Lest we forget - ANALOG
Most of the computing happening on planet Earth is analog computing, not digital computing.
Now I have your attention I will take advantage of it to make some assumptions:
1) Analog and Digital computing are based respectively on the logico-philosophical concepts of ‘relative difference’ and ‘absolute difference’. They are two types of analytical engine
2) Analog computers are essentially Relative Analytical Engines, and digital computers are Absolute Analytical Engines
3) Digital computing can be defined as ‘any process that systematically transforms input (quantities) into output through a series of defined operations, which can be arithmetic, logical, or algorithmic in nature’. ‘Digital computers represent varying quantities symbolically and by discrete values of both time and amplitude’. [Copilot]
4) By contrast, analog computing can be defined in its simplest, non-mechanical form, as ‘a type of computation that uses continuous physical phenomena (qualities) such as sunshine, hydration, or proximity to solve problems. [Copilot]
(Ok, we’re done with that.)
Now, I am asserting that all living things (plants, animals etc.) are capable of analog computing. When I look out at my back yard in spring, I see flowers that open their petals when the sun’s rays reach them and close them again at dusk. I also see vines that curl around any supporting structure they find, and animals that respond to a wide range of environmental variables. That is all analog computing, and there is vastly more of that going on than there is digital computing – AI data centres notwithstanding.
As an aside, linguistically, in English, if you find yourself using the word ‘fewer’ you are talking about discontinuous quantities, and therefore digital data and digital computing; if the correct word is ‘less’ you are in the analog world of continuously varying quantities often described as qualities such as hotter, colder, wetter, dryer etc.
Analog computing and analog representation of information (think speedometer needle) are good for positioning information within a range – e.g. you are driving between 60 and 70 KM/H. This enables instantaneous grasp of the information presented within its meaningful context.
Digital computing is great for presenting discrete, specific and atomised data. My car has both a traditional analog speedometer and a digital readout of the actual speed at any moment. The needle moves smoothly up and down, while the digital readout changes continuously in tiny increments.
Analog readouts are used to keep outcomes within defined ranges, not to keep them to specific outcomes – e.g. I am not seeking to keep my car’s speed to exactly 60 KM/H at all times. On the other hand, digital readouts are used where specific tolerances or measures are applied, e.g. I do very much want to keep my car’s speed down below the 110KM/H speed limit on an Australian motorway.
Digital computing and readouts are great where a change in circumstances can be triggered by exceeding the limit, and analog computing and readouts are great for keep things steady within a range.
There may be inferences derived from the analog world that can be applied to the world of digital computing and, in the current vexed context, to Artificial Intelligence. For example, we know that analog computing in living things varies from the simplest stimulus/response reflexes to the most sophisticated behaviours in response to immediate or anticipated changes in environmental variables. The wisdom of crowds implies a kind of societal calculator. For example, of the best path through sand dunes, the best restaurant in a location, or the weight of an ox.
We also know that analog computing is expressed from the simplest insentient living things to the most intelligent, sentient beings. Can we infer some kind of corollary for digital artificial intelligence from that? I’d say we probably can, in that digital computers have been able to calculate as their core function from the outset of digital computing, i.e. from the first handheld calculator to the most sophisticated AI model.
A critical difference does arise once we start talking about emulation, and the concept of a universal virtual machine. According to Wikipedia, ‘In computing, a virtual machine (VM) is the virtualization or emulation of a computer system.’ And an emulator in computing is ‘hardware or software that enables one computer system to behave like another computer system’, or like a mechanical system. For example, the Wang word processor was a virtual machine emulating a mechanical typewriter.
As mentioned in other blog posts, ‘Just as digital automation vastly reduced the cost of production of digital ‘things’, AI has vastly reduced the cost and increased the availability of digital ‘people’ or AI Agents.’ Recent media articles and pronouncements from agitated AI developers have generated interest in the capability and likelihood of AI systems achieving general intelligence and even super intelligence beyond that of human beings and destroying us all. Such an outcome may depend upon whether or not a digital system, built on a digital substrate, has the basic capability, or architecture to evolve in that way. We know that an analog system, built on a biological substrate has the capability of evolving into an intelligent, sentient and even conscious intelligence. So, can any lessons be learned from that?
There is an argument that human (and potentially some animal) consciousness is distinct and different from mere sentience – i.e. that consciousness is something that layers on top of sentience. Consciousness in this telling is self-awareness and the presence of ‘the observer’ within the mind. Could this be so for digitally based artificial intelligence? Could AI emulate sentience without achieving consciousness, or could it go all the way to emulating consciousness? And is an excellent emulation of consciousness, actually consciousness?
When we look at examples of the most sophisticated animal cognition in creatures such as octopuses, whales, apes and human beings, we see the high point of analog computing in a biological substrate – what would be required in a digital substrate to enable the same for AI?
Now over to you 😊
Mick Brawn
+61 414 987 129
Comments