http://plato.stanford.edu/entries/consciousness-animal/#hist http://people.whitman.edu/~herbrawt/classes/339/Descartes.pdf http://home.cogeco.ca/~drheault/ee_readings/West/Descartes.pdf http://journalofcosmology.com/Consciousness136.html
https://arxiv.org/abs/1611.04135
if you torture the data long enough, it will confess
animals are without feeling or awareness of any kind?
But the greatest of all the prejudices we have retained from infancy is that of believing that brutes think
the soul of the brute is of the same nature as our own
Rene Descartes, “Animals are machines”
The celebrated polymath Rene Descartes is known for – among other things – his dualistic philosophy which held that body and mind are ontologically distinct entities. The mind (or “soul”) is the incorporeal seat of consciousness, rationality, and emotion. Besides for being separate from body, minds are unique to human beings, according to Descartes, and therefore animals (“brutes”), being purely of mechanical composition are “machines” which lack the ability to perceive pain. Thus, he advocated live dissection for science and other practices regarded centuries later as cruel, and believed harming animals was of no moral consequence (unless it encroached on another human being). The protesting cry of an animal is no different from a “[whining gear that needs oil.](http://www.temple.edu/tempress/titles/1359_reg.html”
Of course, few people needed a philosophical justification to practice bull-baiting, ___, and other such activities. But an animal rights movement fomented only after the emergence of the view that animals were not categorically distinct from human beings. After all, animals seek food and warmth, recoil from pain, quaver in fear, compete for resources and practice self-preservation. Our ability to empathize with animals comes from this tendency to exhibit faculties we see in ourselves.
As contemporary machines continue to acquire more of the traits we once associated only with “organic” beings, Hollywood movies like Her and Ex Machina have played on this tendency.
But in Descartes time, the notion was widespread. No animal rights groups existed to protest practices like vivisection, bull-baiting, and other abuses.
This chapter will address the various and numerous sociocultural implications surrounding artificial intelligence and machine learning, including ethical and political considerations. It will be mostly a big survey of these, and may be broken out into smaller chapters in the future.
Increasing pervasiveness of machine intelligence
about speculative problems, singularity hype, movies like Terminator, Her, Ex Machina, Matrix…
…but AI systems are widespread now and already, problems which most of the ill-informed movies ignore. There are already clear and present dangers, and issues we are increasingly confronting.
…rethinking their strategies, which means many complex decisions will be sorted out in the near future.
http://us8.campaign-archive1.com/?u=bdb368b9a389b010c19dbcd54&id=f2e0882b79 http://socialmediacollective.org/reading-lists/critical-algorithm-studies/ https://www.reddit.com/r/science/comments/45k2pv/science_ama_series_we_study_how_intelligent/
google ceo: ”Machine learning is a core, transformative way by which we’re rethinking everything we’re doing,” he said. http://www.pcworld.com/article/2996620/business/google-reports-strong-profit-says-its-rethinking-everything-around-machine-learning.html
credit card fraud,
In personal communications, we are all familiar with spam filters. But as Google and other companies begin rolling out products like Inbox, we are seeing machine intelligence take up e-mail and communications organization more generally, by offering to sort and group your e-mail according to subject analysis. This is already done, not without controversy, in the retail and commercial sectors where products and inventories are organized by algorithms. With interpersonal communications, these questions persist and hit closer to home.
Questions like:
- When these algorithms are permitted to scan our e-mails, what else do they use the analysis for? Do they store the results, do they share them, do they sell them?
- These algorithms affect which of our e-mails we are more likely to see than other ones. How do they make such discriminations? Knowing that setting them up in different ways may sort them differently, how do we decide one method vs another.
Like in the trolley problem, there are no obvious solutions to these, partly because they are subjective questions. So how are these decisions arrived at? We’ve seen that various settings and free parameters, and certainly entire architectures have enormous influence on the behavior of these algorithms. Which of these levers are exposed to the user, and which are conveniently automated? How much agency does the user have with respect to that question?
There is a conflict of interest inherent in the last problem. Other entities have a stake in these decisions, in addition to the user. Are some topics favored against others?
Various implications of
[ Inbox:: I’ll give you Tuesday/Thursday ]
The trolley problem and the issue of agency
There is an old __ . A more concrete example of this is the Google driverless car – forced to make a decision who to kill, how does it decide? These are questions with no obvious answers, and even fewer explicit legal frameworks exist to answer them. many have made much commentary about it.
Algorithmic accountability and bias
columbia journo disseration on algo bias
Access to data, research, and tools
Asymmetric access in public / private sphere to resources.
Surveillance and profiling
Terror tuesdays, skynet
Genetic surveillance
HDH has explored this since her work with stranger visions, and subsequent ones.
Predicting crime from faces
https://arxiv.org/abs/1611.04135 l’homme criminel Propublica recrimination article
Diversity in research and adjacent fields
The research sector in AI and machine learning suffers from what plagues many fields more generally, particularly the overall field of computer science.
WIML The situation with respect to racial and socioeconomic differences is even more dramatic. Fewer than 3%(?) are POC Even less racial and socioeconomic
Rights of man, rights of machine?
Descartes re: dogs, dogs do not have feelings, you may hurt them
thriving <-> abuse [robot walking along, robot getting messed with, robot getting pushed over]
If you think these questions are overly speculative, long shots, or far off, I invite you to watch to view the robot from the Boston Dynamic group Robot. try to watch this without having feelings
As machines demonstrate more autonomy, can they act out of self-preservation? Nature shows us that self-preservation is a helpful instinct to have and thus selected for it. Will humans, in order to make machines more effective, select for self-preservational qualities? If they begin to clamor for more support, how will we react?
Meta-ethics
These questions are relevant even to the design of this book! Certain technical features are encapsulated and hidden for the sake of accessibility. Does it encourage reductionist thinking?
tbd
these are the sorts of issues movies like Her and Ex Machina play to, but there are numerous issues to contend ith
if you’re skeptical that we’ll have a rights movement for robots, remember that descartes said beasts (animals) had no conciousness/were automatons, ergo you can harm them http://nymag.com/thecut/2016/03/sophia-robot-hanson-robotics.html
AMI on ethics https://www.reddit.com/r/science/comments/45k2pv/science_ama_series_we_study_how_intelligent/
- self driving cars – humans outlawed to drive in future?
- what about vs planes
- scifi writers as social theorists
- e.g. asimov caves of steel - jobs + robots
https://en.wikipedia.org/wiki/Trolley_problem
andrew ng “Worrying about killer robots is like worrying about overpopulation on Mars – we’ll have plenty of time to figure it out.” http://www.rollingstone.com/culture/features/inside-the-artificial-intelligence-revolution-a-special-report-pt-1-20160229#ixzz41aAUSmUe “standing on railoraod tracks worried about being hit by lightning”
http://www.nature.com/news/machine-ethics-the-robot-s-dilemma-1.17881 http://www.nature.com/news/robotics-ethics-of-artificial-intelligence-1.17611
facebook stuff from barca workshop JFK, noted hero on the left
http://www.nextplatform.com/2016/09/14/next-wave-deep-learning-applications/ https://www.technologyreview.com/s/602317/self-driving-cars-can-learn-a-lot-by-playing-grand-theft-auto
ml fairness http://blog.mrtz.org/ how ml decision making is unfair: https://medium.com/@mrtz/how-big-data-is-unfair-9aa544d739de#.mwmwj32du
know your terrorist credit score https://re-publica.de/16/session/know-your-terrorist-credit-score
https://vimeo.com/163292139
AI safety https://openai.com/blog/concrete-ai-safety-problems/ AI safety (google brain) https://arxiv.org/pdf/1606.06565.pdf
WIML gender diversity http://stanford.edu/~sorathan/papers/SAILORS-SIGCSE2016.pdf
chatbots argue https://www.youtube.com/watch?v=WnzlbyTZsQY
Building Machines That Learn and Think Like People https://arxiv.org/pdf/1604.00289v1.pdf
The Hidden Dangers of AI for Queer and Trans People https://modelviewculture.com/pieces/the-hidden-dangers-of-ai-for-queer-and-trans-people
enrollment up 600% http://blogs.nvidia.com/blog/2016/02/24/enrollment-in-machine-learning/
fairness in ML delip rao https://docs.google.com/presentation/d/11vNi2CxPMV_wFhfhpXBK8qe1o4KN59hNVaPeD4zagNA/edit#slide=id.g11be5e1541_1_68
Chances are your Models are Racist, Sexist, or both deliprao.com/archives/129
fighting cancer www.futuretimeline.net/blog/2016/04/16-2.htm#.WGKCFrYrJE7
questions of interest to investigative journalists
- how does content recommendation in social media influence public opinion?
- what effect do autonomous financial micro transactions have on global economics?
- what is the future of “predictive policing” in law enforcement?
- how do we deal with job loss as sectors of the workforce are automated (e.g. professional drivers, factory technicians)
- how does bias in training data affect the quality of ML systems? how do we regulate them?
- what are the consequences (good and bad) of openness in AI research and development?
emphasis on techniques for activists, journalists, citizen scientists
machine learning has many implications on security, privacy, encryption. neural nets encode information so they can be used for e.g., generating encryption keys, compressing data, etc. so question is how reliable are they? can original info be recovered/decoded? can they be fooled/tampered?
machine learning can de-anonymize (e.g. infer identities from writing style) and de-ambiguate (infer unstated political affiliations, sexual orientation, etc from social media). the same exact algorithms can be used by activists to expose wrongdoing by crawling publicly-released (or leaked) masses of data which are impractical to scan through by journalists.
machine learning runs micro and macroeconomics. financial policy, investment banking, credit rating, etc are all dominated by machine learning, and transactions are increasingly made without human oversight, usually in milliseconds.
pre-crime type stuff (like in minority report) is real. trolley problem type ethics are more relevant to this. so if you cover silicon valley, you may ask how driverless car companies deal with these questions, how datasets are collected and labeled, what contingency plans are when they fail or implicate people falsely.
automated fact-checking
- report https://fullfact.org/media/uploads/full_fact-the_state_of_automated_factchecking_aug_2016.pdf
- paper http://journals.plos.org/plosone/article/asset?id=10.1371/journal.pone.0128193.PDF
self-driving cars: https://www.technologyreview.com/s/602273/fully-autonomous-cars-are-unlikely-says-americas-top-transportation-safety-official/
social bot interference in global politics http://firstmonday.org/ojs/index.php/fm/article/view/6161/5300
Kate Crawford :: Social Media, Financial Algorithms and the Hack Crash = http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2602857
pound plunges 6% in 2 minutes (rogue automated algo) https://www.ft.com/content/dfb375be-8c23-11e6-8cb7-e7ada1d123b1 2010 flash crash https://en.wikipedia.org/wiki/2010_Flash_Crash
Over the last few years, machine learning technologies have trickled out of research labs and into our daily lives. They are widely used in internet and social media applications like content filtering and recommendation, investment and finance, scientific research, and increasingly in unexpected places, such as law enforcement via so-called predictive policing. Many jobs done by humans are increasingly being automated as well; as these machines continue to claim more responsibilities from us, their influence on our lives will continue to grow indefinitely, pushing us to investigate and educate ourselves about how they work, so we can make informed decisions about how to integrate them into society.
deblurring faces https://arxiv.org/pdf/1702.00783.pdf https://arstechnica.com/information-technology/2017/02/google-brain-super-resolution-zoom-enhance/ http://gigazine.net/news/20170208-pixel-recursive-super-resolution/
Using Deep Learning and Google Street View to Estimate the Demographic Makeup of the US https://arxiv.org/abs/1702.06683
https://arxiv.org/abs/1602.02697
https://twitter.com/hen_drik/status/841643896899850240
https://intelligence.org/files/PredictingAI.pdf http://www.openphilanthropy.org/focus/global-catastrophic-risks/potential-risks-advanced-artificial-intelligence/ai-timelines
https://iamtrask.github.io/2017/03/17/safe-ai/ https://iamtrask.github.io/2017/06/05/homomorphic-surveillance/
easily fooled: http://karpathy.github.io/2015/03/30/breaking-convnets/
how many jobs lost: http://www.marketwatch.com/amp/story/guid/A6E401F0-463F-11E7-980C-8D853B065349
https://medium.com/@yonatanzunger/asking-the-right-questions-about-ai-7ed2d9820c48