Modern Technologies and Conflicts
JAKUB DRMOLA
AI, autonomy and society

weak AI
-machine learning
-deep learning, ANNs, atd…
-fastest growing domain

-Deep Blue (chess, 1997)
-Watson (jeopardy, 2011)
-AlphaGo (go, 2016)
-
-using computing power to overcome human capabilities, not a true general AI


Image result for deep blue garry kasparov Image result for deepmind lee sedol loss

Imitating mind
-natural effort to replicate consciousness
-while we don’t really what it is
-
-often “measured” by ability to communicate
-Turing test versus Chinese room argument
-ELIZA (1966) -> Mitsuku (2017)
-
-sensory perception
-https://quickdraw.withgoogle.com/
-

-true AI still beyond our reach

Related image Image result for hu,man face robot Image result for sophia robot Image result for
ishiguro robot


Motion
-Shakey (1972) – first robot
-
-bipedal balance still difficult
-delicate tough also difficult
-
-Boston Dynamics prototypes:
-Handle
-Spot, 2
-Atlas, 2
-
-
◦

Current use
-monitoring and recording
-news, police work and rescue, research
-transportation
-Amazon, pizza, beer, medicine…
-entertainment
-racing













http://moralmachine.mit.edu/



Image result for car autonomy levels






Future use
 transformation of economy
◦https://www.pwc.co.uk/economic-services/ukeo/pwcukeo-section-4-automation-march-2017-v2.pdf
◦https://www.oxfordmartin.ox.ac.uk/downloads/academic/The_Future_of_Employment.pdf
 home care
◦Robear
◦other types of „care“
 creativity
◦Sunspring, painting, music
◦
◦
◦




Potential problems
 - machine learning outcomes are highly dependent on the quality of source materials being use for
the process
 - machine learning can lead to “black boxes”, when even the creators do not fully understand the
AI and its decisions
 - this will probably get even worse with more advanced AIs
Image result for tay bot Image result for tay bot