Workflow Intelligence
From John Locke to Workflow Intelligence
On Lockean Learning: Liberalism
On Hegelian Abduction: Hegelian Abduction
On Disorganization: Disconnected Workflow
Technology generally makes workflow processing faster, and AI technology continues to increase its facility. However, the meaning of the word intelligence in relation to AI faces the usual problem: it is quickly out-bounded by the word super-intelligence, which lacks an immediate referent. The referent for super-intelligence is, indeed, to the future of intelligence. However, the advent of super-intelligence is also the advent of meaningless.
How so?
Technology, in terms of its ability to process data intelligently and thereby mimic intelligence as we know it, proves itself smarter in terms of the speed of its processing. Technological speed in itself does not necessarily render intelligence unintelligible as it still preserves intelligence at a level of coherence we are able to relate to, but we do need to be situationally aware as this very speed poses an existential threat. ...
On Understanding:
But the out-bounded threat, the undefined, the super-intelligence, is not necessarily related to speed so much as it is related to technology developing an intelligence we cannot relate to.
How so?
For now, not withstanding our efforts to see processes such as crystal formation as intelligent, intelligence is generally defined as activity generated by carbon-based life, and any intelligence outside of a basis in carbon-based life generally falls to artificial intelligence or to theology or to ... .
No less, super-intelligence serves to render our intelligence meaningless in that our intelligence is our ability to understand relationships and an intelligence that understands differently is likely to create relationships we may or may not be able to understand.
And we do not need artificial intelligence to acquire consciousness within this definition to render our intelligence meaningless.
Mostly the opposite!
As consciousness, especially shared consciousness, serves to relate intelligence to super-intelligence, the underlying, but unsubstantiated claim is that consciousness develops in AI as its facility of AI in reasoning advances.
If only it were reasoning! But even if AI were actually reasoning, there is no guarantee of AI acquiring consciousness, perhaps even little likelihood.
And here is a set of discussions as to why AI as computing consciousness is not likely:
Roger Penrose on Consciousness ...
But forgoing the argument Penrose gives for intelligence (Roger Penrose on AI (YouTube)), do we need consciousness to have an understanding or to have intelligence of a kind?
As a mechanism that makes relationships ready-to-hand, understanding works as that which builds the process: an understanding that relates this to that within a process need not be conscious: understanding as a mechanism needs to work as a built-in in a ready-to-hand fashion.
See Also:
- On AI: Gary Marcus
- On The Illusion of Thinking: Apple
On Monkeys Typing Shakespeare:
In terms of the agency in play, the Monkeys Typing Shakespeare problem is near this level of understanding: paper, typewriters, which have a built-in representation of language and probabilistically, the work of Shakespeare, and monkeys, who do the typing without understanding the language in play and whose consciousness is not really related to the solution.
On Mis-Understanding and No Understanding:
Ask a question as though it addresses a position held by the assumed respondent and forgo the stated case most entirely. What is with such, as Harold Bloom might term it, misprision? But forgoing further vegetable words from the bully pulpit, here is a bit of lukewarm stew: What are we without consciousness? This answer seems central.
What is AI without consciousness? It is, at least, what it is now. And consciousness is not central to AI, but consciousness is central to certain types of arguments regarding AI.
What is AI for us should it out-bound our ability to relate to it?
What I have done to the consciousness argument for AI is to show that it is a tack-on proposition, which serves to establish, however tentatively, a means of relation between our intelligence and its lorded super-intelligence.
But my central question remains: should AI go beyond our intelligence, not as a mimic of it, but as something creative, as something that is a maker of workflows, however lacking in understanding it might be, how are we going to relate to it?
On Meaning:
William Thurston wrote an essay on proof in mathematics, and here, the topic of proof is quite analogous to the topic of superintelligence, i.e., whether superintelligence delivers that which has meaning for us or not, as our ability to understand what a proof is largely constitutes its meaning for us: William Thurston on the Use of Computing in Problem Solving.
Read This Too
- On AI-topia: Luisa Jarovksy
- On Interpretability: Latent Space
On Workflow:
One way around the question of developmental intelligence is to shift to a discussion of workflows, which usually assess intelligence as embedded in a process. The process itself is the intelligence, and understanding operates largely a posteriori.
A Philosophy of Computing Controls
On Decision Theory
On AI Workfow:
On Capacity and Process and Governance
Capacity signifies the ability to process input where the process itself may be or may not be known. Process refers to how the input is treated. Intelligence is how the process does or does not fit the input.
Workflows are themselves processes, and they are also composed of processes. So, the intelligence we find in the world involves transformation (i.e., processing input), and we make such transformations artificially by finding means to transform the same kind of inputs and/or related inputs. We tend to model such tasks as workflows.
What I am saying here--'true, but knowing at that level may or may not be a goal. The goal of automating workflows is not [necessarily] to create understanding nor is it [necessarily] to create intelligence ... . It is merely a means to run a process. In this light, "artificial intelligence" is just the intelligence built in to a process'--is that I usually only need to know how the input is treated to use a capacity in a workflow.
The risk that I do not understand everything about a process is mitigated by how well I understand what happens to all of the inputs.
Read This Too:
Computing Power and the Governance of Artificial Intelligence:
The History of Computer Science:
MIT:
The Imposition of Natural Selection on the Progress of AI:
State of the Workflow:
Comments
Post a Comment