There is a new paper in PNAS, Building transformers from neurons and astrocytes, where the abstract stated that “Glial cells account for between 50% and 90% of all human brain cells, and serve a variety of important developmental, structural, and metabolic functions. Recent experimental efforts suggest that astrocytes, a type of glial cell, are also directly involved in core cognitive processes such as learning and memory. While it is well established that astrocytes and neurons are connected to one another in feedback loops across many timescales and spatial scales, there is a gap in understanding the computational role of neuron–astrocyte interactions.”
The computational role of neurons-astrocyte interactions in relation to transformers, the AI architecture responsible for LLMs may play a role in the future of generative applications. The success of LLMs drives regular assessments of their capabilities. There are debates about their intelligence since what they do is to predict the next word. The debates are not whether they access language but that language is not intelligence. Something similar is said of consciousness, where many say language is not consciousness.
Consciousness is described with experience for example, what it means to see red or smell lavender. So, if a word is seen, does that not mean consciousness too? If someone is told to imagine the scent of some air-freshener, and the person does, evoking a bit of what that scent is and pleasure, has language not driven consciousness? With its capability to do so, is it not a part of the apparatus?
When someone makes and sends a message, regular or digital, using the letter “I” does language not express the person’s existence? Isn’t it an extension of the awareness of the person’s being?
In biology, genetics can be said to be what “would” for an organism. While the mind can be said to be what could. What would is mostly expected with time. What could is mostly learned. All that could is derived from what would. Though genetics is at the core, what could or the nervous system, or memory makes determinations as well.
Knowing math for humans is what could. Other organisms don’t know math, exceeding what would and could. Language is how math is learned. It makes it possible for humans who understand it better to express it, then have other humans learn. Language carries the intelligence, contains intelligence and is intelligence.
There are debates on syntax and semantics, linguistic, contextual and nonlinguistic forms. There are deductions in mind, within fractions of seconds that don’t seem to involve language, but it is an option and in cases where it is used or the output, it bears much.
Language is commonly understood as human languages. It is differentiated from communication. However, language could have a broader meaning than some of the rigid cases. When a bird, in a group, takes a sharp flight for some reason and others follow, if moving is communicated, can the action not be loosely described as a language, a type of bird language?
Moving could signal that the one closer to something had seen a threat, so when that others notice, they get that as part of a medium for their communication [or language] and follow. If that move prevents danger, can it not be said to be a constituent of their consciousness and intelligence?
Language, for humans, is a property of the available stretch of the mind. It is not everything but plays a major role, to receive intelligence, to give it as well as for the consciousness of being and experience.
Conceptually, the human mind has two key components, electrical and chemical impulses. Sets of electrical impulses interact with sets of chemical impulses, to produce knowing [of emotions, feelings, sensations, perceptions, regulations, thoughts, memories and others], to degrees. The sets have their features.
Humans are more intelligent than other organisms because of a feature [theoretically] of sets of chemical impulses called stairs or drifts, for rationing or fills. Chemical impulses [neurotransmitters, neuropeptides, neurohormones and others] are rationed when struck by a set of electrical impulses. This rationing to different proportions defines why taste is different from sight, why touch has degrees, why emotions are different from feelings and why memory encoding is different from retrieval.
Since all that is in the mind are electrical and chemical impulses, what becomes experienced is structured by rationing, which electrical impulses induce and takeoff from. Synaptic clefts are part of the rationing destinations.
Drifts or stairs for electrical impulses, that may influence rationing or fills for chemical impulses, are myelinated axons, where some electrical impulses concurrently jump from node to node, over myelin sheaths in saltatory conduction.
There are different brain circuits with more specializations for certain functions than at others. The customization of their localized rationing shapes what they do. Language is available at some of these rations, whose expansions begin early. Communication can be said to be more distributed than language, though signing is a form of language even in non-standard cases.
Humans can think out loud or say things with little to no pause, showing a connection between thought-memory and language. Whatever is thought about is something known. All that is known is in the memory. Thought can be said to be the transport within the memory space. There are possibilities within memory, including intelligence and consciousness. The possibility of language increases how human intelligence can be expanded, as well the extent of human consciousness.
When a bull is on a field, looking back and forth, its consciousness of experience is also using a form of thoughts to make inferences, on risks communicated to it in various forms. It does not have language, but that is not what makes it vulnerable, it is that it does not have enough rationing to understand better, in a limit of its mind. When the bull is in transport, it may observe that it is being moved but be unable to understand how the medium came about.
The limits of other organisms are not necessarily because they can’t speak, type or write, but they lack rations for understanding, complex problem-solving and so forth, which humans have. It may also be more difficult to tame other organisms if they were more intelligent or had more rationing, even without language.
Language makes intelligence easier for the human species. It would have been possible to make strides without it, but would have been more difficult.
AI has language, containing human intelligence. There is no difference between some of the accurate outputs of LLMs and what some experts in some fields would answer in some situations.
Some people are looking for AI with all the other properties of human intelligence that include understanding to categorize AI, but those may not be necessary. Language is used to communicate those and since generative AI has access, it must not have those to be included.
Consciousness can be defined as the rate of knowing for any species, with humans having the highest at 1. Language is included in what humans can know, factoring in human consciousness. Other forms of communications of other organisms or their rudimentary languages also add to their totals. For AI, the language of humans it has, can be estimated for its consciousness.
—
This Post is republished on Medium.
—
Photo credit: iStock
