Lecture
AI takeover — is a hypothetical scenario in which artificial intelligence (AI) becomes the dominant form of intelligence on Earth, as computer programs or robots effectively take control of the planet away from humans. Possible scenarios include the replacement of the entire human workforce, a takeover by superintelligent artificial intelligence, and the popular notion of a robot uprising. Stories about AI takeovers are very popular in science fiction. Some public figures, such as Stephen Hawking and Elon Musk, have advocated for research into precautionary measures to ensure that future superintelligent machines remain under human control.
The traditional consensus among economists is that technological progress does not lead to long-term unemployment. However, recent innovations in robotics and artificial intelligence have raised concerns that human labor will become obsolete, leaving people in various sectors without work to earn a living, which would lead to an economic crisis. Many small and medium-sized businesses could also be driven out of business if they cannot afford or license the newest robotic and artificial intelligence technologies, and may need to focus on areas or services that cannot easily be replaced in order to remain viable in the face of such technologies.
Artificial intelligence technologies have become widespread in recent years. Although these technologies have displaced many traditional workers, they are also creating new opportunities. The industries most susceptible to AI takeover include transportation, retail, and the military industry. Military artificial intelligence technologies, for example, allow soldiers to operate remotely without any risk of injury. Author Dave Bond argues that as AI technologies continue to develop and expand, the relationship between humans and robots will change; they will become tightly integrated into several aspects of life. AI is most likely to displace some workers while creating opportunities for new jobs in other sectors, especially in areas where tasks are repetitive.
Computer-integrated manufacturing uses computers to control the production process. This allows individual processes to exchange information with each other and initiate actions. Although production can be faster and less prone to errors through the integration of computers, the main advantage is the ability to create automated production processes. Computer-integrated manufacturing is used in the automotive, aviation, aerospace, and shipbuilding industries.
In the 21st century, machines have partially taken over many skilled tasks, including translation, legal research, and low-level journalism. Care work, entertainment, and other tasks requiring empathy, which were previously considered safe from automation, have also begun to be performed by robots.
An autonomous vehicle — is a vehicle capable of perceiving its environment and moving without human involvement. Many such vehicles are under development, but as of May 2017, automated cars permitted on public roads are not yet fully autonomous. All of them require a human driver behind the wheel who can take control of the vehicle at any moment. Among the obstacles to widespread adoption of autonomous vehicles is concern about the loss of driving-related jobs in the road transport industry. On March 18, 2018, in Tempe, Arizona, a self-driving Uber car struck and killed the first person.
Scientists such as Stephen Hawking are confident that superhuman artificial intelligence is physically possible, stating that «there is no physical law precluding particles from being organized in ways that perform even more advanced computations than the arrangement of particles in the human brain». Scientists such as Nick Bostrom debate how far off superhuman intelligence is and whether it poses a danger to humanity. According to Bostrom, a superintelligent machine would not necessarily be driven by the same emotional motives — the desire to accumulate power that often drives people — but might instead view power as a means to achieving its ultimate goals; taking over the world would expand its access to resources and help prevent other agents from interfering with the machine's plans. As a simplified example, a paperclip maximizer designed solely to produce as many paperclips as possible would want to take over the world so that it could use all the world's resources to make as many paperclips as possible, and moreover, prevent people from shutting it down or using those resources for anything other than paperclips.
AI takeover — is a common theme in science fiction. Fictional scenarios tend to differ greatly from those envisioned by researchers, in that they involve active conflict between humans and AIs or robots with anthropomorphic motives who see humans as a threat or otherwise have an active desire to fight them, in contrast to the concern of AI researchers that an AI would rapidly exterminate humanity merely as a side effect of pursuing arbitrary goals. [17] The idea appears in Karel Čapek's R.U.R., which introduced the word « robot» in 1921, and can be seen in Mary Shelley's book .Frankenstein (published in 1818), as Victor considers that if he grants his monster's request and makes him a wife, they will reproduce, and their kind will destroy humanity.
The word «robot» from R.U.R. comes from the Czech word robota, meaning laborer or serf. The 1920 play was a protest against the rapid development of technology, featuring industrial «robots» with growing capabilities who ultimately rebelled. HAL 9000 (1968) and the original Terminator (1984) — are two iconic examples of hostile AI in pop culture.
Nick Bostrom and others have expressed concern that an AI with the abilities of a competent AI researcher would be able to modify its own source code and increase its own intelligence. If its self-reprogramming leads to it becoming even better at reprogramming itself, the result could be an intelligence explosion in which it quickly leaves human intelligence far behind. Bostrom defines superintelligence as «any intellect that greatly exceeds the cognitive performance of humans in virtually every domain of interest», and lists some advantages a superintelligence would have if it decided to compete with humans:
According to Bostrom, a computer program that precisely emulates the human brain, or runs algorithms as powerful as those of the human brain, could still become a «speed superintelligence» if it thinks many orders of magnitude faster than a human, due to being made of silicon rather than flesh, or due to optimizations that increase the AGI's speed. Biological neurons operate at a frequency of around 200 Hz, whereas a modern microprocessor operates at a frequency of around 2,000,000,000 Hz. Human axons carry action potentials at a speed of around 120 m/s, whereas computer signals travel at speeds close to the speed of light.
A network of human-level intellects designed to work together and share complex thoughts and memories without friction, able to collectively work as a giant unified team, or consisting of trillions of human-level intellects, would become a «collective superintelligence».
More broadly, any number of qualitative improvements to a human-level AGI could produce a «quality superintelligence», perhaps resulting in an AGI that is as far above us in intelligence as humans are above non-human apes. The number of neurons in the human brain is limited by cranial volume and metabolic constraints, whereas the number of processors in a supercomputer can be increased indefinitely. An AGI need not be limited by human working-memory constraints, and could therefore be capable of intuitively grasping more complex relationships than humans. An AGI with specialized cognitive support for engineering or computer programming would have an advantage in those fields over humans, who have not developed specialized mental modules specifically suited to working with those domains. Unlike humans, an AGI could create copies of itself and tinker with its copies' source code in an attempt to further improve its algorithms.
A serious problem is that unfriendly artificial intelligence is likely much easier to create than friendly AI. While both require significant advances in the design of a recursive optimization process, friendly AI also requires the ability to make goal structures invariant under self-improvement (otherwise the AI could transform into something unfriendly), and a goal structure that is aligned with human values and is not subject to instrumental convergence in a way that could cause it to automatically wipe out the entire human race. Unfriendly AI, on the other hand, can optimize for an arbitrary goal structure that need not be invariant under self-modification.
The sheer complexity of human value systems makes it very difficult to make AI motivation human-friendly. Unless moral philosophy provides us with a flawless ethical theory, an AI's utility function could permit many potentially harmful scenarios that comply with a given ethical framework but not with «common sense». According to Eliezer Yudkowsky, there is little reason to suppose that an artificially created mind would have such an adaptation.
Many scientists, including evolutionary psychologist Steven Pinker, argue that a superintelligent machine is likely to peacefully coexist with humans.
Fear of a cybernetic uprising is often based on interpretations of human history, replete with instances of enslavement and genocide. Such fears stem from the belief that rivalry and aggression are necessary in the goal system of any intelligent being. However, such human competitiveness stems from the evolutionary background of our intelligence, where the primary goal was survival and gene propagation in the face of human and non-human competitors. According to AI researcher Steve Omohundro, an arbitrary mind can have arbitrary goals: there is no particular reason why an artificial intelligence machine (which does not share humanity's evolutionary context) would be hostile or friendly, unless its creator programs it to be so and it is neither inclined nor able to modify that programming. But the question remains: what would happen if AI systems were able to interact and evolve (evolution in this context meaning self-modification or selection and reproduction) and had to compete for resources — would this create goals of self-preservation? An AI's goal of self-preservation could conflict with some human goals.
Many scientists dispute the likelihood of the unforeseen cybernetic uprising depicted in science fiction such as « The Matrix», arguing that it is more likely that any artificial intelligence powerful enough to threaten humanity would probably be programmed not to attack it. Pinker acknowledges the possibility of deliberate «bad actors», but states that in the absence of bad actors, unforeseen accidents do not pose a serious threat; Pinker argues that a culture of engineering safety will prevent AI researchers from accidentally unleashing a superintelligence. Yudkowsky, in contrast, argues that humanity is less threatened by deliberately aggressive AIs than by AIs that were programmed in such a way that their goals are unintentionally incompatible with human survival or well-being (as in the film «I, Robot» and in the story «The Evitable Conflict»). Omohundro suggests that current automated systems are not designed for safety and that AIs may blindly optimize narrow utility functions (say, playing chess at any cost), driving them to pursue self-preservation and the elimination of obstacles, including people who might turn them off.
The problem of AI control is how to build a superintelligent agent that will help its creators while avoiding the unintentional creation of a superintelligence that would harm its creators. Some scholars argue that solutions to the control problem could also find application in existing non-superintelligent AIs.
The main approaches to the control problem include alignment, which aims to bring AI goal systems into agreement with human values, and capability control, which aims to reduce an AI system's ability to harm people or gain control. An example of «capability control» is research into whether a superintelligent AI could be successfully confined in an «AI box». According to Bostrom, such capability-control proposals are unreliable or insufficient to solve the control problem in the long run, but could potentially serve as a valuable supplement to alignment efforts.
Physicist Stephen Hawking, Microsoft founder Bill Gates, and SpaceX founder Elon Musk have expressed concern about the possibility that AI could develop to the point where humans would not be able to control it, with Hawking suggesting that this could «spell the end of the human race». Stephen Hawking said in 2014 that «Success in creating AI would be the biggest event in human history. Unfortunately, it might also be the last, unless we learn how to avoid the risks». Hawking believed that in the coming decades AI could offer «incalculable benefits and risks», such as «technology outsmarting financial markets». In January 2015, Nick Bostrom joined Stephen Hawking, Max Tegmark, Elon Musk, Lord Martin Rees, Jaan Tallinn, and numerous AI researchers in signing an open letter from the Future of Life Institute, which speaks of the potential risks and benefits associated with artificial intelligence. The signatories «believe that research on how to make AI systems robust and beneficial is both important and timely, and that there are concrete research directions that can be pursued today».
Comments