Artificial Artificial IntelligenceIntelligence

bybyJeff PasternackJeff PasternackMike ThackerMike Thacker

55thth century BC century BCAristotle invents syllogistic logic, the first Aristotle invents syllogistic logic, the first formal deductive reasoning system.formal deductive reasoning system.

1616thth century AD century ADRabbi Loew supposedly invents the Rabbi Loew supposedly invents the Golem, an artificial man made out of clayGolem, an artificial man made out of clay

A Brief History of AIA Brief History of AI

1717thth century centuryDescartes proposes animals are Descartes proposes animals are machines and founds a scientific machines and founds a scientific paradigmparadigm that will dominate for 250 that will dominate for 250 years.years.Pascal creates the first mechanical Pascal creates the first mechanical calculator in 1642calculator in 1642

1818thth century centuryWolfgang von Kempelen “invents” fake Wolfgang von Kempelen “invents” fake chess-playing machine, The Turk.chess-playing machine, The Turk.

1919thth century centuryGeorge Boole creates a binary George Boole creates a binary algebra to represent “laws of algebra to represent “laws of thought”thought”

Charles Babbage and Lady Lovelace Charles Babbage and Lady Lovelace develop sophisticated programmable develop sophisticated programmable mechanical computers, precursor to mechanical computers, precursor to modern electronic computers.modern electronic computers.

2020thth century centuryKarel Kapek writes “Rossum’s Universal Karel Kapek writes “Rossum’s Universal Robots”, coining the English word Robots”, coining the English word “robot”“robot”

Warren McCulloch and Walter Pitts lay Warren McCulloch and Walter Pitts lay partial groundwork for neural networkspartial groundwork for neural networks

Turing writes “Computing Machinery and Turing writes “Computing Machinery and Intelligence” – proposal of Turing testIntelligence” – proposal of Turing test

1956: John McCarthy coins phrase 1956: John McCarthy coins phrase “artificial intelligence”“artificial intelligence”

1952-62: Arthur Samuel writes the first 1952-62: Arthur Samuel writes the first AI game program to challenge a world AI game program to challenge a world champion, in part due to learning.champion, in part due to learning.

1950’s-60’s: Masterman et. al at 1950’s-60’s: Masterman et. al at Cambridge create semantic nets that Cambridge create semantic nets that do machine translation.do machine translation.

1961: James Slagle writes first 1961: James Slagle writes first symbolic integrator, SAINT, to solve symbolic integrator, SAINT, to solve calculus problems.calculus problems.

1963: Thomas Evan’s writes ANALOGY, 1963: Thomas Evan’s writes ANALOGY, which solves analogy problems like which solves analogy problems like the ones on IQ tests.the ones on IQ tests.

1965: J. A. Robinson invents Resolution 1965: J. A. Robinson invents Resolution Method using formal logic as its Method using formal logic as its representation language.representation language.

1965: Joseph Weizenbaum creates 1965: Joseph Weizenbaum creates ELIZA, one of the earliest “chatterbots”ELIZA, one of the earliest “chatterbots”

1967: Feigenbaum et. al create Dendral, 1967: Feigenbaum et. al create Dendral, the first useful knowledge-based agent the first useful knowledge-based agent that interpreted mass spectrographs.that interpreted mass spectrographs.

1969: Shakey the robot combines 1969: Shakey the robot combines movement, perception and problem movement, perception and problem solving.solving.

1971: Terry Winograd demonstrates a 1971: Terry Winograd demonstrates a program that can understand English program that can understand English commands in the word of blocks.commands in the word of blocks.

1972: Alain Colmerauer writes Prolog1972: Alain Colmerauer writes Prolog

1974: Ted Shortliffe creates MYCIN, the 1974: Ted Shortliffe creates MYCIN, the first expert system which showed the first expert system which showed the effectiveness of rule-based knowledge effectiveness of rule-based knowledge representation for medical diagnosis.representation for medical diagnosis.

1978: Herb Simon wins Nobel Prize for 1978: Herb Simon wins Nobel Prize for theory of bounded rationalitytheory of bounded rationality

1983: James Allen creates Interval 1983: James Allen creates Interval Calculus as a formal representation Calculus as a formal representation for events in time.for events in time.

1980’s: Backpropagation (invented 1980’s: Backpropagation (invented 1974) rediscovered and sees wide 1974) rediscovered and sees wide use in neural networksuse in neural networks

1985: ALVINN, “an autonomous land 1985: ALVINN, “an autonomous land vehicle in a neural network” navigates vehicle in a neural network” navigates across the country (2800 miles).across the country (2800 miles).

Early 1990’s: Gerry Tesauro creates TD-Early 1990’s: Gerry Tesauro creates TD-Gammon, a learning backgammon agent Gammon, a learning backgammon agent that vies with championship playersthat vies with championship players

1997: Deep Blue defeats Garry Kasparov1997: Deep Blue defeats Garry Kasparov

Modern Times (post-Modern Times (post-Cartesian)Cartesian)

RobopetsRobopetsWidespread viruses, Widespread viruses,

security holes security holes aplentyaplenty

AI-powered CRMAI-powered CRMFaster—and many Faster—and many

more—computersmore—computers

A word about paradigms…A word about paradigms… AI will force a dualistic view of life to change because the AI will force a dualistic view of life to change because the

environment will be inseparable from it. Axiological shifts environment will be inseparable from it. Axiological shifts will occur in defining life, causing society to expand will occur in defining life, causing society to expand current definitions of life (e.g. requirement of a body). current definitions of life (e.g. requirement of a body). Also, the connectedness of the local environment to AI will Also, the connectedness of the local environment to AI will force science away from a reductionist view of this new force science away from a reductionist view of this new life and into a more complex view of interactions causing life and into a more complex view of interactions causing life to arise. life to arise.

On the other hand, most scientists would be happy to view On the other hand, most scientists would be happy to view the brain as a vast but complex machine. As such it should the brain as a vast but complex machine. As such it should then be possible to purely replicate the brain using then be possible to purely replicate the brain using artificial neurons. This has already been done for very artificial neurons. This has already been done for very simple life forms such as insects which only have a few simple life forms such as insects which only have a few thousand neurons in their brains. In principle, it would not thousand neurons in their brains. In principle, it would not be necessary to have a full scientific understanding of how be necessary to have a full scientific understanding of how the brain works. One would just build a copy of one using the brain works. One would just build a copy of one using artificial materials and see how it behaves. artificial materials and see how it behaves.

ETHICAL CONSIDERATIONSETHICAL CONSIDERATIONSUtilitarianism supports the development of AI, but only because of the Utilitarianism supports the development of AI, but only because of the

Christian value of dominion over the environment. AI promises to Christian value of dominion over the environment. AI promises to increase control over life, thus suffering can be reduced. Yet, if AI is increase control over life, thus suffering can be reduced. Yet, if AI is developed and not forced into particular tasks, Utilitarianism may developed and not forced into particular tasks, Utilitarianism may not apply. not apply.

Artificial life may be viewed as more expendable than human life, so Artificial life may be viewed as more expendable than human life, so AI will be used as cheap labor, or perhaps slaves, thus increasing AI will be used as cheap labor, or perhaps slaves, thus increasing profits for corporations.profits for corporations.

We do have to take responsibility for our creations, so if the risks We do have to take responsibility for our creations, so if the risks associated with creating a form of AI are too great, then we should associated with creating a form of AI are too great, then we should not pursue that development.not pursue that development.

We do have to take responsibility for our creations, so if the risks We do have to take responsibility for our creations, so if the risks associated with creating a form of AI are too great, then we should associated with creating a form of AI are too great, then we should not pursue that development.not pursue that development.

Rights Based EthicsRights Based Ethics Once AI programs achieve Once AI programs achieve

a modicum of sentience, a modicum of sentience, should they be given rights should they be given rights on par with other animals?on par with other animals?

Two sides of the argument: Two sides of the argument: 1) No, because sentience is 1) No, because sentience is

impossible to determine.impossible to determine. 2) Yes, because sentience 2) Yes, because sentience

can be proven beyond a can be proven beyond a reasonable doubt.reasonable doubt.

Duty ethicsDuty ethicsConsequence of developing AI is not at issue. What obligations do Consequence of developing AI is not at issue. What obligations do

we have to our biological children? Once AL is created, are they we have to our biological children? Once AL is created, are they to be a Frankenstein’s monster and cast off without help, or are to be a Frankenstein’s monster and cast off without help, or are they to be guided by their creator, like Adam & Eve in Eden.they to be guided by their creator, like Adam & Eve in Eden.

We do have to take responsibility for our creations, so if the risks We do have to take responsibility for our creations, so if the risks associated with creating a form of AI are too great, then we associated with creating a form of AI are too great, then we should not pursue that development.should not pursue that development.

Do we have a responsibility to our biological children similar to AI? Do we have a responsibility to our biological children similar to AI? Are the two exactly the same?Are the two exactly the same?

What rules should we use as categorical imperatives? AI should What rules should we use as categorical imperatives? AI should contain a set of rules that most people share, like “do not kill, contain a set of rules that most people share, like “do not kill, unless in self-defense” or “do not lie, unless the suffering caused unless in self-defense” or “do not lie, unless the suffering caused by honesty is large”. by honesty is large”.

ConfucianismConfucianism Confucianism is, in part, similar to Confucianism is, in part, similar to

Kant’s duty ethics: “don't do to Kant’s duty ethics: “don't do to others what you would not want others what you would not want yourself“ (reciprocity)yourself“ (reciprocity)

YìYì , not , not LìLì: actions should be based : actions should be based on righteousness, duty and morality, on righteousness, duty and morality, not on gain or profit.not on gain or profit.

RénRén: “benevolence, charity, : “benevolence, charity, humanity, love, and kindness”humanity, love, and kindness”

But AI is also an advance in civilization; it makes But AI is also an advance in civilization; it makes it possible for everyone to live more comfortable it possible for everyone to live more comfortable liveslives

Thus, Confucianism seems to encourage Thus, Confucianism seems to encourage development and application of (non-sentient) development and application of (non-sentient) AI, so long as it does not endanger othersAI, so long as it does not endanger others

If the AI is sentient, however, it would have to be If the AI is sentient, however, it would have to be treated humanely without exploitation, and treated humanely without exploitation, and could not be created out of individual or could not be created out of individual or corporate greed.corporate greed.

Though Confucianism can be seen as a religion, Though Confucianism can be seen as a religion, it is also a code of ethics, which suggests that it it is also a code of ethics, which suggests that it would be more open to considering AI to be alive would be more open to considering AI to be alive and deserving rights than other spiritual and deserving rights than other spiritual mythologies.mythologies.

Virtue ethicsVirtue ethicsDifficult to predict what virtues to Difficult to predict what virtues to

give Artificial life because it is a give Artificial life because it is a complex technology. Unable to complex technology. Unable to see the results of these virtues see the results of these virtues may be a problem because may be a problem because there are risks involved with there are risks involved with some virtues overpowering some virtues overpowering others.others.

What preprogrammed “virtues” What preprogrammed “virtues” should computers have to allow should computers have to allow them to be morally right? Can them to be morally right? Can virtues make an AI entity virtues make an AI entity behave morally at all?behave morally at all?

Wisdom, compassion, courage, Wisdom, compassion, courage, strength, obedience, strength, obedience, carefulness…?carefulness…?

ADVANTAGES (Factual ADVANTAGES (Factual Changes)Changes)

Smarter artificial intelligence promises to replace human jobs, Smarter artificial intelligence promises to replace human jobs, freeing people for other pursuits by automating manufacturing freeing people for other pursuits by automating manufacturing and transportations.and transportations.

Self-modifying, self-writing, and learning software relieves Self-modifying, self-writing, and learning software relieves programmers of the burdensome task of specifying the whole programmers of the burdensome task of specifying the whole of a program’s functionality—now we can just create the of a program’s functionality—now we can just create the framework and have the program itself fill in the rest framework and have the program itself fill in the rest (example: real-time strategy game artificial intelligence run by (example: real-time strategy game artificial intelligence run by a neural network that acts based on experience instead of an a neural network that acts based on experience instead of an explicit decision tree).explicit decision tree).

Self-replicating applications can make deployment easier and Self-replicating applications can make deployment easier and less resource-intensive.less resource-intensive.

AI can see relationships in enormous or diverse bodies of data AI can see relationships in enormous or diverse bodies of data that a human could notthat a human could not

Disadvantages (Risks)Disadvantages (Risks) Potential for Potential for

malevolent malevolent programs, “cold war” programs, “cold war” between two between two countries, countries, unforeseen impacts unforeseen impacts because it is because it is complex complex technologytechnology, , environmental environmental consequences will consequences will most likely be most likely be minimal.minimal.

Self-modifying, when combined with self-Self-modifying, when combined with self-replicating, can lead to dangerous, replicating, can lead to dangerous, unexpected results, such as a new and unexpected results, such as a new and frequently mutating computer virus.frequently mutating computer virus.

As computers get faster and more numerous, As computers get faster and more numerous, the possibility of randomly creating an the possibility of randomly creating an artificial intelligence becomes real.artificial intelligence becomes real.

Military robots may make it possible for a Military robots may make it possible for a country to indiscriminately attack less-country to indiscriminately attack less-advanced countries with few, if any, human advanced countries with few, if any, human casualties.casualties.

Rapid advances in AI could mean massive Rapid advances in AI could mean massive structural unemploymentstructural unemployment

AI utilizing non-transparent learning (i.e. AI utilizing non-transparent learning (i.e. neural networks) is never completely neural networks) is never completely predictablepredictable

Mythological ConsiderationsMythological Considerations

Do sentient programs have a soul?Do sentient programs have a soul? Christianity says no because a soul is Christianity says no because a soul is

imparted by God alone, and not by imparted by God alone, and not by Computer Scientists, yet Christianity does Computer Scientists, yet Christianity does dictate that we control the environment dictate that we control the environment around us, thus if cyberspace is part of our around us, thus if cyberspace is part of our environment, would AI allow us to control it? environment, would AI allow us to control it?

Buddhism and Taoism take a different Buddhism and Taoism take a different stance. stance.

Buddhism and TaoismBuddhism and Taoism

The Hua-Yen school of Buddhism offers as the The Hua-Yen school of Buddhism offers as the metaphor for the world an infinite net, at each metaphor for the world an infinite net, at each intersection of which lies a jewel in which exists intersection of which lies a jewel in which exists every other jewel and where every part of the net every other jewel and where every part of the net depends for its existence on dynamic awareness of depends for its existence on dynamic awareness of every other part. This is in line with the axiological every other part. This is in line with the axiological shift that would likely result from developing shift that would likely result from developing Artificial Life; environment and individual life are Artificial Life; environment and individual life are one. one.

Buddhism and Taoism value life above all else, so AI Buddhism and Taoism value life above all else, so AI would be valued just as highly as all other life, once would be valued just as highly as all other life, once developed. Creation of AI would be opposed because developed. Creation of AI would be opposed because AI does not share the tenets of the 8-fold path. AI does not share the tenets of the 8-fold path.

Christianity, Islam, JudaismChristianity, Islam, Judaism Christianity, Islam, and Judaism are (at least in Christianity, Islam, and Judaism are (at least in

relation to AI) very similar: they all state God relation to AI) very similar: they all state God created man in his own imagecreated man in his own image

So:So: How can man create artificial life if God is the creator of How can man create artificial life if God is the creator of

life? (“Say unto them, O Muhammad: Allah gives life to life? (“Say unto them, O Muhammad: Allah gives life to you, then causes you to die, then gathers you unto the you, then causes you to die, then gathers you unto the day of resurrection...”)day of resurrection...”)

If AI programs are sentient and as smart or smarter than If AI programs are sentient and as smart or smarter than humans, is man still the highest of worldly life?humans, is man still the highest of worldly life?

Does sentient AI have a soul? Does it ascend to heaven Does sentient AI have a soul? Does it ascend to heaven when it is deleted? Or when it stops running temporarily when it is deleted? Or when it stops running temporarily (and then is reborn)? How do you baptize software? (and then is reborn)? How do you baptize software? And so on…And so on…

Oh my!

Because they will not accept AI as life, Because they will not accept AI as life, Judaism, Christianity and Islam do not care Judaism, Christianity and Islam do not care about the rights and treatment of any AI.about the rights and treatment of any AI.

Instead, they will focus on the dangers to Instead, they will focus on the dangers to humans.humans.

Christianity concerned with orthodoxy Christianity concerned with orthodoxy (correct belief), while Islam and Judaism (correct belief), while Islam and Judaism concerned with orthopraxy (correct action)concerned with orthopraxy (correct action)

Should you release (potentially Should you release (potentially dangerous) AI software if you have tested dangerous) AI software if you have tested it to your satisfaction or if you have it to your satisfaction or if you have applied a set testing protocol?applied a set testing protocol?

Judaism and Christianity hold that man has free will; however, Judaism and Christianity hold that man has free will; however, Islam is more slanted toward predestination (“By no means Islam is more slanted toward predestination (“By no means can anything befall us but what God has destined for us”).can anything befall us but what God has destined for us”).

Digital AI doesn’t appear to have free will: all inputs and Digital AI doesn’t appear to have free will: all inputs and outputs to an AI program are discreet and reproducible, as outputs to an AI program are discreet and reproducible, as are the AI program’s state and execution (its “memory” and are the AI program’s state and execution (its “memory” and “thought”). Given the same conditions and the same input, “thought”). Given the same conditions and the same input, digital AI software will always produce the same output.digital AI software will always produce the same output.

Notice, however, that this could be true for humans as well, Notice, however, that this could be true for humans as well, but is unverifiable because our inputs, outputs, memories and but is unverifiable because our inputs, outputs, memories and thoughts are not easily accessible or reproducible.thoughts are not easily accessible or reproducible.

Bottom line: in most ways, sentient AI doesn’t make sense in Bottom line: in most ways, sentient AI doesn’t make sense in the context of these religions, and, in some cases, is contrary the context of these religions, and, in some cases, is contrary to their beliefs.to their beliefs.

And thus Christianity, Islam, and Judaism would not accept And thus Christianity, Islam, and Judaism would not accept any AI as sentient, and probably not even as life.any AI as sentient, and probably not even as life.

Applied Ethicist’s StanceApplied Ethicist’s Stance Macroethics-current societal values is Macroethics-current societal values is

dominated by Utilitarianism, thus AI is likely dominated by Utilitarianism, thus AI is likely to continue.to continue.

Microethics-depending on a person’s Microethics-depending on a person’s spirituality, this may influence the codes of spirituality, this may influence the codes of conduct designed into artificial life.conduct designed into artificial life.

Mesoethics-if a company develops AI, it will Mesoethics-if a company develops AI, it will produce a utilitarian creature. If research produce a utilitarian creature. If research institutes develop AI, then it may contain institutes develop AI, then it may contain various ethical standpoints depending on who various ethical standpoints depending on who is doing the research and development.is doing the research and development.

The Future?The Future? Idea of Artificial Idea of Artificial

Intelligence is being Intelligence is being replaced by Artificial life, replaced by Artificial life, or anything with a form or or anything with a form or body. body.

The consensus among The consensus among scientists is that a scientists is that a requirement for life is requirement for life is that it has an that it has an embodiment in some embodiment in some physical form, but this will physical form, but this will change. Programs may change. Programs may not fit this requirement not fit this requirement for life yet. for life yet.

Should we start caring yet?Should we start caring yet? Very sophisticated—perhaps even Very sophisticated—perhaps even

sentient—AI may not be far off; with sentient—AI may not be far off; with sufficient computation power (such as sufficient computation power (such as that offered by quantum computers) that offered by quantum computers) it is possible to “evolve” AI without it is possible to “evolve” AI without much programming effort.much programming effort.

Today, concerns include mutating Today, concerns include mutating viruses and the reliability of AI (you viruses and the reliability of AI (you don’t want software directing your don’t want software directing your car into a tree).car into a tree).

What should happenWhat should happen When programs that appear to demonstrate When programs that appear to demonstrate

sentience appear (intelligence and awareness), a sentience appear (intelligence and awareness), a panel of scientists could be assembled to determine panel of scientists could be assembled to determine if a particular program is sentient or not.if a particular program is sentient or not.

If sentient, it will be given rights, so, in general, If sentient, it will be given rights, so, in general, companies will try to avoid developing sentient AI companies will try to avoid developing sentient AI since they would not be able to indiscriminately since they would not be able to indiscriminately exploit it.exploit it.

Software companies should be made legally Software companies should be made legally responsible for failings of software that result in responsible for failings of software that result in damage to third parties despite good-faith attempts damage to third parties despite good-faith attempts at control by the user.at control by the user.

AI and robotics have the potentially to truly AI and robotics have the potentially to truly revolutionize the economy by replacing labor with revolutionize the economy by replacing labor with capital, allowing greater production—it deserves a capital, allowing greater production—it deserves a corresponding share of research funding!corresponding share of research funding!

And what is going to And what is going to happen…happen…

Most people are willing to torture and kill intelligent Most people are willing to torture and kill intelligent animals like cows just for a tastier lunch—why would animals like cows just for a tastier lunch—why would they hesitate to exploit artificial life?they hesitate to exploit artificial life?

This is further compounded mainstream religious This is further compounded mainstream religious beliefsbeliefs

Even with laws, any individual with sufficient Even with laws, any individual with sufficient computing power could “evolve” AI without much computing power could “evolve” AI without much programming.programming.

Licensing agreements will continue to allow careless Licensing agreements will continue to allow careless companies to often escape responsibility for faulty companies to often escape responsibility for faulty software.software.

Bottom line: ethical considerations will be ignored; Bottom line: ethical considerations will be ignored; reform—if it happens—will only take place when the reform—if it happens—will only take place when the economic costs become too high.economic costs become too high.