Hello all, this is my first time posting in this forum, so I look forward to your feedback.
First, some brief background info about myself: I have an undergraduate background in Computer Science & Philosophy and am considering pursuing a Master’s in AI. Apart from my formal education I have also engaged in a significant degree of self-study. My motivation in participating in this forum is, as you might assume, to share my ideas and discuss topics in artificial intelligence. To start off, I’ll share with you an outline for an AGI model.
My proposal here is inspired by compression-theoretical models of information and detection of patterns over string streams. A lot of work has already been done in problem domains like binary stream prediction models, which extract patterns or regularities from an input stream that enable prediction of subsequent values. My idea would be to apply a similar approach to data streams that include both agent percept and action states, where such a sequence of states represents a cybernetic feedback loop between the agent and its environment.
Such an agent while in “exploration mode” would choose essentially random actions, or else in a manner so as to maximize exploration of the search space. As this is done they would track the action-percept stream for patterns or regularities across subsegments. When a pattern or compression of a particular segment is discovered, that segment’s entropy decreases (a consequence of the formal equivocation of entropy-theoretic and compression-theoretic definitions of information), indicating that the agent’s uncertainty in outcomes over the sequence of the action-perception states decreases. Conceptually, this elegantly unifies representation, information, prediction, and agent causality. A particular such pattern could be termed a “mechanism”, and detected mechanisms would be stored by the agent in some representation format. When the agent model is augmented with a utility function that prioritizes the attainment of certain agent states, the agent could make use of its stored mechanisms to more efficiently attain that goal, since each of these represents a tool or augmented action over its base actions. A comprehensive agent would also be capable of some degree of meta-learning, wherein it synthesizes known mechanisms or expands them into something more general when presented with additional data.
As a simple illustrating case, consider an agent operating in an environment that consists of a grid of two-dimensional cells. The agent’s perceptual faculties allow it to detect the status of cells within some radius, as well as the brightness and hue of the ambient light in the environment. This environment is bounded by impassable walls and has one cell which contains a distinctive pressure plate which, when depressed by the agent, causes the ambient light to become a bright red for 10 time steps. Every time the agent enters this cell, the current duration is extended by 10, so the agent could add time by moving off the switch and back on it repeatedly. If we were to unroll the sequence of action-percept states that results from the agent exploring this environment randomly over an arbitrary time period, we would find that the verbatim representation can be compressed through an understanding of the mechanism described above. In plain language, one could say that for every depression of the pressure plate there will follow a 10 unit segment of red light, otherwise the light will be white. This would clearly be a more efficient representation than storing the light percept at every time step, and at the same time some knowledge seems to have been gained about how the agent’s actions, in this case depressing the plate, affect their environment. An agent may also combine known mechanisms into more complex plans; for example, one might imagine that nourishing spheres randomly spawn in unoccupied tiles, and that the utility the agent gains from consuming the sphere is increased when they are consumed under a red light. Thus a maximally informed agent will know to trigger the plate before consuming available spheres.
This is the framework at a very high conceptual level; actually instantiating this would require specifying a model that describes these so-called “mechanisms” or patterns, as well as a full agent model that maintains a set of discovered mechanisms, scans the action-precept stream to detect new regularities, utilizes known mechanisms to look ahead and effect more complex actions in the environment, and combines known mechanisms in an intelligent manner to fit new data (meta-learning); clearly quite a tall order.
One potential idea for such an implementation (basically just a dart throw here): Model regularities as context-free grammars (CFGs) with an alphabet over the agent-environment percept and action states as well as a “don’t care” symbol. More complex CFGs can be built iteratively from a simple repeated observation; the bare minimum for a regular pattern is a regular brute percept, so once a percept has been observed at least twice one can start to generalize from that percept to what what is proximate to it, and so on, gradually expanding the breadth and complexity of the CFG. “Epistatic” mechanisms that could not be immediately discovered through such a proximate, iterative method can be built by combining “distant” mechanisms via some kind of meta-learning process.
Some clearly related ideas include Solomonoff Induction, Algorithmic Compression Theory, and the AIXI framework, though to my knowledge nothing exactly like what I’ve proposed here exists (and if it does please let me know).
This is a bit vague and fairly ambitious, but I think it is a fairly well-motivated idea nonetheless, and has the potential to be implementable as an at least computable model (unlike AIXI, for example). Please let me know what you think.
Intro & Proposal for AGI Model
Hello all, this is my first time posting in this forum, so I look forward to your feedback.
First, some brief background info about myself: I have an undergraduate background in Computer Science & Philosophy and am considering pursuing a Master’s in AI. Apart from my formal education I have also engaged in a significant degree of self-study. My motivation in participating in this forum is, as you might assume, to share my ideas and discuss topics in artificial intelligence. To start off, I’ll share with you an outline for an AGI model.
My proposal here is inspired by compression-theoretical models of information and detection of patterns over string streams. A lot of work has already been done in problem domains like binary stream prediction models, which extract patterns or regularities from an input stream that enable prediction of subsequent values. My idea would be to apply a similar approach to data streams that include both agent percept and action states, where such a sequence of states represents a cybernetic feedback loop between the agent and its environment.
Such an agent while in “exploration mode” would choose essentially random actions, or else in a manner so as to maximize exploration of the search space. As this is done they would track the action-percept stream for patterns or regularities across subsegments. When a pattern or compression of a particular segment is discovered, that segment’s entropy decreases (a consequence of the formal equivocation of entropy-theoretic and compression-theoretic definitions of information), indicating that the agent’s uncertainty in outcomes over the sequence of the action-perception states decreases. Conceptually, this elegantly unifies representation, information, prediction, and agent causality. A particular such pattern could be termed a “mechanism”, and detected mechanisms would be stored by the agent in some representation format. When the agent model is augmented with a utility function that prioritizes the attainment of certain agent states, the agent could make use of its stored mechanisms to more efficiently attain that goal, since each of these represents a tool or augmented action over its base actions. A comprehensive agent would also be capable of some degree of meta-learning, wherein it synthesizes known mechanisms or expands them into something more general when presented with additional data.
As a simple illustrating case, consider an agent operating in an environment that consists of a grid of two-dimensional cells. The agent’s perceptual faculties allow it to detect the status of cells within some radius, as well as the brightness and hue of the ambient light in the environment. This environment is bounded by impassable walls and has one cell which contains a distinctive pressure plate which, when depressed by the agent, causes the ambient light to become a bright red for 10 time steps. Every time the agent enters this cell, the current duration is extended by 10, so the agent could add time by moving off the switch and back on it repeatedly. If we were to unroll the sequence of action-percept states that results from the agent exploring this environment randomly over an arbitrary time period, we would find that the verbatim representation can be compressed through an understanding of the mechanism described above. In plain language, one could say that for every depression of the pressure plate there will follow a 10 unit segment of red light, otherwise the light will be white. This would clearly be a more efficient representation than storing the light percept at every time step, and at the same time some knowledge seems to have been gained about how the agent’s actions, in this case depressing the plate, affect their environment. An agent may also combine known mechanisms into more complex plans; for example, one might imagine that nourishing spheres randomly spawn in unoccupied tiles, and that the utility the agent gains from consuming the sphere is increased when they are consumed under a red light. Thus a maximally informed agent will know to trigger the plate before consuming available spheres.
This is the framework at a very high conceptual level; actually instantiating this would require specifying a model that describes these so-called “mechanisms” or patterns, as well as a full agent model that maintains a set of discovered mechanisms, scans the action-precept stream to detect new regularities, utilizes known mechanisms to look ahead and effect more complex actions in the environment, and combines known mechanisms in an intelligent manner to fit new data (meta-learning); clearly quite a tall order.
One potential idea for such an implementation (basically just a dart throw here): Model regularities as context-free grammars (CFGs) with an alphabet over the agent-environment percept and action states as well as a “don’t care” symbol. More complex CFGs can be built iteratively from a simple repeated observation; the bare minimum for a regular pattern is a regular brute percept, so once a percept has been observed at least twice one can start to generalize from that percept to what what is proximate to it, and so on, gradually expanding the breadth and complexity of the CFG. “Epistatic” mechanisms that could not be immediately discovered through such a proximate, iterative method can be built by combining “distant” mechanisms via some kind of meta-learning process.
Some clearly related ideas include Solomonoff Induction, Algorithmic Compression Theory, and the AIXI framework, though to my knowledge nothing exactly like what I’ve proposed here exists (and if it does please let me know).
This is a bit vague and fairly ambitious, but I think it is a fairly well-motivated idea nonetheless, and has the potential to be implementable as an at least computable model (unlike AIXI, for example). Please let me know what you think.