Autonomous Goals: Bringing Life To Computers
The rise of artificial intelligence is a hot topic, but what does the end game look like? How do computer systems come to life? What is the underlying cause for a human to think of a goal to pursue long-term? What causes us to sacrifice instantaneous dopamine for pain that leads to a better life? These are the questions that need to be answered in order to acheive fully autonomous intelligence.
import MailingList from ”@/components/blog/NewsletterForm”;
- How do we come up with goals?
- How do which goals to pursue over others (temporally)?
- How do we perform risk/reward assessment, e.g. doomscrolling vs studying?
- How do you decide to puruse a goal when the reward is unknown?
- How do you decide how much loss you should tolerate before opting out?
- When doing literally nothing with your eyes closed how do you decide to do something? I guess we all have a list of things to do but lets say we’re rich an don’t need to do anything. Then what do you do? How do you get to the point of wanting to do something?
- What neurotransmitters + neuromodulators are responsible for curiosity? Is dopamine the only factor?
- How is attention captured?
- How is attention retained?
- How is attention overriden by another thing?
- When we have multiple goals to pursue how do we prioritise one over the other? E.g. I enjoy studying neuroscience however math would be a more relevant and short-term gain than it.
Why autonomous goals?
To give an agent the ability to have curiosity to learn new things and relate them to existing knowledge to discover and live like a human. The aim is to remove human interaction entirely and let them live on their own like a human baby — experimenting through trial and error in the world to gain intuition and understanding. The goal is to survive and thrive (by learning everything to invent galactal conquest).
How do humans create goals?
Something we want to achieve that satisfies a desire that gives us a feeling. Without any feeling afterwards would there be any intrinsic motivation to even be alive? Maybe this feeling is the satisfaction of answering a question, the rush of a movie or talking to a pretty woman, being well known or even being safe. The objective is to satisfy the craving for a feeling or to extinguish it, whether its feeling pleasant or not. And this can mean different things for different people. The question is what causes this feeling of pleasantness?
What causes the sensation of feeling pleasant or not?
How are goals achieved?
We are aware of our bodies in physical space and out limitations. We focus on combining relevant topics to achieve the goal of advancing technology — cars to get to places fast on land, planes to travel to different countries faster than boats, the internet to share information, rockets to explore space. We become interested in a topic and figure out what is needed to succeed in said topic: for rockets you need to know about math to understand physics to build the program to run the rocket which requires in various types of engineering. What type of math is relevant to rockets? The fuel needed to create a thruster force to launch an
x heavy rocket into the atmopshere, etc.
To solve problem
x how do you know about all the relevant pathways to get to the solution? Lets think from birth. Initially we know nothing and we need to learn language to formulate concepts of things. From language we can understand text and speech to associate things together. Then we expand with school to learn math, history, biology, etc. These spark ideas that drive us to go deeper into the topics. The further we go the more we know in our mental ledger of what is possible. We ask more questions and attempt to solve them with the knowledge we have. If we don’t have the answers we ask more questions to try and find the answers. These questions act as a filters to remove irrelevant infromation and pathways so we can hone in on what is actually needed. The sequence continues until we reach the answer.
How do you decide is something relevant?
You have to think of the concept abstractly to remove all the low level details. For a car, we have a general model of what it is: a rectangle-like structure that has 4 weeks with an accelerator, a break and a steering wheel for the front two wheels for direction. What is the minimum amount of things needed for this model to be a function car? How do we know a piece of chicken isn’t relevant to achieving the creation of a car? In actuality it could be: to feed workers to make the parts — so there’s a spectrum of relevancy. So how do we get to the closest part of relevancy? By understanding the properties of materials and the thing being created. Can a frame that holds a car together be made of chicken drumsticks? Yes, but it’ll fall apart unless glued together. What about when we sit in it, it’s not structually strong enough to hold a human’s weight — so you make the base thicker. Then you run into the problem of the food will decay, the large “car” structure will be too big to enter roads and is inherintly unsafe due to structural integrity. So what are the alternative decisions to chicken drumsticks? Metal. You can bend it, it’s strong and can be a sound material to create the frame. So ultimately it’s probability of relevancy. Do the properties of chicken drumsticks hold up against metal? Not so much so we choose to explore metals and anything similar to it. Now we have this framework for a car we can start applying it to other structures like planes, buildings, boats, guns, tools, etc because of these properties. Ultimately, we determing relevancy by creating a model of what things are and their properties, what they can be most useful for, how they can be used, the cause and affect of using them in certain ways and then applying it creativly to ideas.
How are ideas and creativity created?
From where do we get our goals? Two possible places: internal sources — the mental image of who we want to be or our immediate desires such as finding something sweet to eat — and from external sources — the things other people would like us to do, or nonsocial cues from the environment (e.g, being in a library may help motivate to study).
Extrinsic goals are constantly re-represented as intrinsic goals in a process called “assimilation”.
When thinking about why we are motivated to do something now we think of money, status, power, fame, and to feel good. However, we have this compelling force nagging us to pick up our phones and doomscroll rather than going to the gym, learning math, distracting us from goals that actually impact our lives. Why is this? Why do we chase the instantaneous dopamine hit in favour of chasing beneficial gain?
We’ve all experienced it. Starting a task that illicits struggle is hard, despite knowing it will be better off for us. There is this barrier to entry that plays a massive factor otherwise the world would stop reading self-help books and actually do something meaningful with their lives to advance themselves.
motivation in absence of necessity, intrinsic motivation
Can be defined as the intrinsic desire to learn or obtain information. Without it we would never explore and answer questions about the world. Why incur significant costs in energy, effort, and time, to pursue an uncertain reward when our material needs have been acquired? Humans haven’t got unlimited time to learn everything as well so there is some intrinsic value assigned to all sources of information causing selective structured exploration that maximizes long-term advantage.
Share this Article