Artificial general intelligence: savior, slayer, or not happening

At 7 pm we will order supper on individual bills. The talk won't begin till after supper, about 8 pm. A wireless mike and speakers will be available. People are welcome to gather earlier than 7 pm for conversation. --Murray E. Denofsky

In the 1960s many computer scientists predicted that in a couple decades computers would be able to do anything human minds could do. Led by the Department of Defense, specialized expert systems were developed that efficiently answered spoken questions in specific fields. Now research is refocusing on the original goal of artificial general intelligence that may develop into superintelligence.

Advances in technology, especially IT, are accelerating toward the singularity when machine intelligence can improve itself far better than any humans can. Superintelligent machines may then work toward restoring paradise earth or exterminating humans infesting it. Or civilization may well collapse long before the singularity.

The talk will present a summary of recent advances and their potential for good and evil. I’ll present an infographic on the subject and hope to learn more from members who know much that I don’t. Topics may include:

AI history and bottlenecks: Early optimism led to research splitting into various teams that communicated little with each other.

Possible applications: research, product development and marketing, language translation, speech recognition in any language and accent, tutoring, consumer product selection, cancer treatment by big pharma, warrior drones and robots to kill thousands, ideal partners for humans, reforming politics and ending war

Advanced technologies: The current shrinking of transistors is approaching its limit. Further progress depends on new technologies like carbon nanotube transistors, quantum computing, inexact computing, and cognitive computing that learns from interaction with humans and its environment

Intelligent personal assistants: perform tasks for an individual based on user input, location awareness, and access to information online such as weather, traffic, user schedules. Examples of such an agent are Apple (https://en.wikipedia.org/wiki/Apple_Inc.)'s Siri (https://en.wikipedia.org/wiki/Siri), Google (https://en.wikipedia.org/wiki/Google)'s Google Now (https://en.wikipedia.org/wiki/Google_Now), Amazon Echo (https://en.wikipedia.org/wiki/Amazon_Echo), Microsoft (https://en.wikipedia.org/wiki/Microsoft)'s Cortana (https://en.wikipedia.org/wiki/Cortana_(intelligent_personal_assistant)), Braina (https://en.wikipedia.org/wiki/Braina), Samsung (https://en.wikipedia.org/wiki/Samsung)'s S Voice (https://en.wikipedia.org/wiki/S_Voice), LG (https://en.wikipedia.org/wiki/LG)'s Voice Mate (https://en.wikipedia.org/wiki/Voice_Mate), SILVIA (https://en.wikipedia.org/wiki/SILVIA), HTC (https://en.wikipedia.org/wiki/HTC)'s Hidi, IBM (https://en.wikipedia.org/wiki/IBM)'s Watson (https://en.wikipedia.org/wiki/Watson_(computer)), and Facebook (https://en.wikipedia.org/wiki/Facebook)'s M.

Smart personal agents: perform ongoing tasks such as schedule management and restaurant reservation changes autonomously.

Cognitive computing such as IBM Watson: Understands natural language, extracts information from unstructured documents, and provides best rated answers. In 2011 Watson soundly beat the best human players at Jeopardy, and now has shrunk to the size of a desktop PC and its services are offered on free trial to consumers and businesses.

Dangers of Super AI: Depending on who first develops it, whether military, foreign powers, academia, or fastbuck corporations, results will vary. Presently 50 nations are developing AI warrior robots that may become stronger, faster and smarter than humans, able to swarm through a city, break down doors, and kill all who are judged dangerous, a Dick Cheney dream. Super AI may decide that humans are so warlike and mentally ill that to save AI and the planet, they must be exterminated by nuclear weapons, a power grid meltdown, or a virus pandemic.

Stages of AI: Superintelligence will grow in 4 phases:

1. Humans improve it until phase 2

2. Recursive Self-Improvement, better at designing itself than humans can, could do social manipulation and hacking, could be stopped if it causes trouble

3. Covert preparation, strategizing for long term goals

4. Overt implementation, can make heaven or hell on earth

Other: Transhumanism means humans and computers joined into cyborgs; the rights and responsibilities of human-smart robots and their programmers; downloading a human brain; Ray Kurzweil predicts desktop computers will equal human processing power in 2029, the singularity in 2045

by Robin Harper. I've:

• been with the group from the start

• been an early adopter of IT since 1981

• learned to write using many kinds of information structures, not just pictures & paragraphs as in most books and websites

• identified human instincts that evolved for survival & empire building, yet threaten our existence as a species