The Future and Ethics of AI Flashcards
What is the eventual goal of AI?
achieving capabilities equal to or greater than that of human intelligence
What are 2 classifications systems of AI?
weak and strong AI vs. narrow, general, and super AI
What kind of AI do we have currently regarding the weak vs strong AI classification system?
weak AI
What does weak AI do?
simulates some human capabilities
What are some examples of weak AI?
a robot that plays chess, chatbots, spotify shuffle, spam filters, smart assistants, autocorrect, self driving cars, navigation maps
What is an example of strong AI?
technology like in the movie Blade Runner
When was the concept of weak vs strong AI first used and by who?
John Searle in 1980
Is AlphaGo weak or strong AI?
weak
Is Watson weak or strong AI?
weak
What is the definition of strong AI?
AI that can perform various tasks at a human level
True or false: Strong AI would be able to have emotions and think at human levels of consciousness?
true
What are some characteristics of strong AI?
capable of general use in many fields, intelligence similar to human level, can learn independently, heave emotions, and self-awareness
What is narrow AI?
AI that specializes in one task
What is general AI?
human-level AI
What is super AI?
AI that is superior to human intelligence
What are examples of narrow AI?
DeepBlue, Watson, AlphaGo, facial recognition, autonomous cars, Siri
Do we currently have any general or super AI systems?
no
True or false: General AI systems can have thinking, social, and creative abilities?
true
What are characteristics of super AI?
have creativity, general wisdom, and social skills far superior to that of humans
What is a concern of using super AI?
it will try to dominate humans with it’s superior intelligence and abilities
What is a singularity in mathematical terms?
a point at which an equation becomes undefined
What is singularity regarding AI?
a hypothetical future point in time at which AI creates new civilization by surpassing human capabilities
What is the idea behind the Explosion of Intelligence?
the starting point where AI surpasses human intelligence and when super AI emerges
Who first proposed explosion of intelligence?
Irving Good
How might AI develop faster than human intelligence?
superintelligent machines make more superintelligent machines that are better than themselves
Who argued for singularity in technology and in what year?
Von Neumann in 1958
Who popularized the singularity?
Vinge
When is singularity predicted to be reached according to Ray Kurzweil?
2045
When will the explosion of intelligence occur according to Yudkowski?
in 30 years
Who criticizes predictions of singularity?
Pinker and Kaplan
What is transhumanism?
the belief that human physical and mental abilities can be improved by using scientific technologies such as AI
What does philosopher Nick Bostrom believe about super AI?
that humans may become extinct due to super AI
What is an idea to counteract the threat of AI turning on humans?
designing AI with a love for humans
What is the biggest obstacle we have to overcome in order to realize super AI?
overcoming human creativity
What are other concerns for AI regarding humans and safety?
black market sale of AI weapons, powerful governments using AI to develop weapons
What is a current example of an AI weapon?
an autonomous attack drone with facial recognition that can target specific people, genders, and races
As a result of the dangers of AI, what is being done to prevent these?
AI rules and standards to govern AI developers
What is an example of a chatbot that cause problems due to lack of ethics standards?
Tay, made racist and misogynistic tweets
What is an example of AI ethics standards?
the EU AI ethical principles announced in 2017
What are the general ideas of the EU AI ethics principles?
AI must develop stably, personal information must be protected, AI development and management must be transparent, and there must be strict liability regulations
What is another AI ethics standard?
Asilomar Principles of AI from a conference held in Asilomar in 2017
How many principles are in the Asilomar AI principles?
23
What are some ideas presented in the Asilomar AI principles?
AI must be useful and beneficial to humanity, AI -based arms races must be avoided, research funding should focus on beneficial AI use, research must be transparent and collaborative, privacy should be taken into account
