None of this article is written with any A.I. assistance.

This is a work in progress, my scratchpad of ideas if you will.

2024 Jan - Ki

A LOT of people ask me about A.I. 

As I posted on Twitter recently... 

'I literally invented A.I. that controls and flies weaponised military vehicles autonomously decades ago, and I have to defend why I call ChatGPT a parlour trick to script kiddies that took a course on YouTube last week.'

This is an overview written by one of my close friends who also wrote something akin to points I make below AI: Hello World - Simon Funk

'But then everyone is talking all day to an intuitive AI who's programmed to alert the authorities when it's little red flags are raised, which means we have pre-crime overnight too, based on presuppositions ultimately chosen by the likes of Magneto or Moderna. Welcome to the future.'

Before we start, please watch this.

Everything I have to say on this topic was perfectly captured by the Monkees

2015/16 - Chinese company starts the heavy change 
 

Let's start with ChatGPT and DALL-E

I’m going to write up the most important things I have to say on this topic here and will update this when I get any new information.

First, since I hate undescribed abbreviations/acronyms:

• DALL-E – is a portmanteau of Wall-E and Salvador Dali’s name. This lets you describe an image and it pops out ‘something’ that will probably impress you.
   
• ChatGPT – is Chat Generative Pre-trained Transformer – In other words, it will make new stuff out of old stuff it found laying around the web.

Do not confuse A.I. and Machine Learning (ML)

Machine Learning is a process for mining (sifting through) data, finding patterns and statistics, and then making guesses that seem very much like humans.  Not all A.I. uses ML.  But ML is a subset of A.I.

Is it Artifical Intellegence?

This is the real question, and thus we have to agree on what A.I. means. This is akin to ‘What is life?’

Here is my summed-up take:

A.I. is anything that appears or behaves in a way that makes us think it makes decisions.

Therefore, even some very simple mechanical devices could be considered A.I. in my mind.  There used to be people that rang bells in churches so everyone knew the time. Most lost their jobs to wind up springs.

Is a wind-up spring A.I.?

Yes! When set up to do what a human was doing.

If I made a simple scale of 1-100, where 1 is that spring I mentioned, and 100 is a human of 100 IQ, then I place A.I. today well below 20 IQ, not even a pool of drool, but it would be a savant pool of drool. And magically would spew and sputter out amazing sentences. It belongs in a Circus.  It is not smart, it does not 'understand' anything.  But it can do some cool tricks.

Summed up, the current version will/can:

• Save some people time - Allow smart people to waste less time.  Programmers should leverage this very well. It might replace much of StackOverflow (a tool programmers use to troll each other, and pretend to answer questions. This is good since they put their energy here as opposed to trolling their friends and family).
   
• Search engines - They might replace Wikipedia for the masses as a resource to answer questions. Microsoft, Google, etc will be installing this as their new search engine. In general, though, normal people won’t make much use of it yet. But it will be like having a neurodivergent (ND) savant friend they can ask endless questions of (cool, we're tired of answering all your questions).
   
• Marketing - It will write better prose - Marketing people will use it well, and non-marketing will use it to be a little better than normal.
   
• Writers - It will help people write good.

Will the next version be way better [smarter A.I.]?

No, but it will do the same parlour tricks better.

What is a State Machine?
 

You need to understand that State Machines can create the illusion of intelligence.

You probably have a light switch on the wall, and it is connected to a lamp on the ceiling. That is a machine with two states: Off, and On.

Now, imagine adding a second light switch.  It goes to the same lamp, but a different light bulb.
Now you can have Light and twice that light.

Thus three light levels:
 

• Off
• 1st switch on - Bright
• 2nd switch on - x2 brighter

But there is another option, you could put 2 bulbs of different brightness in the lamp, perhaps 500 lumens and 100 lumens.

Now you have 4 states:

• Off
• 500 lumens (1st switch on only)
• 1K lumens (2nd switch on only)
• 1.5K lumens (both switches on)

If we keep going with this analogy, you can see perhaps how we can eke out more and more tricks from a simple machine.  We might even make it appear analogue (smoother, human-like, or biological in nature).  Much like how music can be stored on discs digitally, but sound smooth. 

We all accept this now, but there was a time it was a huge (silly) debate if digital could even make something sound good.  In computers, this is ultimately just machine states. Passed then to the next function, that acts on them.  For now, there is no proof that the path of current AI-related technology like DALL-E or ChatGPT will garner us smarter or more intelligent machines.

My opinion is standing now for 20 years, they will start replacing humans more than they create new jobs, and at a rate where the jobs will go before the prices for consumer goods, housing, and support (elec, water, gas, etc.) drop.  This is the formula that matters.

My best guess right now is that we will enter a 20-year period of pain.  Jobs are lost, but prices don't fall fast enough.  Odds are this created strife, and massive civil unrest.  We will need armed robots to stop people from destroying unarmed robots.  And so it begins.
 

Current thoughts on ChatGPT

Let's first break ChatGPT down into several concepts to demystify it.
(this is super simplified for lay people, and not in this specific order)

Trick #1 is to build a system that can write sentences following the rules of a given language, starting with English.
This alone is actually a lot of work, and there are a lot of exceptions.  But one great way to do this is to review millions or billions of sentences and look at word replacements.

'I pick up the ball' can be replaced with many nouns: phone, wallet, plate.

But things get murkier when the object is larger than you.  But still possible, e.g. 'I pick up a truck.'  
Rarity is a noun, 'I pick up the rarity' would be a rare sentence.  So rare, ChatGPT will probably not write that.

And that is our #2 trick - simply avoid writing things people would not say or write.  That alone will help ChatGPT appear more human.

Trick #3 is more complex, we need to abstract all of this, and focus on the goal of what is being said, and then bring that all back together and pop out not just sentences, but whole paragraphs that express an idea.  But, all said, it is the same concept as above, just extrapolated out.

But it is the next trick, trick #4 that I believe is what makes ChatGPT seem eerie smart.  Even if we toned down the other tricks I mentioned, if we were talking about a recipe for food, this would be salt, sugar, and butter and fry it.  Meaning, this makes almost anything taste good.

It's called continuity.

What ChatGPT is doing is having a conversation with YOU!

And as you type your questions, it is tracking what you said, then what you say next, and what you didn't say, and it is focusing the direction of the conversation to produce better and better results (hopefully).

When you ask a second question, it sort of puts it together with the first question, and uses its own first answer to guide the answer to the second question.  This is already better than most clerks can do when answering questions.  [one day I will insert a great joke here].

Lastly, and this needs to be mentioned, it does one thing that I hate when humans do this, which is to try to give you an answer, even when it doesn't know.  Most humans don't know stuff, but they do their darndest to make you think they do.  'I don't know' is a great answer much of the time.
 

A.I. will take jobs, not make them

MOST jobs fall into just one of several major categories:

Vending machines

A person takes money, and hands you an object.

This includes Liquor store clerks, cashiers, bank tellers, etc.

Driving machines

Taxis, Aeroplanes, Helicopters, Trucks, Boats, etc.
Do you get it? If it drives something, it will be a robot.

Babysitters

People that watch things, like security guards.

But this extends to people that tweak small things (on par with wiping a baby's arse) like pool repair, HVAC repair, etc.

They take the diaper off, they wipe it down, and they put the diaper back on.  They swap out filters, replace o-rings, swap oil in a car, 

Also, people who clean, Janitors, gardeners, and even painters; they keep 'things' clean.

Every single one of these will be replaced by a robot, pretty soon. I own half a dozen robots already.

But even jobs like Teachers. Most teachers regurgitate information. So they are closer to babysitters.

Harsh, but, sad. And, for me, this is a form of noise. I don't really want to engage people that are not directly pushing society forward.

The great  Automation Dividend (the big crossover)

I don't know the exact dates, but I welcome anyone explaining to me this is now our reality soon.

But also, after the great crossover, things should be great!

REF

Technological Unemployment Crisis
 

#automationcrisis

The 3rd A.I.

This is where I'm leaning in terms of fixing the problem of overloading people with Tasks.

1. The first A.I. is already inside us as humans.  There is us, but there is a second thing that is the model of the world we have to deal with that is ever more complex.  (being animals on the planes is our baseline).  This goes deeper into an area I've been researching, and attempting to find tests and models that might disprove my theory, but I'm finding more to support it, which is that we are each a hive mind, with one mind taking the lead in any moment, creating the illusion of continuity.  
   • Zen - It would explain easily how we can NOT be focused on a given task, and yet the solution POPS into our mind out of nowhere.
   • Dreams - our minds (unchecked) running amock without oversight. Our awake mind, the one that forces us to deal with reality, the moment, the now, is off duty while the brain (the physical) recovers).
   • Schizophrenia - or as I see it - temporal communication lag with trust.  Said another way - our minds dream constantly, and the accountable mind of being awake has a blind spot, and that is information that is time-shifted.  Messages from one mind to the other are delivered with trust (our minds trust one another).  But here, they should not, because the messages are old.  Even by milliseconds.  As a result, we 'believe' dream-like messages from other minds running around doing their own business.
      
2. The next A.I. is crap like ChatGPT.  Tools we think are going to help us.  But it is also just the collective of tech and social and all that OUT THERE.  Remember, A.I. is a general catch-all word (read above).
    
3. The 3rd A.I. - is a needed system that is a buffer, or proxy between a given person and the world.  It monitors your well being, ensures you are not over-taxed, stressed, crippled even.  It lets you sleep.  Building this shield seems like a wise model to start thinking about.

Why do we build A.I.?

This is a capture of something Sam Altman expressed.

Ineffective people

As one on the spectrum, I agree with Atlman.  

One person responded to that post thinking it sounded schizophrenic.  While completely wrong, and that person should never practice medicine, I find it interesting, in that - this is how we (on the spectrum) view most of you, neurotypical people.  NTs are random and emotional, and most seem to suffer from ADD... squirrel!

Another used the word 'gross.'

'Gross' is a great word. In fact, it's so strong, and sharp, it is a sword, and swords cut both ways. Let's be harsh here, and if you are NT, be honest here... aren't 'most' people, on the whole, ineffective? Don't be nice, be truthful. 

What are we measuring against?

We view most tasks as repetitive and boring.  We view this as being enslaved.  As best I can tell of my peers in this field, and I definitely am motivated by this, we are building tools to break these shackles.

But, what we find, and is expected, people don't like change, even if it is good change.  All change is bad (psst, this is where NTs and NDs are the same, NDs perhaps even vastly more so).

The future of art

Fill this in

• The writing was on the wall, 1980, why I left art.
• You're not competing with the quality of your art, you are competing with the sheer quantity of stuff that looks like art to others. 
• Do artists have a future?  Animators?  etc.? 
• What does art look like in the future?
• Building construction of art.  Overseeing projects, and beautifying the world.  Picking the 'best'
• Upping one's game.

Recapped statement 

AI is not in competition with us, we compete with the new economics created by a new species on our planet.

If we can get it to drop the price of food and shelter, and free our time to make as opposed to clean, we will return to just living. Trust me, when you reduce stress, sex happens.

REFERENCES

Lump of Labour Fallacy

In economics, the lump of labour fallacy is the misconception that there is a finite amount of work—a lump of labour—to be done within an economy which can be distributed to create more or fewer jobs. It was considered a fallacy in 1891 by economist David Frederick Schloss, who held that the amount of work is not fixed

What’s Really Going On in Machine Learning? Some Minimal Models - Stephen Wolfram

'It’s surprising how little is known about the foundations of machine learning. '

= END =