Terms and Concepts
Terms and concepts commonly used in discussions of AI and AGI as well as some specific to the SynthCog blog.
A software/hardware platform that employs artificial general intelligence technology.
One who is skeptical that machine intelligence at human-level or beyond will be possible until sometime in the distant future if at all.
One who is alarmed by the possibility of human-level or beyond machine intelligence and believes that it will pose a potentially existential threat to humanity.
One who is cautiously optimistic about the possibility of human-level or beyond machine intelligence and the potential benefits it may provide at some point in the future while keeping in mind the potential pitfalls and dangers that may come with it.
A software/hardware platform that employs AI technology.
One who is excited about the possibility of human-level or beyond machine intelligence and believes it will provide an extraordinary bounty for humanity.
Artificial General Intelligence (AGI)
The field concerned with the creation of machines that approach or surpass the level and quality of human intelligence.
Artificial Intelligence (AI)
The contemporary field that includes machine learning and symbolic logic and typically involves techniques such as the statistical analysis of massive data sets, weighted evaluation networks, self-modifying feedback loops, and hierarchical relationship networks.
A concept of rational decision making that recognizes the constraints inherent in any real-world situation in which constraints will exist on cognitive ability, knowledge, and time.
The mental action or process of acquiring knowledge and understanding through thought, experience, and the senses (as opposed to intelligence, which is the result of those processes).
A software or hardware module that is capable of synthetic cognition. An AGI system would employ a cognition engine to exhibit intelligence.
A being, biological or synthetic, that is capable of cognition and exhibits intelligence.
Conclusion of Convenience
The conclusion that the outcome of an advantageous conjecture will occur within the lifetime of the person making the conjecture.
The concern that an AGI system, particularly a superintelligent system, would be difficult or impossible to control, contain, or terminate and would instead attempt to control, contain, or terminate others to preserve itself.
Cooperative Inverse Reinforcement Learning (CIRL)
An AI learning model involving a human and an artificial agent in which the agent is not given the reward parameters of its utility function but instead must infer them from observing the human. The intent is to maximize the realization of human values.
The direct manipulation of an organism's genetic code using technology.
The drive that motivates an intelligent agent to prevent alterations of its present ultimate goals so as to ensure that those goals are more likely to be achieved by its future self, whatever form that future self takes.
A model of intelligence as a phenomenon based on attaining goals and governed by an algorithm designed to maximize the attainment of those goals. GOUFI is an acronym for Goal-attainment Optimization driven by a Utility Function as Intelligence.
Instrumental Convergence Thesis
The conjecture that for a wide range of potential ultimate goals pursued by an intelligent entity, we can identify a number of subgoals that are likely to be pursued. These subgoals have the potential to be detrimental to humanity even if the ultimate goals themselves are harmless.
An intermediary subgoal formed to facilitate achieving an ultimate goal. There may not necessarily be an obvious correlation between an instrumental goal and the ultimate goals it facilitates.
Rationality that is confined to some subdomain of endeavor or circumstances.
Intelligence is that quality which allows an entity to solve a wide range of deductive and inductive problems, extract and prioritize information from the environment, infer causal as well as correlative relationships from both small and large data sets over many known and novel domains, generalize knowledge from a known domain to another known or novel domain, extrapolate probable outcomes from both factual and counterfactual circumstances, recognize in its own cognition both the potential for fallacies and the fallacies themselves, synthesize existing knowledge to form original concepts, and acquire awareness of its own cognition and of itself as an independent and unique entity distinct from other entities and from its environment.
The supposition that if humans were able to develop an AGI system even slightly more intelligent than themselves, then that system would be able to do the same, and then the subsequent system would be able to do the same, etc. This would result in a superintelligent system far beyond the level of humans.
A postulated infinitely vast intelligence able to know the exact movements of every atom in the universe. This intelligence is at the heart of a thought experiment in which it would be possible to determine all future events in a deterministic universe.
Large Language Model (LLM)
A Large Language Model is a neural network built using a predictive deep learning model with billions of parameters. They are typically trained on extremely large data sets with self-supervised or semi-supervised learning, and they excel at a wide range of tasks.
Law of Accelerating Returns
The premise that analyzing the history of technology demonstrates that technological change is exponential, and that when any technological barrier is reached, a new paradigm will be found to surmount it.
Intelligence anywhere on the spectrum from AI to superintelligence that is exhibited by a machine rather than a biological organism.
A field of computer science that involves algorithms that automatically improve through experience. These algorithms frequently involve the use of artificial neural networks, although other techniques are also used.
A hypothetical machine that operates at the nanoscale and can manipulate individual atoms and molecules with atomic precision.
Originally coined to refer to technology that was created at the nanoscale (1 billionth of a meter) and operated on individual atoms and molecules. Later co-opted to refer to science that involves materials with at least one-dimensional attribute at the nanoscale.
The conjecture that intelligence and goals are not directly correlated, and consequently one can't assume that some level of intelligence would guarantee a particular subset of goals and exclude some other subset of goals. Conversely, one can't assume that a subset of goals would be guaranteed to be pursued or not pursued by any particular level of intelligence.
A thought experiment involving an AGI system that’s designed with the benign goal of maximizing the number of paperclips it creates. Inevitably, it tries to turn the entire universe into paperclips. The thought experiment is meant to demonstrate that the behavior of such a system is not predictable, and that even benign goals can result in a dire effect on humanity.
A model of human behavior originally used in economic theory to represent a consumer and the choices that consumer is most likely to make. Later used in various theories and speculation involving artificial general intelligence.
The method of discovery that involves careful observation and applying rigorous skepticism to what is observed given that cognitive assumptions can distort how one interprets the observation. It involves formulating via induction hypotheses based on such observations; experimental and measurement-based testing of deductions drawn from the hypotheses; and refinement (or elimination) of the hypotheses based on the experimental findings.
A level of intelligence radically beyond that of any human.
Cognition that is synthesized rather than occurring naturally, i.e. not biological in nature.
Synthetic Cognitive Entity
A strictly non-biological being that is capable of cognition and exhibits intelligence.
A moment in time when the rate of technological change advances so rapidly that we're unable to conceive of what might lie past it. At this point all our conceptions of humanity and society quickly fall away as the nature of technology and our relationship with it engender a new reality, a future discontinuous with our past. Many believe that the driving force behind this event is the accelerating evolution of intelligence itself, particularly the development of artificial general intelligence.
A situation that arises when a group of individuals talk past each other but are unaware that they’re doing so. It frequently occurs when discussing complex subjects. The individuals who don’t understand the subject offer essentially incoherent questions or statements, while the individuals who do understand the subject try to answer or respond without acknowledging the incoherence.
The algorithm guiding an intelligent agent's actions and designed to maximize the agent's ability to achieve its goals.
Value Alignment Problem
The issue of negative outcomes arising due to an AGI system's goals not aligning with human values.