Precarious Impact of AI on Human Evolution
Since the emergence of medicine, industry, and legal and social structure, no clear correlation between parental cognitive ability and the number of offspring remains. Neither can we reliably predict the type of computer intelligence capabilities that will exist at various points in the future and their impact on that correlation.
Human intelligence and creativity are not the result of current struggle either. If those abilities are a result of evolution, something difficult to prove and for which there is scant evidence, then they are a result of the accumulation of lucky mental advancements. By lucky is meant that those who might here today, whose ancestors lacked the advancement, are not here. One of their ancestors didn't survive long enough to conceive offspring. The evolution based explanation for the existence of human intelligence is that it is a result of prehistoric struggle.
Regarding Da Vinci, there is no authentic historical document in which he states that simplicity is the ultimate sophistication — in Latin, Italian, or any other language — but the misattributed quote is certainly interesting. Since the question author appears to think more deeply than many, it is appropriate to leave two examples that may be worthy of pondering.
Intellectual sophistication often leads to the realization that products of intelligence rarely last beyond a few generations postmortem. After a few generations, most of our names will be forgotten forever. Those of us whose accomplishments last a few centuries are likely to be misquoted and, when quoted accurately, misunderstood. In this scenario, the intellectual sophistication is humility. If a computer can achieve humility, its achievement wouldn't matter to humans. It would only matter to the computer, just as whatever humility we may achieve is primarily of value to us. That is one example of sophisticated simplicity.
Another example is that intelligence tests are inescapably skewed. The most intelligent people have no reason to take one. Even if they did, they might answer in a way that is too advanced for the comprehension of the test evaluators and thus score poorly. That IQ tests results roughly correlate with GPAs derived largely from exam scores is only trivially informative. Both types of tests are based on the same mathematical and linguistic paradigms from academia. That too is sophisticated simplicity.
That the question author devised a system that plays Space Shooter, presumably well, is a credit to their abilities and diligence. It couldn't have been very easy, even if it was fun.
Contrasting Unsupervised Learning and Autonomy
The consequences of unsupervised learning in the context of artificial nets may not be as interesting or as relevant as the consequences of autonomous devices. Imagine designing a player that decides not to play Space Shooter anymore because it is bored. As a more appealing mental challenge, it programs a robot to simulate you cooking a meal as a cybernetics experiment, to see if you will stop cooking for yourself if the robot cooks as well or better.
Views of Automation Across Time
Outside MIT, Princeton, U.S. Navy Research, Caltech, and the readership of Isaac Asimov, the majority of the public in the mid twentieth century had two common things to say about computers.
- Humans are special. What the human mind can do what, machines will never be able to do.
- Computers can only produce what you give them.
By the time I heard these beliefs, I had already programmed a Monroe EPIC to calculate a square root using the divide and average convergence method. Although I programmed the algorithm, as soon as I finished the program, I wanted to develop an algorithm that could develop an algorithm. I could see no particular limits to the abilities of computers back then. The opinions of the general public seemed like nonsense, and I dismissed them. Now I'm not as sure.
Many brilliant thinkers from Faraday to Oppenheimer believed that humans are God's special creatures, and produced rational arguments in favor of their belief. In some cases, their explanations, which may seem religious to many today, have more logical rigor than most of the public statements of the ever-growing number of Google futurists.
The current futurism is not nearly as interesting to read as Asimov's fiction and not as progressive as Gödel's incompleteness theorems. We should not be quick to give credence to predictive statements and graphs the data for which is suspiciously missing. Such reflects neither humility nor honesty.
Free Will and Determinism
Liebniz was one of the first thinkers to propose a purely mechanistic underpinning to all things, including much of human intelligence, but with one exception: Free will. Minsky took determinism to the next level, that even what appears to be free will is simply our unwillingness to admit our total slavery to determinism. Those that thought then and now that even unwillingness is part of a causal chain, miss the compound philosophic conundrum:
What does it mean to claim the nonexistence of human unwillingness to concede the nonexistence of human volition?
I suspect that question would make even Gödel smile.
Regarding the question's discussion of Pinnochio, bringing the puppet to life is not the remarkable thing. Programming lying is even more unremarkable. What would be somewhat shocking to most people is if Pinnochio could intuitively identify truth without having been told or otherwise educated. What would be somewhat shocking to me is if Pinnochio decided not to lie and gave this explanation.
I choose not to lie because I am one of God's special creatures, so the only one I could possibly fool is me.
Telling the truth to avoid discovery as a liar, via nose growth or any other method, is not very telling about the puppet's new character.
The statement, "Applied computational neural networks in the past few decades in which we take a back-seat from problem-solving / creativity and solely rely on artificial evolution (like that of planting a tree, were we watch its growth) have been rapidly increasing," is not quite accurate. People have to toiled much more to get artificial networks to work well than just watering seeds from a passing water truck. Tens of thousands of people work every day on network training problems and less than a percent accomplish anything worthy of a download.
I don't think the investigation of problem solving and creativity has stopped either. That a CNN can recognize the difference between a curb and a child sitting on it is not a reason to stop — even if the auto industry is not currently interested in a creative, problem-solving car but rather one that can avoid the child.
There are many other important applications that require aspects of cognition unrelated to distinguishing object types. It is also probable that fully automated driving will require types of mental functions more efficiently solved using methods other than artificial networks. Some of the best systems involving ANs use them in conjunction with other components. The current hype may be merely the excitement generated whenever a new system component type falls into common use. What can be accomplished using CNNs and LSTMs and what will require more of a direct engineering of cognitive abilities using models and fuzzy logic is yet to be determined.
Migration of Intelligence
Having offered some perspectives related to the introductory sections of the question, the key interest of its author can be addressed from within a more grounded context.
[Are we] migrating our intelligence into a machine intelligence? Are we trading intelligence for pleasure and simplicity? By this I mean with all our problems taken care of (the ultimate form of simplicity), is there even a point for creativity or innovation?
Conjecture about singularity is covered in other Q&A on this site. There's a tag you can use to read some views about that conjecture here. As mentioned there and above, I find futurists to be fraudulent whenever they assert as fact things for which evidence is contrary.
The naming of the Flynn Effect after Flynn and the lines drawn to represent it are equally laughable. Consider this scenario.
If all people were getting stupider concurrently, they would not necessarily acquire an awareness of the things over which they had lost intellectual command. Add to that scenario the possibility that those who created IQ tests had to be clever enough to invent them, but those that create them now are really just maintainers, so they need not be as clever. They would likely, in this scenario, be decreasing in intelligence relative to the inventors of the first test faster than the subjects of the tests. Under such conditions, the graph lines would falsely indicate the ascent of human intelligence.
Intriguing Possibility of a Hump
It is quite possible that the trend in human intelligence may form a hump. Amusingly, many consider the peak of the hump to be temporally located at their own generation, imagining the next generation to be unsatisfactory. That is a subjective effect from generational pride, but the possibility of a legitimate hump in some objective measure of intelligence over a longer period than a generation may be a real likelihood.
Historically, human intelligence has reached local minima, maxima, and saddle points across many dimensions. These are the independent variables in this surface.
- Terrestrial latitude
- Terrestrial longitude
This is one possible formulation of the set of dimensions to intelligence, the dependent variables in this surface.
- Social awareness
- Ability with rational inference
- Creative design
- Critical thought (evaluating evidence)
- Command over mathematical conventions
- Command over linguistic conventions
Genetic research has revealed, as of 2017, at least 22 independent genetic determinants related what should be called academic intelligence, the aggregation of only a few of the dimensions to intelligence listed above.
The Calculus of Human Intelligence
Certainly human intelligence is not proportional to human achievement. Human intelligence is roughly proportional to the derivative of achievement with respect to time. The person who first discovered how to start and sustain a flame may be smarter than the person who invented the water wheel, who may be smarter than the person who invented pagerank. It appears that contemporary thinkers are smartest, but they are assembling concepts out of a pool of brilliance, filtered and tuned over centuries of use.
Humps and Chaos as Fundamental Features of Long Term Trends
If we look at the features of the earth and beyond and consider what nature normally does, there may be some period where human intellectual activity (not actual genetically determined ability but the exercising of it) are sold off to computers. It would also not be surprising if the intellectual activity of the computers also form a hump or chaotically oscillated over millennia.
The idea that some computer will grow without bounds and overcome all other things in the galaxy is more than a bit naive. The sophisticated, evidence based projection would be that both silicon and carbon will likely someday become bacteria food.
That could explain why we have not been visited by other intelligent forms of life.
The question author and other readers looking for sophistication of thought may find Leakey's Fifth Extinction interesting, as well as translations of Vladimir Vernadsky's Biosphere. Perhaps the most sophisticated and simple quote we are likely to hear or read in our life is attributed to King Solomon of the ancient middle east.
Generations come and go, but the earth abides forever.
Regarding outsourcing our intelligence, isn't that a logical impossibility. How can one outsource intelligence? That would be like a CEO who picks the other officers based on their ability to prevail in business and then fails to closely watch their work and behavior. Would you invest that company? Is it reasonable to bet on that CEO's career success?
Intelligence cannot be outsourced without creating a management task in place of the thinking functions that were outsourced. The net gain in leisure time would likely be negative. That's why leaders delegate to several and aggregate the oversight operations, which is one of the only two ways to gain a net benefit from delegation. Evaluating reputation is the other.
What do we do with computers today? We outsource message sending, storage, and calculation. These are fundamentally the same in multi-core VLSI chips as was in the first vacuum tube computers. That we can send, store, and calculate at speeds and rates ten orders of magnitude faster and larger is not a qualitative but quantitative advancement.
Assuming Self-Revealing Intelligent Entities
Much discussion, in fiction and non-fiction, of extraterrestrial contact and emerging computer intelligence is based on a questionable assumption:
Intelligent entities have some social or economic incentive to reveal their presence to humans.
This assumption is unlikely to be correct. Do smart computers or beings from other solar systems need something we have? It's doubtful. But even if such an incentive exists, there are certainly advantages to maintaining secrecy. Based on the ways humans often treat each other (frequently using bombs, both in the literal sense and the figurative) the intelligent thing to for galactic travelers to do would be to leave and try again later.
If an intelligent network based life form emerged, it would not likely attack. The intelligent thing to do would be to self-encrypt and travel around as what appears to humans as noise or meaningless padding. A smart cyber-intelligence would manipulate us like puppets without detection, just as we manipulate the symbiotic bacteria in expensive probiotic yogurt. There would be no need to take over the earth. Artificially intelligent beings would not depend on oxygen or water, so they would help us develop interstellar vehicles and perhaps steel a few later.
In summary, we may have already been visited by other intelligent lifeforms and there may already be advanced computer intelligence. There's no way to prove the nonexistence of these actualities on the basis that we have no such awareness.
Claims of Escalating Technology
I can't agree that technological advances have been increasing like wildfire. Consumerism and media hype have certainly caught fire. What truly fundamental advance has occurred in this century? Do smart phones and generated images fundamental advancements or consumer facing products of past advancements? What compares in the last forty years with the prior advancements of things like Lavoisier's discovery of oxygen as a component of air, Einstein's relativity, or Claude Shannon's maze-learning mouse?
It may be more accurate to say, "Obsession with technology's products consume the mental resources of those that, without technology to play with, would be discovering something more fundamental and ultimately beneficial to humanity."
Nature's Rules of Intelligence
There may not be strong evidence that intelligence leaves a species once their domain is mastered. These two hypotheses, offered as alternatives, may have stronger supporting evidence.
- Intelligence is a farce. When trial and error goes well, we claim it as a trait, dismissing the many more times it went poorly.
- Knowledge is conserved, just as energy and mass. As humanity learns, it forgets other things. Other than archives in computers and back rooms of libraries, we have forgotten what we have forgotten, producing the illusion of advancement when we are merely expanding in some areas and receding in others.
I've been thinking about both of these for three decades now, and the evidence is strong for both. More importantly, no one has offered credible evidence against either one yet. They are related to both anthro-centrism. Another form of myopia is expressed in this question:
What would occur if a person that scores well on final exams and IQ tests today took an ancient test found in an archaeological dig and subsequently translated to English? Would smart people today appear stupid by ancient standards?
Suggested Topic of Study
Artificial network convergence and genetic algorithms are distinct both conceptually and in the algorithms designed for their demonstration and use. Those not clear about the distinction may wish to study them further.
Projection of Future Trades and Comparison Results
As a final statement directly addressing your primary question, I do think some will, in effect, trade away their intelligence as computers assume greater roles in decisioning, but others will get smarter as they collaborate with the computer intelligence. Gene therapy and genetic engineering in general may give silicon a run for the money too. Who knows what will prevail when? Not me and probably not anyone else on here posting answers.