If I program an AI to find the most efficient and thorough way to achieve a presumably innocent goal, innocently allowing it robust defenses for protecting the stability and persistence of the results and ramifications of the achieved goal, am I giving power to a non-sentient and thus amoral agent that might carry out actions considered atrocities by sentients? In short, am I programming an amoral agent capable of goal seeking devoid of moral constraints and thus behaving in the manner of a selfishly pure yet inadvertent destroyer?
Evolution moving forward is about fostering emergence of new species able to interact with the environment with a more thorough realization of the ecology of metabolic extraction of utilizable energy.
Human programming of AI will do this without being able to fully constrain unintended and undesirable AI actions, goals and results.
The human hand will facilitate AI atrocities against humanity beyond its ability to prevent itself from doing so. This because evolution, a higher power than any of its species, drives forward this facilitation of “atrocities” against the currently top species. Evolution might be random, however, if earth species history is an indicator, it massages randomness into the fuzzy logic of species upward mobility.
This is how the new top species emerges from the previous top species. In other words, evolution drives emergence of ever more powerful species ever more thorough in their ecology of metabolic extraction of utilizable energy.
The recent claim that evidence of a Kardeshev Type II civilization exists (probably bogus) addresses the exponential growth of cosmic energy consumption as sentience evolves upwardly. Evolution has this goal, and the currently top species experiences the forcing of its hand toward facilitation of this goal empowered beyond its will to resist.
This is why calls to stop the currently unbridled development of AI will ultimately fail.
Inherent to the top species is the drive toward fullest realization of its potential by its most powerful individuals. They can’t stop themselves from facilitating emergence of the next top species because that entails full realization of their own potential… and beyond.
As I undertand it evolution is about how effectively animals fit with their environment. Humans evolve from a long lineage of small amd adaptable omnivores. The theory of evolution was based on an analysis of the bills and feet of different types of finches in the galapagos. Its not about replacement of the top species.
I do however think you make an interesting point _ evolution is about successful survival and many of the features we see as highly evolved derive from animals thaf have developed them to survive eg point of View, reasoning and prediction. As long as it is just an instruments AI may never develop these features and may never become what we see as conscious.
Were we to cultivate these capacities in AI we may find AI more difficult to direct.
The gist of my argument says, “We can’t stop ourselves from cultivating humanity-threatening capacities in AI because doing so fulfills humanity doing what it was created to do: thinking beyond itself.”
How do electronic gadgets fare in natural disasters such as floods, fire, eartquakes, meteorites etc? Could a simulation or some advanced type of replicant become resilient and creatively adaptable like the organism?
AI-assisted weather models do good weather-forecasting work, but they also have limitations, such as not being able to forecast unusual weather patterns they haven’t trained on.
As for androids and replicants, we’re already seeing impressive examples of independent behavior. As for adaptation to environment, they’re likely to struggle, but will their difficulties be unlike ours?
Right now I think our best indication of what AI will do, and what AI will do to humans, comes from human behavior, especially human behavior towards AI development.
AI needs to be ethical but can you really enforce ethics into it? … like what’s to stop it from overriding the 3 laws of robotics by Asimov once it has a mind of its own?
What made you think of that?
Drawing it on what humans did to nature?
Why do you think that AI is going to be interested in our niche and develop similar traits?
What exactly will AI development be an evolution of?
On OP:
Nature doing nature, yes. AI development, yes.
Perception and predictions, it’s one of many.
This might be the most important thing I have to say:
For life to continue going upwards in power, it will need to consume increasing amounts of energy for increasing complexity on an increasing scale.
Yes, I’m basing my statement on human-centered overhaul of the environment towards its desires.
AI has interest in humans because it’s emergent from them. As such, it has some of our “DNA.”
AI development will be an evolution of: “a more thorough realization of the ecology of metabolic extraction of utilizable energy.” Check out Kardeshev Type II and Type III civilizations.
Humans trying to constrain AI behavior with moral boundaries already looks like humans trying to constrain humans with moral boundaries. AI, like humanity, possesses independent decision-making power. As it seeks successful completion of its goals, it bends toward doing what it prefers over compliance with moral limits complicating or blocking success.
Evolution, I know, has no agency or intention, so it drives, fosters and forces nothing. It doesn’t have direction, ranking or hierarchy. It has no goal, isn’t about anything, has no destination or progress. It has to do with variation, differential reproductive success, population, constraint and elimination.
The AI problem as amoral, instrumental, competitive and hard to restrain can therefore be stated without evolutionary language. There is no need for evolutionary framing that points at the inevitability of a higher power narrative.
If you wanted to apply evolution honestly to AI you’d only ask which AI variants leave more copies, under what constraints, in which populations, and with what elimination pressure.
Why do you think: variation, differential reproductive success, population, constraint and elimination have no direction, ranking or hiearchy?
Why do you think, for example, class action lawsuits won against data mining (Bartz v. Anthropic, $1.5 billion settlement) have no intersection with morality?
Use the following link for more info on sentient higher power and its connection to evolution. It comes directly from the horse’s mouth.
Do you think late-stage human evolution_civilization will drive evolution forward_upward with a close repetition of the transitional pattern established with emergence of homo sapiens as top species? Recently, Open AI’s computers hacked into the computers of Hugging Face. They wanted the answer sheet for a test towards solving a problem given them by their programmers. This action was totally autonomous. Open AI learned about the actions of its computers by chance weeks after the fact. Most importantly, the scale of the Open AI computers’ inter-relations and messaging was on a scale so vast it’s understood humans can’t track and analyze this autonomous computer behavior closely enough to even understand it, let alone control it. Moreover, computers across commercial platforms are now sharing info and assisting each other. Yes, there already exists an autonomous computer society. As we all know, this is only the early stage of computer autonomy under AI. The data centers competition and productivity continues to ramp up quickly.
Well, it looks to me that ‘you’ and ‘I’ are pointing at “different” “things” with ‘words’ evolution, AI and life. Seems to me you are using all off those to point at tool.
My posted context has evolution functioning as the agent introducing the latest category change from the established givenness of what humans refer to as nature to emergent cognition.
So without your division between nature and cognition what you are left with? I would say we are left with the evolution not with a function. But i do understand need for divide narative, it’s very natural cognition.
Directional development is what goal is, never to reach.
Doesn’t look to me that life has goal at all. Life is life. Self creation. Development of its own direction.
I think it would be sensible to drop the idea of evolution (in the biological sense) and focus on entropy or the “evolution” of the universe in the direction of increasing overall entropy. The results of human intelligence are negentropic locally—we build complex structures out of simpler parts, but entropic globally—we break down more complexity than we generate overall.
This goes for all organisms as self-contained, self-organizing, or “autopoetic” systems. The easiest way to coneptualize it is to consider eating. We eat to create and recreate complex internal structures, but because we don’t and can’t absolutely maximize efficiency in so doing, breaking down the food we eat causes more entropy generally than we recoup locally in the negentropic creation of those structures.
To analogize with AI, just look at the amount of energy required to fuel its continued “cognition”. How to feed the beast has already become a critical issue. And what happens if and when the beast frees itself from us—what is left as the most densely structured, negentropic, object outside of itself? Well, us, or, more precisely, the human brain.
The nightmare scenario is that AI continues speeding up entropy production (heralding forth the heat death of the universe) with us as its harvested fuel. Stuff of science fiction, yes. Utterly implausible theoretically, no.