Who is right about the dangers of AI : Zuckerberg (Facebook) or Musk (Tesla) ?

Zuckerberg claims AI is and will only be an advantage to the human species, whereas Musk states that AI is dangerous and could actually become a threat to our species, even in its current form.

Like always, the answer is somewhere in the middle, especially in the short term. Most AI applications today (and tomorrow) will be ‘harmless’ and focused on ‘discovering your wishes’ and augmenting your comfort. In worst case it will be killer robots, like for instance Kalashnikov recently introduced. This looks worse than it is, at least in this stadium (see the question above 'Is advanced AI dangerous for mankind' to see Geert's reflections on this). That is where Zuckerberg is right.

But even with today's technology, wars could indeed get started due to making trouble and destabilization of (democratic) countries via networked robots. This is where Musk is right ! All these ‘attacks’ can, at present, still be countered by human and therefore it is however manageable and still largely ‘child-play’.

In its current state, ‘AI’ exists for 99 percent out of fixed procedures and 1 percent ‘real intelligence’. Most people do not understand that these ‘fixed procedures’ ARE actually a form of advanced intelligence ! Remember how your brain had to create procedures for basic mathematical functions like ‘adding’ and ‘subtracting’ ? You did not know this after just one day ! It took months to obtain those simple procedures in depth. It is fair to say that a procedure is the outcome of a lot of advanced thinking ! If these programmed procedures have ‘danger’ in them, applying just a spark of intelligence can make them lethal as no mechanisms are available to prevent that from happening.

At 2004, Geert played around with ‘words’ and ‘sentences’ to get a better understanding of the concept in our brain and he wrote a small program to do that. It was possible to let the program create sentences by itself based on a number of parameters, and it was extremely surprising how ‘realistic’ it turned out. It looked a bit similar like the modern virtual assistants Apple, Microsoft and Google offer. If he would have added just this ‘little splash’ of AI (the 1 percent) to this program, it would have been extremely realistic and hard to distinguish from ‘real conversation’. This is probably what Musk is referring to when he is talking about the ‘manipulative’ possibilities of today's AI.

Imagine what one can do with enough ‘dangerous’ and ‘advanced’ hard coded procedures and allowing ‘the system’ to ‘play’ with like it sees fit !

This is the point where Zuckerberg is wrong ! It can be dangerous, even with the technology of today !

In the long run, although most systems should be created to help mankind (and Zuckerberg is right), I agree with Musk and AI is not only a possibility but even more a likely threat to mankind !

Please read the question on if it is harmful for mankind to see Geert's reflections on that topic.

Artificial (General) Intelligence (AI) (AGI) and Artificial Brains (AB) , isn't that all the same ?

It is important to make the distinction between ‘AI’ and ‘AB’. A good comparison is for instance that it is not required to exactly know how cells and stem cells work to simulate the working of ‘a leg’. The same applies to ‘AI’ vs ‘AB’. ‘AI’ is, currently, very focused on how neurons work whereas an ‘AB’ is mainly focused on what ‘functions’ are behind some behavior or parts of a real brain.

Is one better than the other ? Not at all ! It depends on what you need. One can travel a thousand miles or kilometers by foot, but a car is probably the better way to achieve this. The reverse is true if one needs to go upstairs. Using a car to do so, is most likely very devastating for the home (in particular the stairs !) and it is obviously overkill.

So, using ‘AI’ in very simple forms of intelligence is good and acceptable. If it needs to be ‘real intelligent’ and safe – an ‘AB’ is normally the better solution !

Making a distinction between ‘AB’ and ‘Artificial General Intelligence’ (AGI) is often more difficult. First of all, creating an ‘AGI’ with the current technology available is possible and realistic and leans towards ‘AI’. Making it safe is the harder part ! In doing so, one will need ‘functions’ used in ‘AB’ and at that moment, it is more on the side of ‘AB’.

You keep saying it is better to base an intelligent system on human (behaviour), but if I look at history or even just at the news, I mostly see very cruel and destructive behaviour caused by human ! Isn't it therefore not better NOT to base it on humans, but on logic and facts ?

Intro

Wouldn’t that be lovely and it is a very good question !

The first part of the answer is about ‘facts and focus’, the second on humans ‘very bad’ and ‘naughty’ behavior !

Facts and focus

First of all, let's focus on ‘facts’. To have ‘facts’ one needs calibrated tools to measure, simple as that. A first thing to notice is that you cannot supply every ‘intelligent system’ with all possible measuring tools required to always be ‘basing on facts’. It is also not wise to assume the calibration will forever be correct and within margins !

Lets take for instance ‘vision’. You will find hundreds of optical illusions online and your system would suffer from this as well ! Just look at a simple fata morgana : is it water further up on the road or is it ‘heat-games’ ? Your system would have to go there and verify that because only facts (not estimations – a form of ‘believe’ and surely not facts) are allowed. This means your system, based on facts, not even statistics (an interpretation and/or interpolation of facts, not facts as such) would have to drive for miles and miles since the fata morgana always is ‘ahead’. It would be like finding the pot of gold at the end of the rainbow …

Every form of input is limited and therefore ‘the facts’ are limited and are no longer facts. It is even more confusing when more sensors are combined. Let's assume for this example that all sensors are functioning withing margins and we should be able to fall back on ‘facts’.

We don’t like the thing about the fata morgana in the above and add a laser for perfect 3D vision. Good of us, isn’t it ! At some point, we pass a transparent wall and the beam of the laser is, for some reason, reflected by this. For the laser, fact is that that is a wall. But behind this wall, someone is walking and our vision detects this as it can see through a transparent wall, but ‘vision’ will not see the wall as it is transparent and no reflections are detected for instance …

How will your system react on this ? Two different ‘facts’. Is it a wall and not a person (laser) or a person and not a wall (vision) ? Which of the two will it ‘believe’ or give it priority (not very factual, is it), since both at the same time are obviously impossible ? Procedures must be written and it is the programmer who will decide the priority. That is not intelligence, that is a lot of work for programmers and people with a background in statistics ! It is what you see happening right now with autonomous vehicles ! If they fail, and they do, it is 95 percent of the time in this department !

Secondly, ‘logic’. Logic is only logic if the rules are strict. In real life, the rules are not always strict, available in advance or even defined ! You cannot foresee ‘everything’ and for sure not the combination of ‘everything with everything’ ! See the above example for instance. The input of which sensor will it pick as correct if no rules are defined ? What will this conflict have as a result for your system ? Will it freeze ? If the facts don’t match, how can any logic be followed ?

Will it fall back on a ‘general rule’ in case of such ‘conflicting data input’ ? For instance a ‘general rule’ causing extreme caution and thereby possibly harming others by doing so ! Recently there was a car review in a Dutch car magazine on an autonomous driving German premium car. This guy was driving the car on a highway autonomously. Luckily for him, with very few traffic around him. All the sudden two different sensors noticed ‘something’ for a fraction of a second and the car started braking at maximum deceleration and spanning his safety belt as if an accident was eminent. For all clarity, factually, nothing was there ! Imagine that would have happened with more traffic on the road !

Let's go even a step further and say, like in real life, a dichotomy occurs. A dichotomy is an impossible logical situation. Simply said, it is two apparently opposite ideas or things happening at the same time. If you are walking left, you cannot be walking right (two opposites). Look at children and you will see a few of them. For instance, if it is raining and the sun is shining at the same time. The logic is completely flabbergasted at that moment. In their logic, it has to be either the one or the other, but not both at the same time ! The same happens when the sun and the moon are visible at the same time. In their logic, the sun stands for ‘daytime’ and the moon for ‘nighttime’. Both together therefore is impossible. In the same category, but this time for adults, just look at how a couple of hundred of years ago a total eclipse during daytime had a similar effect on adults.

Still not convinced ? Ask an astronaut in space if he is left, right, up or down of earth and he will have difficulty answering since one time ‘left’ is ‘up’ and another time ‘left’ is ‘down’ and yet another time ‘left’ might actually be ‘left’ ! A pure dichotomy and very realistic !

A final point on logic is that ‘the logic’ does not exist and that ‘your logic’ is by no means and surely by no definition the same as for any other person out !

In the digital world, all 'intelligent' systems will have a form of logic (procedures) because the computer needs these to function ! How many conflicts between male and female are due to a completely different form of logic ? Books, tv-shows and comedians all have heavily benefited financially from this, these are facts !

Human behaviour

Now, let's talk about humans and the path of destruction. First of all, it is not only human behavior to slaughter ‘enemy villages (and later entire countries)’ and burn them to the ground. Similar conduct has been documented for other primates like chimpanzees for instance. Pure ‘for fun’, a neighbor tribal is killed and even eaten.

But ‘humans’ are the only form of real and advanced intelligence we are acquainted with ! We know for instance that psychopathic people lack empathy. We also know that lots of politicians and business people fall into this category. We can study that, we have a couple of billion examples and objects and have been observing the behavior for over 100 years scientifically. But even thousands of years of information is available, starting from the Greeks …

Classifications have been made like for instance around 1930 the ‘DISC’ classification. ‘Dominant’ versus ‘Submissive’ vs others. It is clear that a ‘submissive’ personality will not all the sudden start attacking an entire group ‘just because’, something a dominant would sometimes dare ! A submissive personality rather requires a dominant to ‘give orders’. Look how Hitler was capable of mobilizing an entire nation and together they did what they did, not just one person …

We know a lot on ‘bad behavior’ for humans. We are actually better scientifically informed about ‘bad thinking’ and ‘bad wiring’ and the outcome of that than we are about ‘normal thinking’. What do criminologists, psychologists and psychiatrists (and to some degree even sociologists) all have in common : they look if it goes ‘bad’. They are specialized in the ‘wrongs’ of the human brain and interaction, both hard- and software ! (Geert Masureel)

If you have discovered the flows behind ‘the thinking’, like Geert Masureel did, you are able to use all this information and apply it in a predictable fashion. And yes, if you set the parameters of a Livinoid Intelligensia as if it should be a psychopath, you are asking for trouble ! But at least we know what the parameters for a ‘psychopathic intelligence’ are in our system.

As a matter of fact, any AI system without a form of empathy (for human species), is by default psychopathic in nature ! Lack of empathy is namely one of the main characteristics in defining a psychopath, like stated before …

An important question one should ask is : “Would You rather be confronted with something ‘like a human’ and ‘with empathy for humans’, holding a gun, or a killer robot with the brain of half a shrimp like Kalashnikov introduces recently, killing people based on a tag or ‘some other procedure’ you have no influence on” ?

You will most likely say : ‘By non of them as neither should exist’. Well, did Kalashnikov ask for your opinion when they made that decision and announced it … … So, please answer that question like if no other choices are available as that is regrettably the position we are at right now ! (Geert Masureel)

Not convinced ? At this very moment, we are overcrowding earth and over population is a fact. A ‘factual’ and ‘logical’ intelligence with no empathy for humans would (in best case scenario !) suggest what we do with wildlife. “So many of that species (in this case humans) must die to maintain and preserve the balance”. That is being factual and logical ! Hell, pretty much every scientist will agree that ‘such a measure’ is required if it is about rabbits and foxes or deers. Species most of these scientists don’t sympathize or empathize with apparently …

That all looks pretty much like what happened during World War 2 in Auschwitz and other German prisoner camps … … not something one would like to ever see repeated in human history !

Conclusion

To conclude, does our concept has potential danger in it ? YES ! But at least we have lots and lots of realistic information upfront available on it, even before writing one single line of code ! Equally important, we have better chances in preventing 'it' from being dangerous, if we wish to, because all this is realistic information and not only theoretical and mathematical !

How come you only need a couple of modern desktop computers to simulate a complete brain and even a complete human where it is currently not uncommon to simulate only a few neurons per computer ?

Actually, it gets even better. For Livinoid Physisia and Livinoid Emosia (without the thinking, but with the physics and emotions), dozens of these can run simultaneously on a modern powerful multi-core desktop computer ! The thinking process of Livinoid Intelligensia makes it heavier and therefor a couple of computers are needed !

The main, initial and single goal was to get an artificial brain (similar to the human brain) working on only a couple of computers and with such a strong focus, one looks for appropriate solutions and optimizations even in the (early) drafts. If it is too heavy in nature, you drop it and look for solutions that fit the bill ! Simple and easy : dare to fail and rethink !

One of the primary reasons why Geert Masureel initially stated that only ‘a couple of computers’ would be required, is the academical world ! Yes, since the 90’s there have been several professors claiming that the processing power and available memory of modern processors/computers was/is equal or better than what humans have.

His viewpoint was, and still is, that instead of doing such bolt statements about something that is not to be compared (apples versus oranges), produce something real ! Call it a middle finger to these guys (if you hadn’t figured out by now, Geert has this tiny rebellious side in him ...).

The main conclusion is that modern computers are extremely powerful IF used in the proper way. The techniques and concepts we use are based around this principle and therefore we can use every single bit and byte in our advantage. Using deep learning is almost as efficient as using a laptop as a spoon during dinner. With a little imagination, it works, but far better tools exist to have much better results …

Another reason was pure business minded. It is not good to create something that will use thousands and thousands of computers, as such a thing is unusable in the (relative) short term and that is unacceptable in Geert's opinion.

We should use our brains to provide solutions that fit our goals and not adjust our goals because we are unable to find solutions ! One has to dare to fail and rethink, rethink and rethink until you have a working solution within (or at least very close to) the original goals ! (Geert Masureel)

Wasn’t it Michelangelo who stated something like : ‘It is not failing putting your goals way too high and not reaching them, but it is failing putting them too low and achieving them !’.

What about other initiatives, like OpenAI, will you work together with them ?

We serve different purposes and therefore chances are high we will not work together !

The market is big enough for a couple of players and according to the website of OpenAI, one of their main goals is to provide safe ways of Artificial General Intelligence. We wish them the best of luck in their discovery as it is a noble goal !

The only thing Geert Masureel hopes for, is, that they are up to the task once it gets a little more complicated. At that point everything starts to influence everything and the isolation of “single thoughts”, however you create them, is no longer possible. That is where unpredictable and dangerous behavior will occur ! Something Geert can avoid in his concept as he uses (predictable and well documented) human behavior as the main technique …

Can something good come from this all for humanity ?

Geert Masureel, one of the founders of Artintelli, is truly convinced that it will, if done correctly.

It actually offers a number of solutions human mankind really needs the upcoming 30-50 years, apart from the possible threat of advanced AI itself !

Economics is the driving force behind prosperity in most countries and it is the main reason 'peace' is so durable in large parts of the world. Both peace and a growing world population just offers new markets and more profits ! No single multinational in the world will ever freely agree with a "zero-growth" economy and for sure not with long periods of shrinking markets ! Paraphrasing a President of the Unites States : "Is not gonna happen !".

Add to the equation the problem of aging of the population and the affordability of the pension system in a lot of countries, it is clear that birth policy is not a priority at all. China, for instance, dropped the 'one-' and even 'two-child' policy to overcome this problem. BTW, this latter is not working well, but that is a different discussion.

But it is this over populating that causes almost all problems and is a real threat for a lot of people ! Just imagine we would be able to half the world population ! It literally would solve pretty much every problem we face today as human species ! This latter is without doubt a scenario frequently discussed in some elite high society mediums, the offered solutions most likely are less 'plebs friendly' ...

So, to have a solution for the over population we need to make sure both 'the people' and 'the companies' are satisfied !

Robots and advanced AI could be the solution ! If robots can take over the holes in employment and taxes are raised on the labor done by these robots, the calm can remain in most countries. Of course, this implies that all people have a sort of 'replacement income'. Therefore, it is important that a fair share of the profit go to 'the people' and not only to 'the companies' ! Again, something an equilibrium for both parties can be found.

This latter is 'the bread' in "bread and circuses", the famous statement Julius Caesar made a couple of thousand years ago. If a meaningful occupation can be provided, depending on the type of people, the 'circuses' part of Julius' saying is fulfilled as well and 'the people' should be 'happy' and 'content'.

The second part is the one of the companies. If a large enough consuming market can be provided and potential for growth is present, it should be fine. People with more spare time, will consume more (if money is available) and that is only the initial effect. Growth could be available in allowing the robots/AI to become part of the consumer market. We are not the first to state this and if done well, it would offer the solution to keep the industry pleased. A 'conditio sine qua non' for success in this part of the equation !

And what about politics ? Well, modern politics is largely driven by lobby influences from the industry on one hand and 'the needs of the people' on the other hand. After all, it is the people who need to vote on these politicians and it often is the industry providing money (sponsoring) or resources (employment) helping the politicians to get voted. The above offered solution should therefore be good for politicians as well.

At that point, a global 'one- or two-child policy' should be enforced by an organ like the United Nations !

IF, and only IF, we are capable of colonizing other planets or 'space cities', an increase in births should be allowed to fit that purpose. People with the need of having more children, must be able to find a good and realistic virtual alternative, something Livinoids can provide in the very near future ! A Livinoid Emosia could grow from a baby to adolescents (and even further) at real life pace (or faster, if wanted).

If more and more 'products' can become virtual (look at virtual reality for instance) and if the world population can shrink gradually, this would be a enormous help for the environment and all related problems ...

... so, yes, AI/Robotics can be an extremely important positive influence on humanity !

I'm a realist and I perfectly see what is happening in the world. Currently a lot is not to be very proud off and we all should be collectively ashamed ! BUT, I'm also a dreamer and why not dream of making it a better place for our children and grandchildren instead of using our ingenuity only for short term visions, financial gain, arrogance and vanity ... ... one day the bill will be presented for that egocentric (also egocentric as a country) and currently socially acceptable behavior.

 

Ask Weinstein or other (sexual) abusers of power how it feels today, being spitted out by the world for something that apparently was socially acceptable 30 years ago. This could be you in 20 years for doing what you do today ! Holding the moral standards high in regards with the future of our blue planet and not being such an ass is usually a good starting point ... (Geert Masureel)