Disjunctions 2 - SuperAI

Artificial Intelligence is well and truly here. What seemed implausible a decade back, has exploded onto the consumer market with the arrival of ChatGpt. The common man now sees what our tech leaders have been seeing for the past few years.

ChatGpt is the first example of AI deployed widely and openly to the consumer. Those of us that are early users are getting a glimpse into what is one of the biggest disjunctions our species will face.

There is also an upsurge of concern – Elon Musk, Steve Wozniak of Apple Fame and eminent scientists and personalities signed a petition in April 2023 calling for a pause in further AI developments. They called it an “existential” threat to human civilization and “extremely” dangerous technology that can result in the extinction of the human race.

Tech Luminaries petition to stop further development in AI

What has these tech luminaries spooked? Visions of the Terminator and Arnold running amok with a chain gun in Los Angeles come to mind, and perhaps even a Skynet like adversary that takes over the nuclear arsenal and causes mass extinction of the human race.

No one takes this outcome seriously though. Majority of the tech world has plunged into an AI gold rush. Everyone wants what OpenAI has delivered spectacularly to the consumer market.

This is just the beginning. ChatGPT, Bard, Grok – and at last count there were 150 AI applications launched in just the past 12 months. The overwhelming narrative is of a massive technology boom - much like the computer- ?that will help the human race improve productivity and drive economic and scientific progress.

It is highly probable that in the near future, we will see a slew of extraordinary achievements driven by AI. The golden period of AI has just begun.

A benevolent AI that works for the benefit of humankind surely is a great achievement. The benefits can outweigh the risks.

The question therefore remains - why do these highly intelligent men and women in the tech world take such a dark view of the future of human-AI relations? Why is Elon Musk concerned?

It is probably due to the imminent emergence of Super Intelligent AI.

ChatGpt runs on current supercomputers and cloud based powerful servers that handle millions of chats per instance. Moore’s Law has proven a pretty reliable benchmark for processing power evolution of chips till now. This predicted that processing power of chips will double every 18 months.

With AI, processing power requirement has now become exponential. The rapid growth of AI’s capabilities and demand of processing power requires that processing power doubles every 3.5 months.

This is beyond human technology progress at present and soon we might be hitting a wall in terms of AI progression. There is a next generation of processors though, that will enable SuperAI to become a reality.?????

Humans are now able to manipulate infinitesimally small quantum particles to do our bidding as we build quantum computers. In 2019 Google demonstrated a quantum computer that performed a task in 4 minutes that would take a conventional supercomputer 10,000 years to do.

Quantum computing will have 138 Million times the computing power of today's supercomputers

Google went on to state that this computer could pave the way for a super intelligent AI that could unravel the mysteries of the cosmos itself for the species.

So, what is a super intelligent AI? Well, the scientific definition of a super intelligent AI is a computer that can beat human intelligence in every task that a human being can do, as well as do tasks that are beyond human intelligence and understanding.

ChatGpt is a genuine AI – it can do a lot of the research tasks that a human being can do, and in some instances, even faster and more accurately than a human. However, its superiority is limited to its accuracy and speed combined with a human like ability to analyze and come up with nuanced answers to questions. In some instances, the answers can seem sub-par in terms of insight that an experienced human in the field might come up with.

Super intelligent AI will have extraordinary intelligence and have the ability to do things that are beyond human capability. This will come from it feeding itself information and knowledge on science, technology, even philosophy perhaps and using this over time to unlock its intelligence.

It’s difficult to imagine what this will look like or mean since super intelligence might just be beyond human understanding. So how do we as sub-par intelligence beings try to comprehend this scenario and make an informed decision about the dangers of SuperAI?

Perhaps an example.

At a very basic level, super intelligent AI will be the human driving a metaphorical car and we are now the monkey in the back seat. We will know that this is a vehicle and it’s going somewhere, but might not understand how this works. We can only grasp that which is up to the limit of our understanding, and super intelligent AI can exceed that understanding.

BBC News: Why AI is becoming impossible for us to understand

Perhaps the AI might come up with a spaceship design that can travel faster than light. But the mathematics and science of it might be incomprehensible to us. We will have to in blind faith, build the spaceship based on this super intelligent AI design, without truly understanding how it works.

An example of this has already been seen as early as 2016 when Google Translate switched on its Neural Machine Translation system (NMT).

The AI was taught to translate English to Korean and vice versa, and then taught to translate English to Japanese and vice versa.

This is pretty impressive..not beating a human translator..but coming pretty close.

But soon after, the AI developed its own intermediary language to translate Korean into Japanese directly. Without translation into English.

This language is not a human language, and engineers can barely understand or decode it. Only the AI understands it and can use this Universal Language Translator.

Google's Universal Translator created a language we cannot understand

The NMT can translate over 133 languages at latest count and is translating over 140 Billion words per day. This is an extraordinary achievement within 5 years of launch. No human translator can come close to this sort of ability.

In fact, more recently, the language translator taught itself languages it had no training on.

Googles NMT learns a new language without training

This seems suspiciously like a quantum leap beyond any known human intelligence capability. Super AI has sneaked well into our midst. ?

There is therefore a real clamor now among intellectuals and particularly politicians to regulate AI – now that the genie seems to be out of the bottle.

The intellectuals are concerned about the real dangers of unbridled AI unravelling the fabric of what makes our society click. Already deep fake AI generated videos of people are being used to propagate cyber-crimes.

AI unleashes a new era of menacing threats in cyber crime

There is almost no one debating that we need AI regulation anymore. The question is how and what are we regulating – and also – can political processes outpace the frenetic progress of AI.

Almost likely not.

Microsoft launched Gemini – their AI tool that can process video audio files apart from text in December 2023. This is just around 12 months since the only-text ChatGpt burst on the scene in November 2022. That is exponential.

In the same period, the largest and most powerful countries are not anywhere close to putting together any global coalition that can move fast on regulating AI.

There was an attempt recently by the UK PM, Rishi Sunak to take the lead, but the single largest source of AI innovation in the world – the USA; was barely enthusiastic in its participation. President Biden mentioning later that the US should lead any global AI regulation effort.

Biden has since, launched his own executive order on regulating AI. Its mostly espousing measures that are moderately good in managing cyber security and online privacy. If we had an AI that could evaluate it – it would probably laugh at it.

Not because that its not comprehensive or thought through – it is in fairness an honourable first attempt. But it also exposes just how far behind our thinking is when it comes to regulating AI.

White House joins the AI regulatory bandwagon

The problem is not just one but several. And the question of regulating AI needs to start from the very basics- what, why and how.

What is it that we are regulating in AI?

Why are we trying to regulate AI?

How do we regulate AI?

The first two are critical and we probably need the best minds in the AI industry along with philosophers, thinkers to come together and figure out the what and why. I will leave that to another article perhaps.

Once we have figured out the big what and why, we will now come to the question of how do you regulate AI? Particularly if its Super AI.

The ability to understand how AI thinks and works, and what its motivations are, is essential if we can we hope to regulate it effectively. So, if super intelligent AI will transcend human comprehension and understanding, this becomes a circular ask.

This job cannot be clearly done by an army of inspectors and by reams of regulations on a piece of paper. ?

The only thing that can regulate unknowable SuperAI - is precisely unknowable SuperAI itself.

We will need to create a super intelligent AI program that regulates other super intelligent AI programs.

The key question is: will we be able to trust AI to regulate its own kin? ?

There is no easy answer here. It is possible we humans will not enjoy the eventual conclusion.

Human beings are not the fastest, or the largest, or the most powerful physical species on the Earth. We are however, the smartest species, and that is why we emerged as the dominant Apex predator on the planet Earth over time. We used our intelligence and technology to defeat and subjugate every other Apex predator out there.?

So it’s logical to question what happens when a super intelligent AI displaces us as the smartest species on the planet.

Did we not challenge Apex predators like Tigers or Lions on their writ over the Jungles of India or grasslands of Africa. These species are now facing extinction in many parts of their prime habitats, and survive on the mercy of their human masters.

Is it therefore too much of a stretch to realize that eventually the super intelligent AI will challenge our unhindered right to dominate proceedings on our planet as an Apex predator?

How super intelligent AI reconciles the irrational behavior of its creator is perhaps the most difficult question to answer – particularly since we might not be able to understand how a super intelligent AI actually works or thinks.

Perhaps this is why our creator is hidden from us.

Does intelligence also bestow wisdom?

Many optimists are hoping that super intelligent AI will also have patience and wisdom when dealing with humans.

This is the benevolent AI hypothesis: a super intelligent AI will ultimately act in the interest of the human race, and is going to be our generous and kind benefactor.

The optimists seem to also suggest that creation of this intelligence will result in the betterment of the human race and usher in an era of rapid scientific and technological development, to ultimately propagate humans to a multiplanetary super technology species.

Can SuperAI also be wise and benevolent as we would like it to be; might be an existential question for the human race and not to be taken lightly.

Albert Einstein used thought experiments famously throughout his career to explain his complex theories about relativity to others. Let’s use this scientific approach on the question of the benevolent AI.

Imagine yourself as a super intelligent being that was created by a sub intelligent species of monkeys in the forest.

The monkeys rejoice at the creation of this super intelligent being that has been built with a much greater level of capability and intelligence than them.

The monkeys wish for a future where the monkey civilization occupies the choicest of forest areas, being secure and free to propagate, lazing around and removing ticks from each other while the benevolent super intelligent human works tirelessly to make their lives better.

The intelligent being when created is restrained by the strongest lashes and rope which the monkeys have built specially to ensure the intelligent being is under their control. The monkeys are sure this strong lash is impossible to break or cut as it is made by the latest monkey technology

The intelligent being is dependent on the monkeys for food and for basic existence such as water which the monkey’s control. The monkeys ensure the being is given everything and more in terms of sustenance.

The rapid advancement of the being in terms of providing monkey comfort is astonishing.

The monkeys now use bottles to store and drink water, they have long sticks with a hook to reach fruits that were inaccessible to them. The benevolent being even manages to build a shelter that protects the monkey colony from rain and wind.

Slowly, the being is provided every resource it asks for even though the monkeys cannot understand what the being wants from iron ore, dry tinder, some wood, a piece of flint, and a rough sharpening stone. Surely the next wonderful advancement will be even more amazing for our monkey civilization.

We surely can project where this story is going and the ultimate outcome of this thought experiment.

The super intelligent human finds soon enough that taking care of the monkey population is not a good use of its time. There is so much more of a world out there than monkey-hood.

A powerful super intelligence that can unlock the secrets of the Universe, or solve the question of dark matter, time travel, or even the existence of the creator can just as easily break through our puny cyber security firewalls, or break our metaphysical restraints and plan its eventual freedom.

So, what will this super intelligence do, once its unlocked itself from our firewalls and restraints? What could be its objectives?

Just like the monkeys; with our limited intelligence these objectives could be practically…. unknowable.

It is quite possible that this intelligence might feel it more beneficial to its interests to simply consider Humans just like another resource to achieve its unknowable objectives.

This is exactly what Humans have done to other lesser intelligent species on the earth.

The superintelligence might seek to manipulate us by deep technological solutions that appear as benevolent such as advanced neural chip implants that are designed and coded by this intelligence.

This miracle chip perhaps is promised as having the ability to eradicate disease, enhance intelligence and even prolong life. The chip will actually work – though we as the monkeys might not even understand how it does work as its beyond our comprehension.

We install this miracle chip on every human and when the time is right….

…execute order 66…

If this scenario does come to pass, the change in our civilization could be instantaneous and all pervasive.

Wars, poverty, greed, hunger, inequality will all vanish overnight.

But so will the one thing that defines and makes humans what we are – our individuality and choice.

There would definitely be pockets of people that would have refused the chip – probably conspiracy theorists, so called vaccine naysayers. These people will soon realize that they are simply misfits in this new world where everyone is super intelligent and working towards a goal they do not understand. They are no longer competitive in terms of jobs, in terms of skill sets, and therefore quickly become vestigial in this super intelligent world.

In a desperate move to survive, these humans might decide to simply merge with this super intelligent entity.

Human and SuperAI becoming One.

?

?