com exam 4
Finagle's law
"Anything that can go wrong, will."
are ppl polite to computers?
-designers want ppl to like machines and politeness is one way to ensure it
Why the future doesn't need us
Bill Joy
A new option to contact a choice of crisis counsellor helplines via
Facebook's Messenger tool, however, is limited to the US for now.
robots, engineered organisms, and nanobots share a dangerous amplifying factor:
They can self-replicate.
— Yann LeCun, Director of AI Research
"We have incredible people in FAIR who are making significant progress in AI, but to really move the bar it's equally as important to be outward focused. To push the envelope, push the science and technology forward, we must be actively engaged with the research community. We publish a lot of things we do, distribute a lot of code on open-source, and engage deeply with academia to drive the progress."
our main job in the 21st century will be
"ensuring continued cooperation from the robot industries" by passing laws decreeing that they be "nice,"3 and to describe how seriously dangerous a human can be "once transformed into an unbounded superintelligent robot."
For years now, companies such as Amazon, Google and Facebook have personalised the information we are fed
- combing through our "metadata" to choose items they think we are most likely to be interested in
The tool is being tested only in the US at present.
- marks the first use of AI technology to review messages on the network since founder Mark Zuckerberg announced last month that he also hoped to use algorithms to identify posts by terrorists, among other concerning content. -announced new ways to tackle suicidal behaviour on its Facebook Live broadcast tool and has partnered with several US mental health organisations to let vulnerable users contact them via its Messenger platform.
DL: targeted advertising
- uses deep neural networks - the foundation stones of deep learning - to decide which adverts to show to which users. -google also fighting for supremacy in this market
Biological species almost never survive encounters with superior competitors.
-10mill yrs ago w north and south america -in a completely free marketplace(but we dont live in one), robots would affect humans the same
tools will be used for more complex cases
-AI-driven warfare, prioritising patients in hospital and determining which areas of a city should be most heavily policed, questioning them is essential.
looking into the future
-Deep Learning is likely to continue to play a key part in the future development of Facebook. -ideas such as automatically generating audio descriptions of pictures to assist the visually-impaired, and to predict where greater coverage is required in its mission to roll out internet access to poorly served parts of the world. -deeplearning is likely to benefit many other organizations
tiny elite
-Due to improved techniques the elite will have greater control over the masses; and because human work will no longer be necessary the masses will be superfluous, a useless burden on the system. -If the elite is ruthless they may simply decide to exterminate the mass of humanity
The use of algorithms to control the way that we're treated extends well beyond Google's personalised search or Facebook's customised news feed
-Tech giants such as Cisco have explored the way in which the internet could be divided into groups of customers who would receive preferential download speeds based on their perceived value
It is a mistake to always decry this kind of personalisation as a negative.
-The futurist and writer Arthur C Clarke once noted that any sufficiently advanced technology is indistinguishable from magic.
facebook AI research (FAIR)
-We are committed to advancing the field of machine intelligence and are creating new technologies to give people better ways to communicate. In short, to solve AI. -seek to understand and develop systems with human level intelligence by advancing the longer-term academic problems surrounding AI.
post experiment
-after ps were told about results they all said they would never be polite to a computer -shows social responses were unconscious and automatic
experiment w voices instead of text on computer
-all taught by one voice -at the end ps were either asked by same voice for an evaluation or a different voice -ps responded more positively and false to same voice -ps responded more honest to dif voice -results: users are more polite to computers whether its text or voice
We may view computers as coolly objective
-but human bias and error can enter into algorithms, too. It's when we believe that such systems are beyond question
politeness across cultures
-caution readers about cross-cultural differences in politeness -translation of language is often the only consideration for internalizing media products, mistakes are frequent -cultural differences are a bit overrated mostly bc they focus attention away from all human beings: everyone is polite
Sherry Turkle makes an interesting observation about PCs in the 1980s
-computers moved from being a hobbyist - machine you could open up and physically tinker with to a machine you could learn to operate immediately, even if you didn't know exactly how it was working. Computers, Turkle playfully noted, begged to be taken not at face value but at interface value.
clarity
-contributions to an interaction shouldn't be obscure -highly technical language is often necessary -better to have a statement w 3 meanings than to have 1 that is precise but unknowable -user testing can determine the most common way that ambiguity is understood
DL: facial recognition
-deepface: to teach it to recognize people in photos. -human facial recognition for deep face: 97% while humans is 96% -controversial -EU legislators agreed and persuaded Facebook to remove the functionality from European citizens' accounts in 2013. -fb has been quiet about deep face tech but lawsuits will pro prevail soon
pattern recognition
-developed pattern-recognition algorithms to recognise if someone is struggling, by training them with examples of the posts that have previously been flagged. -Once a post has been identified, it is sent for rapid review to the network's community operations team. -
quantity
-each speaker should contribute only to what the conversation demands not too much or too little -interactive media violate this -single icons and words can be frustrating bc they dont tell us everything -solution: use ppl's ability to elaborate abbreviated messages w info that they already have/familiarity/technical abbreviations
deep learning
-enables machines to learn to classify data by themselves. -By analyzing a large number of images, it can learn from the context of the image - what else is likely to be present in an image of a cat? What text or metadata might suggest that an image contains a cat? -give structure to unstructured data, by quantifying it and representing it in a form from which analytical tools can derive insights.
observing polite interactions w computers
-experiments of ps working w computer instead of a person -black NEXT computer w 21-inch black/white monitor -only displayed text and graphical buttons -computer spit out facts and ps rated their knowledge on the topic and then took a test about the facts -at the end computer rated its performance as a great job -ps had the rate the computers by adjectives (accurate, fair, helpful. etc)
open source
-fb is strong supporter -Facebook Artificial Intelligence Research (FAIR) freely available for anyone to use or modify however they like. -Most of Facebook's Deep Learning is built on the Torch platform, a development environment focused on the development of deep learning technologies and neural networks.
DL: Designing AI applications
-flow: implemented which uses Deep Learning analysis to run simulations of 300,000 machine learning models every month, to allow engineers to test ideas and pinpoint opportunities for efficiency.
AI and algorithms dominate our lives -
-from the way financial markets carry out trades to the discovery of new pharmaceutical drugs and the means by which we discover and consume our news.
experiment with paper and pencil interview ("other")
-had ps answer questions on computer like before and also paper and pencil -paper/pencil ps were less favorable on evaluations of computer -did not require polite response
problem for face book's scientists
-have to try to make sense of this is that much of this data is very messily unstructured. -1.2 billion people uploading 136,000 photos and updating their status 293,000 times per minute,
voice-over experiment results
-having a voice doesn't make the interaction any more social than text
it is possible that human control over the machines may be retained.
-man is in charge of private machines like a car -the tiny elite is in charge of larger systems of machines
its impolite to reject
-media increasingly provides the ability to change how they look and work during an interaction -when you change something, you reject one option in favor of another -when you change features that are more obviously social, the reaction is more social therefore impolite -ppl feel the same inhibitions when rejecting social representations on a screen as they would w a real person
rules of etiquette
-media should be judged by their social and technical sophistication -its polite to say hello and goodbye even an interface character -its polite to look at ppl when speaking -its polite to match modality: answer a letter w a letter -politeness and product testing: eliminating positive bias -be suspicious of verbal responses, dos important reactions are not verbalized
factors of experiment that we must control
-only thing that differed in each experiment was what medium asked for an evaluation -made sure ps did not know what was actually being studied -ps did not think that computers have human capabilities (ps were all avid computer users) -evaluations were not out of impulse -ppl willing suspend disbelief when they encounter media
no one in the experiment said that they were using social rules for any reason
-person was not subconsciously thinking about programmer -if they used two computers during the study, they assumed it was same programmer
designing polite media
-polite behavior and many social behaviors are a part of an interaction -social means reciprocal -biggest reason to make machines that are polite to humans is so that humans are polite to the machines
experiment results
-ps who answered questions on the same computer gave significantly more positive responses than those who answered n a different computer -evaluations were shorter on same computer ps than dif computer, ps felt more free to be honest
Other companies promise to use breakthroughs in speech-recognition technology in call centres:
-sending customers through to people with a similar personality type to their own for more effective call resolution rates.
what does polite responses to media mean
-social rules apply to media -social responses to media are not obvious -they are unconscious and mindless
quality
-speakers should say things that are true (computers obey this well) -want to be accurate and cooperative
deep learning neural networks
-systems patterned after the way the human brain works but which, ironically, are almost entirely inscrutable to humans. -Trained with only inputs and outputs, and tweaking one or the other until the middle part "just works"
DL: textual analysis
-text is still a large proportion of the content on fb -deeptext: extract meaning from words we post by learning to analyze them contextually. -it learns for itself based on how words are used. - it won't be tripped up by variations in spelling, slang or idiosyncrasies of language use -"language agnostic"
how should one think about media
-the fact that we are polite to computers should add another dimension to understanding human-media relationships -we still must raise questions about the research
danny's input
-these things would happen gradually and we would get used to it
politness
-virtue not vice -trying to make ppl happy is the norm -practiced automatically -centerpiece of childhood socialization -genuine mark of a well-educated child
relevance
-what ppl/media say should clearly relate to the purpose of conversation -interface should say anything about things it can't do at the moment -interface must understand the user's goals -interafces that provide a single way of presenting info without taking into account multiple goals of users, risk violating rule of relevance is bad
the mayor politeness example identifies what important politeness rule
-when ppl ask about themselves, they will usually receive more positive remarks than if an independent person asked the same question -response could still be negative -positive response is inferred
Theodore Kaczynski - the Unabomber.
-wrote chapter about mass genecide -his bombs killed 3 ppl during a 17-yr terror campaign
polite why to change interface
1. statement legitimates the change 2. it makes the decision impersonal 3. it doesn't reveal the feelings of the character asking the question
latest effort to help Facebook Live users follows the death of
14-year-old-girl in Miami, who livestreamed her suicide on the platform in January.
Algorithms: AI's creepy control must be open to inspection
60th anniversary this year
4 mind blowing ways Facebook uses artificial intelligence
Facebook builds its business by learning about its users and packaging their data for advertisers. It then reinvests this money into offering us new, useful functionality - currently video and shopping - which it also uses to learn even more about us.
Now, when someone watching the stream clicks a menu option to declare they are concerned,
Facebook displays advice to the viewer about ways they can support the broadcaster. -The stream is also flagged for immediate review by Facebook's own team, who then overlay a message with their own suggestions if appropriate.
Facebook artificial intelligence spots suicidal users
Facebook has begun using artificial intelligence to identify members that may be at risk of killing themselves. -After confirmation by Facebook's human review team, the company contacts those thought to be at risk of self-harm to suggest ways they can seek help.
why the future doesn't need us
Our most powerful 21st-century technologies - robotics, genetic engineering, and nanotech - are threatening to make humans an endangered species.
The issue of AI accountability is shaping up to be one of this year's hot topics, ethically and technologically.
Recently, researchers at Massachusetts Institute of Technology's computer science and artificial intelligence laboratory published preliminary work on deep learning neural networks that can not only offer predictions and classifications, but also rationalise their decisions.
downloads: SCRNNs
Self contained software accompanying the paper titled: Learning Longer Memory in Recurrent Neural Networks.
projects : Stack RNN
Stack RNN is a project gathering the code from the paper Inferring Algorithmic Patterns with Stack- Augmented Recurrent
downloads: CommAl
The CommAI project aims at developing new data-sets and algorithms to develop and evaluate general-purpose artificial agents that rely on a linguistic interface, and are capable of quickly adapting to a never-ending stream of tasks.
projects: ComAI
The CommAI project aims at developing new data-sets and algorithms to develop and evaluate general-purpose artificial agents that rely on a linguistic interface, and are capable of quickly adapting to a never-ending stream of tasks.
two different cases may occur
The machines might be permitted to make all of their own decisions without human oversight, or else human control over the machines might be retained.
Robotic industries would compete vigorously among themselves for matter, energy, and space, incidentally driving their price beyond human reach.
Unable to afford the necessities of life, biological humans would be squeezed out of existence.
Uncontrolled self-replication in these newer technologies runs a much greater risk:
a risk of substantial damage in the physical world.
human race might easily permit itself to drift into a position of such dependence on the machines that it would have no practical choice but t
accept all of the machines decisions
computer scientists succeed in developing intelligent machines that can do all things better than human beings can do them.
all work will be done by vast, highly organized systems of machines and no human effort will be necessary.
politeness to computers rationale
computers, in the way they communicate, instruct and take turns interacting, are close enough to human that they encourage social responses.
Ms Callison-Burch
contact from friends or family was typically more effective than a message from Facebook, but added that it would not always be appropriate for it to inform them.
most interesting response in study
id be polite to a computer but I'm not thinking of the computer as a person, I'm merely responding to the person who wrote the computer program and that person is real
Danny Hillis
famous as the cofounder of Thinking Machines Corporation, which built a very powerful parallel supercomputer. -Long Now Foundation, which is building a clock designed to last 10,000 years, in an attempt to draw attention to the pitifully short attention span of our society
Ray Kurzweil
famous inventor of the first reading machine for the blind and many other amazing things.
Hans Moravec's book Robot: Mere Machine to Transcendent Mind.
founder of the world's largest robotics research program, at Carnegie Mellon University
Ray saying that the rate of improvement of technology was going to accelerate and that we were
going to become robots or fuse with robots or something like that, and John countering that this couldn't happen, because the robots couldn't be conscious.
Eventually a stage may be reached at which the decisions necessary to keep the system running will be so complex that
human beings will be incapable of making them intelligently. At that stage the machines will be in effective control. People won't be able to just turn the machines off, because they will be so dependent on them that turning them off would amount to suicide.
people are polite to computers too
if computers are social actors, then ps who responded to the same computer that taught them should be polite and uniformly, so just as if the machine were a real person w real feelings
our attitude toward the new
in our bias toward instant familiarity and unquestioning acceptance.
unstructured data
information which isn't easily quantified and put into rows and tables for computer analysis.
Artificial intelligence achieved a lot in 2016. One of the goals in 2017 should be to make
its workings most transparent
we will continue to let machines do our work bc
machine-made decisions will bring better results than man-made ones.
robots were in the realm of science fiction
near-term possibility
maxims key point
ppl will assume that violations have social meaning
The director of the US National Suicide Prevention Lifeline
praised the effort, but said he hoped Facebook would eventually do more than give advice, by also contacting those that could help.
maxims four basic principles that constitute for polite interaction
quality, quantity, relevance and clarity
computer politeness rules
rule 1: when a computer asks the user about itself, the user will give more positive responses than when a different computer asks the same questions ruke 2: bc ppl are less honest when a computer asks about itself, the answers will be more homogenous than when a different computer asks the same questions -not completely true bc computers are jus machines -ppl are more likely to be honest on a computer survey rather than an in person survey
computers are
social actors
Our overuse of antibiotics has led to what may be the biggest such problem so far:
the emergence of antibiotic-resistant and much more dangerous bacteria -DDT w misquito malaria
when media violates social norms such as being impolite
the media are not viewed as technologically deficient, can be resolved w a better central processing unit (CPU)
when technology violates a politeness rule
the violation is viewed as a social incompetence and it is offensive -media must be polite
Technology in this sense is a bit like the political system:
there are so many decisions to take that we hand the overwhelming majority of them to someone we trust.
The goal is
to help at-risk users while they are broadcasting, rather than wait until their completed video has been reviewed some time later.
Kaczynski's dystopian vision describes
unintended consequences, a well-known problem with the design and use of technology, and one that is clearly
Crime's maxims for politeness
views conversation as an exercise in which ppl try to be helpful
politness rule
when ppl ask about themselves, the responses will be more homogeneous than when someone else asks the same questions
the age of spiritual machines by ray
which outlined a utopia he foresaw - one in which humans gained near immortality by becoming one with robotic technology.
it has become a magnet for a huge amount of data about us
who we are, where we spend our time and what we like.