Mind Blowing Interviews with Joscha Bach on Life, the Universe, and Everything!
Everyday in the morning, when I wake up, I have a mind blowing experience.
I open my eyes, and the world looks like a bunch of fuzzy blobs. Then I put on my glasses and suddenly, everything comes into crisp focus. Things that were just fuzzy and vague outlines become sharp and clear and vivid. The whole world suddenly becomes a lot more meaningful and interesting. It’s a beautiful feeling and I think I am addicted to it.
Which is why, I think, I am constantly on the lookout for crisp insights into pretty much any topic. There are many things in life, the universe, and everything that I have thought long and hard about, but have managed to comprehend only vaguely. Just like those fuzzy blobs I see when I open my eyes in the morning. But, once in a while, I come across someone who throws a bunch of insights at me in rapid succession, and suddenly, some of those fuzzy blobs turn into crisp and clear and vivid forms. I absolutely cherish such moments, such insights, and such people.
I call such people and their insights Deep Turtles. One such person I came across recently was Dr. Joscha Bach. I watched his interview on YouTube taken by Lex Fridman (linked below) and I was instantaneously hooked. I could immediately sense that I was in the presence of a Deep Turtle and I watched the whole interview with awe. It set my brain absolutely on fire! Then I watched his other interviews (also linked below) and finally decided that I should try to summarize some of the things that struck me as deeply insightful.
Some of what follows is directly based on things he says in the interviews, but I have also explained some things in my own words and added some of my own related ideas.
Intelligence and Sentience:
There are many definitions of intelligence, but Joscha uses a definition that emerges out of AI, or computation in general. He defines intelligence as the ability to create models. A model here is basically a set of parameters and relationships among them that are isomorphic with some real world entity or phenomenon. Intelligence, when presented with a set of patterns, is able to see a structure to them, represent them as models, and predict the next set of patterns. The main benefit of this definition is that it is more concrete and precise than other definitions and thus useful for doing further analysis.
It also goes without saying that the higher an entity’s intelligence, the better its ability to create more and more complex models in more and more domains and with higher fidelity.
Joscha also defines “sentience” as the quality of an entity possessing such models, including a model of itself in relation to its world. And the larger, more complex, more comprehensive, more effective someone’s library of such models, the higher their sentience.
Note that, for someone to be sentient, it does not matter how they acquired such models. They may have been born with them, or learned them by imitation from a parent or a mentor, or through their own experimentation and experience or just mediating hard about reality.
Also note that someone may be more sentient but not necessarily more intelligent. And the converse can also be true i.e. someone may be highly intelligent, but possess a lower level of sentience.
Why does this matter?
I think definitions such as the above are very useful because they allow you to break intelligence down into its components, visualize how it works, and even try to recreate it in a machine. This allows you to test your ideas and improve upon them. Pretty much no other definition allows this sort of scientific analysis.
For example, the above definitions helps us see that our ability to effectively navigate reality depends primarily upon our sentience, not on our intelligence. As long as we have an adequately comprehensive and precise library of models required to navigate the part of reality we are in, we can be successful. Even if our intelligence, i.e. our ability to discover models, is limited. (For example, a person with a map will be more effective in navigating a city rather than a more intelligent person without a map.)
Of course this relationship may change if the reality you are inhabiting is dynamic. In a dynamic reality, one needs to constantly update the models needed to navigate it. So in a more dynamic world, the correlation between intelligence and success in life should improve.
This also explains the importance (and shortcoming) of education as a way to become successful in life. Education is the process of transferring models from teacher to student. Given that human ignorance of reality is vast, even an extremely intelligent person will never be able to discover all the models of reality that would be needed to become successful in general. At the same time, if the world is very dynamic, no amount of education will suffice to navigate it successfully unless the person is very intelligent.
I think these definitions help one think more clearly about intelligence and its relationship to success in life. Also, the definitions should be useful in building AI and AGI, which brings us to the next topic.
Turing Test and AGI (Artificial General Intelligence):
We all know that the Turing test is a test to see if an entity has human level intelligence. But according to Joscha, the true Turing test is to ask a system “what is intelligence?” If the system can answer that convincingly, then it is intelligent. According to him, Turing (probably subconsciously) designed the test because he wanted to see if he was generally intelligent i.e. an AGI himself!
The intuition behind this is that we will need to understand what intelligence is in order to build AGI. So the way we, as in human beings ourselves, pass the Turing test is by building an AGI ourselves and thus being able to answer the question “what is intelligence” convincingly.
Thus, the entire enterprise of AI is ultimately in pursuit of passing this test ourselves. Most people tend to think that we are building AI / AGI because we want to automate various tasks and scale them up and make our lives easier in some way. But in my opinion, developing AGI is more of an intellectual endeavor.
Ultimately, everything we do in science and technology is, at its core, in pursuit of answering the deep questions of who we are and what reality is. All these goals like making our lives simpler or making people happier or improving the economy etc. are just carrots and sticks we hold in front of ourselves to motivate ourselves.
Why does this matter?
The development of AI / AGI is one of the two most important and far ranging philosophical projects ever taken up by human beings. (The other equally important philosophical project is Physics.)
Helping us clearly and crisply understand the physical universe down to its lowest constituents, and the universe inside our brains down to its lowest constituents (and, moreover, be able to recreate it inside a machine) has far ranging implications on all of human philosophy and spirituality.
Once you start thinking about AI and philosophy, you of course come to the question of consciousness.
As we saw earlier, intelligence is about acquiring models, running simulations, making predictions, and decisions based on them. So basically this involves creating simulations of reality within our brains.
A generally intelligent system needs to solve general problems. Some of the problems are so general that the system needs to model itself in relation to its environment. For example, a child becomes aware of its own psychology because it notices it does certain things despite it noticing itself wanting different things. It then realizes that it has complex structure within itself and if it wants to predict its own behavior, it needs to model itself in relation to its environment. This is the first step towards consciousness.
The brain also needs a mechanism to focus its attention since there are many things going on and it needs to focus on things that are important. And, this attention mechanism also needs to focus on its own simulation of itself. According to Joscha, this is what we usually experience as consciousness.
Also, it should be easy to see that such a consciousness can only emerge inside a simulation. It happens when an intelligent agent tries to model itself in relation to its environment and pays attention to it (and also pays attention to it paying attention to it).
Another way of putting it is that you are a biological machine that thinks it’s you!
Why does this matter?
This line of thinking is extremely promising because it illuminates a path towards developing artificial consciousness in AGI. And as he says earlier, such attempts at creating an AGI might help us understand ourselves, including our consciousness, better. Thus, if we are able to implement such a mechanism in an AGI, and it wakes up and tells us that it is conscious, and it can debate effectively with us on this topic, then we would have to agree that it is conscious. IMHO, this represents our best hope in understanding consciousness thoroughly. Every other approach that we have come up with so far involves some amount of hand waving or appeal to authority.
Why stop there? Let us see what he has to say about the most commonly asked question about life!
The Meaning of Life
The scientific way to think about the meaning of life is to look at the scientifically discovered unit of life, i.e. the living cell. The first cell never died, it only split and continued to split and every cell in our body can be thought of as an instance of that original cell. You could think of this entire collection of all the cells that originated from that first cell as a giant hyper-organism.
A cell consists of a negentropy extractor, a replicator, and a Turing machine, and it participates in evolution. And a hyper-organism also has the same characteristics and wants to do the same thing i.e. survive by extracting negentropy, replicate, and participate in evolution. And since all of us are a part of that hyper-organism, we are instinctively compelled to help in this process.
When the search for finding meaning in our lives leads us to want to belong to something bigger than ourselves, as it most commonly does, we are really talking about belonging to this hyper-organism and helping it to survive and replicate and evolve. So this leads to a science-based way of describing the meaning of life.
All other creatures do this instinctively. But human beings are more complex. Our brains are programmable through “programs” i.e. thoughts / ideologies / laws. Religions, philosophies, and governments have used our need to belong to something bigger than ourselves and the ability of our brains to be programmed to their advantage. All of them basically create programs that are downloaded into a large number of brains to make them march in lock step.
We evolved this because of the Prisoner’s dilemma. Prisoner’s dilemma tells you that the best outcome for everyone can only be achieved if everyone sacrifices a little and trusts each other. In the old days, this could happen naturally because communities were small and everyone knew each other. So people could notice that someone wasn’t sacrificing sufficiently for the overall good and that person would be ostracized. But once societies became too large, this could not happen. People could get away without sacrificing for the overall good. And trusting people also became a problem.
Hence we evolved governments (and religions). They are basically agents that offset everyone’s payoff matrix to make their local Nash equilibrium compatible with the common good. (This is a mathematically precise way of saying that governments and religions make everyone sacrifice a little in order to help their common good, even if their incentives tell them to not do so.) And they help establish trust among strangers by allowing them to state their allegiance to the common government or religion. The addition of law enforcement makes the picture complete.
Why does this whole thing matter?
Now, I know that some people might ask why are these ideas and conclusions any better than any of the other ideas and conclusions that people have come up with for millennia. Take any corner of the world and you will find sayings and books and doctrines full of answers to these philosophical questions that people have been asking for ages. Entire religions and cults have been created based on them and many of them have millions of happy followers. So what’s the big deal here?
The point here is that the ideas mentioned above are more precise and more science-based or evidence-based than most similar ideas that you may have heard before. They can be explained step-wise from the very basics of science and mathematics, and some even have the potential to be experimented with in reality so they can be verified. They aren’t asking you to believe some authority or magic. They aren’t asking you to just rely on your intuition either. Plus, you can actually conduct experiments to verify the concepts and build useful real world mechanisms based on them.
An analogy is that of the law of gravity. People have been seeing things fall down to earth for millennia. Most people just took this phenomenon for granted and didn’t ask any further questions. Most of those who asked “why?” were given some answer like “that’s how God created the world” or something to that effect.
Then, one day, Newton came along and thought about it deeply and discovered the phenomenon of gravity. Not only that, but he figured out the formula for it. And, once he had discovered the formula, other people were able to understand it, verify it, and take it further to discover various other phenomena and conduct various other experiments that eventually led to us sending rockets into space that land on other worlds! None of this could have happened if we had just accepted the answers we had thought of before Newton. The importance of doing this sort of science-based rigorous analysis is that it makes it amenable to more widespread belief, further analysis, and inventions.
For a long time, questions such as what is the meaning of life or what is consciousness or even what is intelligence etc. have been treated as being part of the realm of philosophy or religion. Or, whenever they have been thought of in scientific terms, they had rather unsatisfying or imprecise answers. Some people even went to the extent of saying that science is fundamentally nihilistic.
But I believe we are at a point where these questions can be brought into the realm of hard science. I believe we now have a way to bridge the gap between religion / spirituality and science and advance both of them further. I see these ideas as making really great inroads in that direction.
Finally, here are the interviews from which most of these ideas have been taken: