Date of publication: 2017-09-04 05:32
The specific responses that such robots would have to specific stimuli or situations would differ from the responses that an evolved, selfish animal would have. For example, a well programmed helper robot would not hesitate to put itself in danger in order to help other robots or otherwise advance the goals of the AI it was serving. Perhaps the robot's "physical pain/fear" subroutines could be shut off in cases of altruism for the greater good, or else its decision processes could just override those selfish considerations when making choices requiring self-sacrifice.
Note: If you register online, and register during the late period or request any additional services, you must enter a credit card to pay those fees before submitting your registration.
ACT, Inc. operates within a framework of internal policies and procedures that protect the personal information of its customers. ACT does not sell or provide any personally identifiable information, including ACT test scores, to any test preparation companies. It is not ACT's practice to call students registering for the ACT test or other customers to sell them test prep services or to request credit card information over the phone.
Qualcomm is working with Southwest High School in North Carolina to improve student test scores using smartphones. Called Project K-Nect, Qualcomm has distributed smartphones in select courses, and teachers hope the devices will introduce high-tech applications to students who don’t have access to the internet at home. So far, the program has encouraged administrators after they determined their kids performed 75 percent better than classmates without smartphones on a final algebra exam.
Humans sometimes exhibit similar behavior, such as when a mother risks harm to save a child, or when monks burn themselves as a form of protest. And this kind of sacrifice is even more well known in eusocial insects, who are essentially robots produced to serve the colony's queen.
Also, organizing chimpanzees into an intelligence is hard because chimpanzees are difficult to stitch together in flexible ways. In contrast, software tools are easier to integrate within the interstices of a collective intelligence and thereby contribute to "whole is greater than the sum of parts" emergence of intelligence.
In 7556 I discovered Nick Bostrom and Eliezer Yudkowsky, and I began to follow the organization then called the Singularity Institute for Artificial Intelligence (SIAI), which is now MIRI. I took SIAI's ideas more seriously than Kurzweil's, but I remained embarrassed to mention the organization because the first word in SIAI's name sets off "insanity alarms" in listeners.
We are coming upon a humanity-altering era, and “we must now choose between the pursuit of unrestricted and undirected growth through science and technology and the clear accompanying dangers,” if we believe these dangers warrant enough concern to relinquish research in genetics, nanotechnology, and robotics (Joy). Our desire for better health, longevity, and technology may be too far along to stop. Therefore, relinquishment may not be an option.
I find these kinds of scenarios for AI takeover more plausible than a rapidly self-improving superintelligence. Indeed, even a human-level intelligence that can distribute copies of itself over the Internet might be able to take control of human infrastructure and hence take over the world. No "foom" is required.
#7: Once an AI passes a threshold, it might be able to absorb vastly more content (., by reading the Internet) that was previously inaccessible.
Bostrom suggests that AI belief systems might be modeled on those of humans, because otherwise we might judge an AI to be reasoning incorrectly. Such a view resembles my point in the previous paragraph, though it carries the risk that alternate epistemologies divorced from human understanding could work better.
Simulating trajectories of planets with extremely high fidelity seems hard. Unless there are computational shortcuts, it appears that one needs more matter and energy to simulate a given physical process to a high level of precision than what occurs in the physical process itself. For instance, to simulate a single protein folding currently requires supercomputers composed of huge numbers of atoms, and the rate of simulation is astronomically slower than the rate at which the protein folds in real life. Presumably superintelligence could vastly improve efficiency here, but it's not clear that protein folding could ever be simulated on a computer made of fewer atoms than are in the protein itself.
Given this, it would seem that a superintelligence's simulations would need to be coarser-grained than at the level of fundamental physical operations in order to be feasible. For instance, the simulation could model most of a planet at only a relatively high level of abstraction and then focus computational detail on those structures that would be more important, like the cells of extraterrestrial organisms if they emerge.
so far as I know, AlphaGo wasn’t built in collaboration with any of the commercial companies that built their own Go-playing programs for sale. The October architecture was simple and, so far as I know, incorporated very little in the way of all the particular tweaks that had built up the power of the best open-source Go programs of the time. Judging by the October architecture, after their big architectural insight, DeepMind mostly started over in the details (though they did reuse the widely known core insight of Monte Carlo Tree Search). DeepMind didn’t need to trade with any other Go companies or be part of an economy that traded polished cognitive modules, because DeepMind’s big insight let them leapfrog over all the detail work of their competitors.