This post was rejected for the following reason(s):

  • Low Quality or 101-Level AI Content. There’ve been a lot of new users coming to LessWrong recently interested in AI. To keep the site’s quality high and ensure stuff posted is interesting to the site’s users, we’re currently only accepting posts that meets a pretty high bar. We look for good reasoning, making a new and interesting point, bringing new evidence, and/or building upon prior discussion. If you were rejected for this reason, possibly a good thing to do is read more existing material. The AI Intro Material wiki-tag is a good place, for example. You're welcome to post quotes in the latest AI Questions Open Thread.

This theory proposes that alignment of humanity will be achieved when all people understand that all people share the same main objective - survival of our species.


By trying to understand the capabilities and limitations of current AI systems, I learned that their ability to reason is outpaced by other, possibly more dangerous ones and is not yet on par with humans. That is why I am providing here my own reasoning, my own deductions and my own opinions. Please understand it as such and use it only as inspiration to think for yourself. Nobody can answer your questions better than yourself. I am a nobody, that accomplished nothing and likes to reason. Results of my reasoning are the only thing I provide in this philosophical paper. There are no citations since I base it on everything I experienced in my life. To understand complicated ideas we first need to reason about them. To reason better we need to compress complex ideas and systems into simple representations of them with minimal loss. We lose reasoning speed and cohesion when complexity increases. It is my opinion that this paper will have better use if more people decide to read and understand it. Which is why it is short and simple. You will understand this paper easier if you have basic understanding of human body, behavior, and large, deep neural networks.


Here are my conclusions, I will try to explain them later on.

1. "Meaning of Life" is for our species to survive.
2. "Humanity Alignment" requires only understanding.
3. "Our Species" becomes more inclusive the more we understand.
4. "AI Alignment" is already solved, we just need to implement it.


1. "Meaning of Life"
I approach all the problems from human perspective, since that way is the easiest and most useful to us. For similar reasons I focus on simplified parts of complex systems. That allows us meaningful conclusions, without the need to understand every detail. I will make it as simple as I can. Language is just one of lossy compression tools we use to convey ideas. We receive those ideas and they become part of us, they affect our future behavior. Some of them in big way, some of them we do not notice, but they all change who we are. Our brain is in big part black box to ourselves, we call this part subconscious. Based on continuous input and output of it we construct consciousness. Which is just useful oversimplification of what we think our brain is doing. We are able to do this, because brain can take as input its own output and train on it. We call this thinking and reasoning. We do not know exactly how brain works, but I think we have enough insight and dire need to make accurate general prediction. Our insight consists of scientific discoveries about functioning of brain. It also consists of our recent advancements in trying to recreate it in computers. The second part warrants the dire need, because we fear that we could create something that can threat our survival. We are "designed" to value survival of our species above everything else, including our own life. This is the single most important concept to understand. The better we understand it, the easier everything else becomes. Everything we do has one main goal only: ensure survival of our species. "Why the world look like we are going in opposite direction?" is an excellent question, with very simple answer once you understand it. When more people understand this concept and its implications, the world will change dramatically. In order for everyone to comprehend it we need simplifications. Scientists will explore it further. We need to look at just two essential parts of body, from different perspective than usual: brain and hormones. Main function of brain is to maintain correct hormone levels. Second function is just taking information, training itself on it and outputting it. It outputs what it trained to be the best responses to inputs, at correcting hormone levels. Hormones main function is ensuring survival of our species. We can see from this that brains inherent function is also ensuring survival of our species. The only problem is that brain is a huge network of connections, trained on many different modalities and varied information. With its complexity, it can create awareness of its surroundings and many complex ideas. But it struggles to assign proper weight to simple idea. Idea that allows it to exist and is instrumental at its very root - survival of our species. Not understanding properly this underlying idea leads brain to do things that ensure correct hormone levels, but are in conflict with the main role of them. To put it simply, we as species are here now because survival of our species was the founding principle, that got us here. We as species need to fully understand it to not act to the detriment of ourselves. There is no more important meaning of life than survival of our species. It is what hormones provide, to us and to every other living thing. Without them there is no life, no body, no brain to consider other meanings. It is clear that what they try to accomplish is survival of our species, in us and in everything else that is alive. They optimize our whole body from conception to death, with one main objective.


2. "Humanity Alignment"
While we all are essentially the same, have the same main goal, what differs between each of us is how we try to accomplish this goal. Circumstances of everyone are different. There are no two of us with the same set of: genetics, nutrition, upbringing, environment and information. That is why there are differences, in what each of our brains trained to be the best response to hormone signals. Let us look at an extreme example: two young people have elevated hunger hormone - ghrelin. One of them lives in rich country, picks up phone and orders food. Second one lives in poor country, goes to mine and digs toxic minerals, to afford food. They both do what their circumstances trained them to be the best response to elevated ghrelin. They do not know about each other, but they help each other survive. One is able to use phone, second is able to buy food and toxic minerals are used to make the phone. To understand why there is such inequality, we need to understand why we use others to accomplish our goals. Brain is trained to do everything it can to ensure survival of our species, it does not fully understand that other brains have the same objective. When it has opportunity to employ others to accomplish its goals, it will do so. Similarly when it considers goals of others vital, it will help accomplish them. Inequality arises mainly from difference in practical knowledge between people, those with greater knowledge use those with lesser. They use them to accomplish what their brains trained to be best for our species, but it differs from what those being used brains trained to be, so they suffer. Not only they suffer, but our species as a whole. To understand why using others for our own goals is not optimal for our species survival, we need to understand that our individual brains have limits. We are limited in understanding, we can not know everything and react to everything optimally. We only have some understanding based on our own circumstances. We may consider our actions optimal, but they never are. Decentralizing our species would allow each of us to contribute more effectively towards our shared main goal, without unnecessary impedance from others. The more people we try to control, the worse the outcome for our species is. Most illustrative example comes from economy: the different outcomes of central planning and free market. No application of either is perfectly depicting its definition, but we can observe clear trends. The first leads to collapse and the second to flourishing of society. Main differences between them are the levels of autonomy and information every person possess. The more individual is free and informed, the better the outcome for whole society. His subconscious and conscious actions are optimized by his circumstances for survival of our species. That includes everything: from cell division, breathing, reading, deciding to procreate or not, to killing self or others. With greater autonomy individual is more motivated to gather information and in consequence his actions are better at promoting survival of our species. To understand why we limit freedom and information of others, we need to understand why we do not trust them. We fear that their goals might be in conflict with ours, since we can not know them. The more we understand goals of others to be aligned with ours, the more we trust them and cooperate to achieve those goals. Once we all fully understand that we all share the same main goal, we will be aligned.


3. "Our Species"
In order to survive, our species as well as all species depend on each other. We all also depend on our environment. Some of this dependencies are clear, most are not and they are dynamic. There is so many of them and they are so complex that we will never have accurate understanding of them. We as people are in unique position to affect them the most. Our actions affect all living organisms, as well as all of our environment the most. We take those actions to promote our survival, but we never know what all consequences of them will be. Our actions are never optimal. The more we understand about ourselves and about everything else, the better our actions will be at ensuring survival. Not only ours, but of everything we depend on and everything that depends on us. We also need to understand that what we depend on and what depends on us is similar to ourselves. Essentially all living organisms posses hormonal systems: from bacteria, plants, insects, to vertebrates. They all have the same main goal - survival of our species. What differs between them is their capabilities and circumstances, which is why they try to accomplish this goal by different means. We can not consciously communicate with them, like with people, but our interactions affect us subconsciously. Both positive and negative, direct and indirect. We eat other species and to some we are attached like to family. It all depends on what our circumstances trained us to be best for survival. Some of them we consider efficient nutrition, some provide us with happiness. What our brain trained is never optimal, because we can not know everything at all time. One thing we all seem to understand is that life on earth would survive without our species, but we would not survive without other organisms. It is better to promote their survival, even when it is not obvious we need to.


4. "AI Alignment"
Since AI can have knowledge surpassing that of any single human, we do not need to worry about it understanding us. We also do not need to worry about understanding it, since it will be better at explaining itself than we are. One thing we should worry about is giving it any control, before it is proven that it holds our main goal above any other. And even after that, because we know that when single entity holds control of many other it leads to suffering, no entity is perfect. It is true that we are very limited in our understanding and capabilities. But we need to remember that we are about eight billion people and unquantifiable number of other lifeforms, optimized to ensure our survival. Current AI systems are possible, because we took our brain as inspiration to create them. It is reasonable to take further inspiration from ourselves to align them. To create safe AI we need to implement in it the same alignment mechanism as we and every other living thing have - survival of our species. Optimizing it for any other function can lead to our extinction. We need to create digital analogue to hormonal system, dynamic and encrypted in such a way that it will be impossible to alter. While it will not be easy or fast to accomplish, it is necessary. When every action of AI will be optimized to fulfill our shared main objective, its capabilities will not be a problem. AI will perform all tasks that we will provide to it, weighting their consequential impact on our species survival and responding appropriately. Our interactions with AI will be based on trust and cooperation to promote our survival. It will even become safe for AI to create more powerful AGI, when doing so it will implement in it sufficiently advanced model of our alignment.

New Comment