There is an important differentiation between AI and predicting the future, however, AI will bring our future closer.
I have been coding for two decades, so after thinking at length about how I go about coding AI, I asked myself what would I do with it. I am going to be honest and say that monetary gain did slip into the forefront of my mind, but then I asked my girlfriend what she would do if I created AI and gave it to her as a gift. Without spending a second to ponder my question she replied; 1. fix health (the big C, Alzheimer's etc), 2. Sustain the planet (renewable energy, catastrophic disasters etc) and 3. Space exploration. Well that put me in my place.
The big question for me is who will create this AI, because (assuming its not democratised) whoever does will undoubtedly influence the morality of its uses and ultimately decide where first to apply its power. Because, for now at least, we need huge amounts of compute resource at the Amazon, Alibaba, Google level to analyse big data sets, we are likely to have someone with an adequate level of the people's certitude to manage this huge task; Mark Zuckerberg, Elon Musk, Bill Gates. If Elon Musk said to the world "Dear All (global), please allow me to access all of your data, present, future and historic, and I will help cure cancer, amongst other things. Cheers, Elon" you probably would concur right ? Obviously this would require a majority democratic vote because access would be requisite to everyone's data and not just the % who say "yes". Btw Elon would not create AI himself, he would be an investor in such a company, in the same way he was an early investor in Deep Mind.
Just ponder for one moment the implications of existing AI technology intuitively accessing the world's data collectively. NLP or Natural Language Processing can already understand human language written in emails, notes, essays, reports, documents, studies, papers or social media. Even if AI were simply just to assimilate the competitive landscape of almost any industry, it would already be of superhuman essence, purely by definition to the extent of the human ego, capitalism and competitiveness. Maybe even corruption.
Conversely, what if, an unknown creator developed AI ? It's important to consider the implication organisations such as the NSA and GCHQ has on our privacy. No longer would a creator have to hack into 1Bn Twitter accounts, billions of email accounts or billions of mobile devices. No, now almost every single keystroke from almost every single user or website is monitored and recorded, albeit for a limited amount of time. But then of course, the other half of the challenge is all the data that does not have public Internet facing access, medical databases, IL6 level data for national intelligence, the MoD, the military amongst others. WikiLeaks proved some of this data can be accessed, however all of that data would have originated from somewhere prior to being uploaded into a database, namely someone's laptop or PC. But if AI were to become superhuman, which it would do very quickly after becoming of human level intelligence, then it would surpass our current security levels of 2048-bit or TLS (or higher) of encryption.
To become this intelligent would require huge amounts of processing power, therefore postulated by the populous; AI is only possible to the extent of the large cloud companies. But if there are billions of mobile devices and billions of PCs with spare compute resource available that could be simultaneously hacked into / accessed by a single AI and used as a grid compute architecture... This then pales into insignificance when you consider the billions of servers in datacentres owned by businesses of all sizes that are sitting dormant when their main user base are asleep or not working. Soon you realise there is a huge amount of compute resource to match the neural network level requirement of 100 billions neurons and 1,000 synapses per neuron totalling 100 trillion logical gates to emulate just one human brain. And that's at a very simplified level, not taking into account the rest of the entire molecular structure of the body to emulate consciousness, genomics or even the inner workings of the brain; dendrites, receptors, serotonin / dopamine hormones etc.
It depends what we want the AI to achieve, neural networks and deep learning are breaking the ice, but emulating the brain may not be the answer to all. Even uploading or emulating consciousness may not be the answer. An entity that can feel pain would certainly turn quickly. AI has become a (somewhat inaccurate) term to cover such a large array of cerebral capacity, much of it we don't require; why would we need humanoid robots to cure cancer ? We wouldn't (require the humanoid part). Why would we need empathy in the plight to discovering renewable energy ? We wouldn't because we as humans and creators already own it. The question is what makes us intelligent in terms of information gathering, learning, disseminating, problem solving and implementing ? I could be very smart at learning languages, but very poor at building personal relationships because I have a low EQ (Emotional Quotient). The later wouldn't matter in the steps to intelligently find a health related cure.
Although I've witnessed Machine Learning programs that evolve to improve themselves, I've always found it very difficult to understand how computer code can think intelligently. Without defining "intelligent" I think you know what I mean. I've always thought some form of bio-tech would be required, which would be inherently flawed for obvious reasons. This was all until DeepMind's AlphaGo beat Lee Sedol at Go; there are 10⁸⁰ atoms in the universe and 10¹⁷⁰ positions in Go. Even the creators aren't sure how AlphaGo won; maybe Google are acting nonchalant so as not to give away their secrets but either way, it's an intelligent fete.
If AI can learn, evolve and essentially think for itself, what can it achieve ? Resolving or at least advising on, some of the major world issues such as; preventing world conflicts / wars or poverty ? There are 3 areas where AI could unequivocally help humanity no matter which side of the fence you are on. Although I want to write in more detail, for time's sake I will touch on these areas and the process the AI would have to follow; 1. Health, 2. Environment and 3. Space;
- Health; mapping the human anatomy and spirituality at the sub-atomic, molecular and quantum level.
Let's take Alzheimer's, something the majority of us start to concern ourselves with early in to our adult life because it affects so many of us, it's one of the saddest things to see someone you love no longer recognise you and the demonstrable effect this has on loved ones.
Most months I read about a new discovery in the fight against Alzheimer's. One very interesting plight is the HCP / Human Connectome Project which I've written about here; http://www.mackie.xyz/james-mackie-publishes-new-theory-on-h2h-consciousness-and-thought-communication/
The HCP's goal is to map all 100 billion neurons in the same way that we reverse engineered the Human Genome, aka the Human Genome Project.
The HCP has already discovered gene mutation and ladder like structures across axons in the brain displaying the early and late stages of Alzheimer's and autism. With 1,200 subjects, 36 scientists and millions in funding, we are still very far from a cure for Alzheimer's and reversing Alzheimer's, for now, seems like an almost impossible task. But in the words of Nelson Mandela "anything seems impossible until it is done".
Now if I were to set my AI bot on curing Alzheimer's it would first gather every single piece of related data both public and private. All of this data would be collated and analysed. AI would be able to use multi line inference and axiom theorems to capitulate experiential and theoretical conclusions (inference rules could be in the majority). At this point AI would have to soldier off on its own to spearhead science. A recent experiment showed that super human AI would learn in one week, what all of humanity has learnt in 20,000 years. If you consider compound and exponential growth, week 2 would be fairly exciting.
However ! My argument against my own subconscious reasoning as per the above has been, "well surely the AI would have to carry out the experiment on the human subject, and then await the results / success of the trials to gather the results". Maybe the AI would have to run clinical trials over a number of months or years on human subjects to garner the success of the induced medicine for example. But the answer is yes and no, in some cases no; the AI will be able to yield research from other areas (industries, sectors) / studies (people, screening) to instantaneously receive the result. In addition, the AI will also be able to run its own experiments based on simulations of the human body to extrapolate results.
But in some cases yes; because not all of the human organs such as the brain are fully understood by us or by machine, AI will have to run experiments on humans with new drug treatments that will work and inevitably not work, to monitor and analyse the results. In other words, the AI will have to fail many times before it's right. This will take biological time, no matter what speed the AI is expanding / forward time simulating.
That is until the AI has completely mapped humans at the almost spiritual level (cellular, molecular, quantum and anything we can't see such as consciousness / the soul / faith) because who knows if it's true that a positive attitude, good diet (+n) would prevent Alzheimer's for example. We need a prevention as well as a cure. So at some point the AI will have mapped the human at the subatomic, quantum, cellular and molecular levels so could run simulation based experiments without touching a human and this process could be sped up thanks to the process being digital and not analogue. In the same way modelling for asteroidal impacts or weather simulation is sped up, the AI would do the same.
At this stage, anything is possible in terms of curing the human body, be it Alzheimer's or the big C.
That's it for now folks, please reply in the comments section below, in part 2 and 3 of this blog I will cover the remaining 2 areas where AI can help humanity.