk is seen to inhibit human excellence, Timothy explores that on the tennis court; Pro’s are not struggling with the outer game i.e holding the racket correctly, they are struggling with the inner game. The game between these two selves, the game which is so engrained into our very nature, and how to not get in our own way.

The declarations

This dyanmic relationship between an command giver and the action do-er I have realised manifests in an assortment of places

  • The human brain (for one)
  • Input and output systems (AI prompt engineering and Large language models, a recent rennovation)
  • Scaling mechanisms for high system demand (Declarative programming languages)

How many times have you input into Chatpgt a prompt, and what it returns you critisize.

“No no no not that, focus on this” and “Okay now change that”

Does the model always do what you want? Sometimes. Often not first try, this in essense models the learning process. Focus on one thing and the brain forgets everything else, or the millions of transformers in this case, due to our feedback.

As a side note, an interesting fact I read, when we as the reviewer ‘Self 1’ in this analogy are working with LLM’s. We are dealing with an exponential amount of processing. If every floating point calculation used to train GPT was a drop of water, then the accumulative calculations would fill the Pacific Ocean, purerly incomprehnsive numbers. Our instant gratification mechanisms (dopamine systems there again) say “No no no I didn’t want that… Focus on this”. We completely undermine the the nature of what we are dealing with. After all, funnily enough these model’s processing are based on the brain - and the same problems occur, its all recurive. We become the ‘Self 1’ the LLM is the ‘Self 2’. It’s everywhere.

The idea of ‘human in the loop’ is used in Large language model training (GPT-) we are the ‘Self-1’ and the model being trained is ‘Self-2’. This is reinforcement learning (Dopamine system ey,ey) this is feedback or declration of right and wrong; But the ‘Self 1’ does not belittle the ‘Self 2’, it does not call it silly, and neither does it think it’s incapable - so why is it human nature? I’m sure familiar to us all to inhibit our own performance - we are the ultimate evolution machines! Yet we have no self confidence, this relationship is futile and on a side note, reverse engineering our toxic self talk and unlearn the habbits of intereference is key to exploring self-furfilemnt and enjoyment of all.

To cover, there are many ways to train a LLM, Supervised learning and Unsupervised learning - and I bet you can guess which has feedback systems (supervised)and which doesn’t(unsupervised) and consequentially which is compartively better at discovering new pattterns and relationships in raw, unlabled data (unsupervised).

I put the anwsers in brackets on a final read… Realised it wasn’t that clear

So declration... It's the future

This whistlestop of an obserever and machine agents, the exploration of selves and model training is just the tip of the iceberg. Genetic programming, a declaritve means of solving a problem by defining scope and requirements, and letting the ‘program’ learn the best solution architecuture is an established method of problem solving for both P,NP and NP hard problems, which is another blog post in itself.

As a taster, if we figure out how to compelete super mario in Polynomial time, the cure for cancer is there as well (theoretically… Not my words). Stay tuned.
Mario Link

NP hard Medicine

Conclude and shut up

I will, as you the reader can see I am a huge beliver of mimicry of nature for man-made solutions. Industries are hurtling away from either natural or man-made objects to hybrid, genetic and in all honestly terminator like society. What I highlight is cross domain, Psycologists, Philosphers Bio-Chemists and Computer Scienstists all have a future in the coming wave of Artifical Intelligence. All I wish to say is pay attention to what you pay attention to.

To quote Stephen Hawkings - “Be brave, be curious, be couragous. Overcome the odds, it can be done”.

Thanks