Oz’s Tik-Tok to the Mechanical Turk, from Neural Nets & Genetic Algorithms to Chess & StarCraft, from fighting the Coronavirus to flying Killer Drones, from Facial Recognition to Fakes, Deep Fakes, & Anti-Fakes, Artificial Intelligence (AI) is everywhere today. How did it start? What do we mean by AI? What are the basic AI techniques? How is it being used? What are the benefits? risks? and how should we manage AI going forwards?
2. http://timeandquantummechanics.comArtificial Intelligence/John Ashmead Philcon 2020
AI’s are Everywhere
three of the 22 here are real
Talos — Jason & the Argonauts
The Mechanical Turk — popular chess playing fake (18th century)
TicTok — Wizard of Oz
Robot Maria — Metropolis
Joe (transparent robot) — The Proud Robot
Roy Batty (replicant) - Bladerunner
R2-D2 & C-3PO — Star Wars
Terminator — Terminator
Rommie (ship avatar) — Andromeda
Android Gunslinger — West World
Commander Data — Star Trek Next Generation
Mecha — AI
Sonny — I, Robot
BB-8 — Star Wars
Eve & Wall-E — Wall-E
Asimo — Honda robot
Johnny 5 - Short Circuit
Sophia — The First Robot Declared a Citizen by Saudi Arabia (2016)
3. Janet — The Good Place
Ava — Ex Machina
Samantha — Her
Denise Virtual Assistant — NextOS (now Realbotix)
Real are Asimo, Sophia, & Denise
many from https://www.zdnet.com/pictures/15-of-the-best-movies-about-ai-ranked/16/
https://www.goodreads.com/book/show/19003281-the-proud-robot
4. http://timeandquantummechanics.comArtificial Intelligence/John Ashmead Philcon 2020
History
• Talos (way old)
• Antikythera
(200 BC)
• Mechanical
Turk
(1770-1854)
• Difference
Engine (1833)
Mechanical Turk: https://spectrum.ieee.org/tech-talk/tech-history/dawn-of-electronics/untold-history-of-ai-charles-babbage-and-the-turk
In 2014, a study by Carman and Evans argued for a new dating of approximately 200 BC based on identifying the start-up date on the Saros Dial as the astronomical
lunar month that began shortly after the new moon of 28 April 205 BC.[15][16] Moreover, according to Carman and Evans, the Babylonian arithmetic style of prediction
fits much better with the device's predictive models than the traditional Greek trigonometric style.[15] A study by Paul Iversen published in 2017 reasons that the
prototype for the device was indeed from Rhodes, but that this particular model was modified for a client from Epirus in northwestern Greece; Iversen argues that it was
probably constructed no earlier than a generation before the shipwreck, a date supported also by Jones.[46]
The Turk, also known as the Mechanical Turk or Automaton Chess Player (German: Schachtürke, "chess Turk"; Hungarian: A Török), was a fake chess-playing machine
constructed in the late 18th century. From 1770 until its destruction by fire in 1854 it was exhibited by various owners as an automaton, though it was eventually revealed
to be an elaborate hoax.[1] Constructed and unveiled in 1770 by Wolfgang von Kempelen (Hungarian: Kempelen Farkas; 1734–1804) to impress the Empress Maria
Theresa of Austria, the mechanism appeared to be able to play a strong game of chess against a human opponent, as well as perform the knight's tour, a puzzle that
requires the player to move a knight to occupy every square of a chessboard exactly once.
The Turk was in fact a mechanical illusion that allowed a human chess master hiding inside to operate the machine. With a skilled operator, the Turk won most of the
games played during its demonstrations around Europe and the Americas for nearly 84 years, playing and defeating many challengers including statesmen such as
Napoleon Bonaparte and Benjamin Franklin. The device was later purchased in 1804 and exhibited by Johann Nepomuk Mälzel. The chess masters who secretly
operated it included Johann Allgaier, Boncourt, Aaron Alexandre, William Lewis, Jacques Mouret, and William Schlumberger, but the operators within the mechanism
during Kempelen's original tour remain a mystery.
5. http://timeandquantummechanics.comArtificial Intelligence/John Ashmead Philcon 2020
What do we mean by Artificial
intelligence?
• Evolving
understanding
• Sense of others
• Sense of self
http://www.decadecounter.com/readingroom/kuttner.htm
A robot with self-regard
Why now:
Big Data
Big CPUs
Big Wins
we will be doing weak AI here: machine learning: minimum cutoff: unsupervised & evolving
No context, no ontology, no sense of self or other, still amazing for what it is
And if you are a programmer, go here!
6. http://timeandquantummechanics.comArtificial Intelligence/John Ashmead Philcon 2020
Where is AI now?
• NOT strong or general AI. No common sense.
No ontology. Invariably narrow focus.
• BIG DATA: Google search, Twitter, …
• BIG CPUs: (parallel processing, graphics
processors, especially NVIDIA)
• and BIG WINS: GoogleTranslate, ImageNet,
AlphaZero (DeepMind), Watson
• Here AI means: unsupervised & evolving
machine learning
• Which can perform at better than human levels,
at least with respect to some characteristics
8. http://timeandquantummechanics.comArtificial Intelligence/John Ashmead Philcon 2020
Covid-19
https://www.understandcovid.org
At every angle of attack
there are web sites to collect all Covid-19 datasets
used to find new attack mechanisms used by the coronavirus
new lines of attack against it
estimates of the spread
of the lockdown effectiveness from cell phone usage!
drug discovery
diagnosis
minimizing contact
disseminating correct information
9. http://timeandquantummechanics.comArtificial Intelligence/John Ashmead Philcon 2020
driverless car
terminator view
fatal accidents rare but not unknown
expect the unexpected
nearly safer than humans already I suspect, especially for long haul truck driving
in amazon’s own feature upload a self-driving car is used as a murder implement: they are trusted so much that the victim is unsuspecting
watch out for unicorns:
https://www.nytimes.com/2019/08/26/opinion/3d-printed-unicorns.html
10. http://timeandquantummechanics.comArtificial Intelligence/John Ashmead Philcon 2020
AlphaZero
• Learns by itself
• Beat AlphaGo
• Plays chess,
checkers, shoji, …
that’s not sexy, sexy is taking starfighters to the ends of the Galaxy & blowing up things with tentacles
games: reason: well-defined problem space
chess the first big win
— in fact first chess AI built before first computers — founders fascinated
alpha Go biggest win
but it was fed games
alpha zero plays itself
and then learns other games as well
using GAN
fix the problem of how to get enough data.
DeepMind Technologies is a UK based artificial intelligence company and research laboratory founded in September 2010, and acquired by Google[4] in 2014. The
company is based in London, with research centres in Canada,[5] France,[6] and the United States. In 2015, it became a wholly owned subsidiary of Alphabet Inc.
11. The company has created a neural network that learns how to play video games in a fashion similar to that of humans,[7] as well as a Neural Turing machine,[8] or a
neural network that may be able to access an external memory like a conventional Turing machine, resulting in a computer that mimics the short-term memory of the
human brain.[9][10]
The company made headlines in 2016 after its AlphaGo program beat a human professional Go player Lee Sedol, the world champion, in a five-game match, which was
the subject of a documentary film.[11] A more general program, AlphaZero, beat the most powerful programs playing go, chess and shogi (Japanese chess) after a few
days of play against itself using reinforcement learning.[12]
15. http://timeandquantummechanics.comArtificial Intelligence/John Ashmead Philcon 2020
Twittering Machine
• By imitating thought,
understand it,
• By seeing what is
mechanical, have a
better appreciation for
what is human,
• And change & extend
what we mean by
“human”
Paul Klee
another crank, and not the last
Photography did not replace art, complements, as in last slide
22. http://timeandquantummechanics.comArtificial Intelligence/John Ashmead Philcon 2020
Swarms
https://www.solutions4u-asia.com/emailc/SwarmIntelligenceForOptimization.html
inspired by ants
parallel processing: know your losers
crowdsourcing
drones very literally — tho curiously I don’t know how much the actual drones communicate
key to success: 100s or 1000s of entities, but if you kill the rest when one or two succeed, the economics are not bad
24. http://timeandquantummechanics.comArtificial Intelligence/John Ashmead Philcon 2020
DreamCatcher
/AutoDesk
• Antenna on right 2x
more efficient
• AI mechanical/
logical/Vulcan?
• Actually organic/
alien biology
• Neural net designed
by a genetic
algorithm
picture of the crane
advantages:
cheap at scale
reliable
explore whole solution space
https://www.autodesk.com/research/projects/project-dreamcatcher
https://medium.com/intuitionmachine/the-alien-look-of-deep-learning-generative-design-5c5f871f7d10
https://medium.com/intuitionmachine/the-alien-look-of-deep-learning-generative-design-5c5f871f7d10
https://medium.com/intuitionmachine/the-alien-look-of-deep-learning-generative-design-5c5f871f7d10
What happens when you have Deep Learning begin to generate your designs? The commons misconception would be that a machine’s design would look ‘mechanical’
or ‘logical’. However, what we seem to be finding is that they look very organic, in fact they look organic or like an alien biology. Take a look at some of these fascinating
designs.
These Long Short Term Memory (LSTM) are designed by an algorithm and shown to be more effective than the conventional LSTM (Note: These are Neural Networks
designed with memory elements). These are generative neural architectures, machines that learn to learn, more like meta meta-models. Learning apparently is not
25. uniform and I highly suspect that meta-level reasoning is a primary mechanism in learning and that is reflected by its biological manifestation. After all, isn’t learning
enhanced by diversity as well as adaptability? The same ingredients to biological survival?
26. http://timeandquantummechanics.comArtificial Intelligence/John Ashmead Philcon 2020
Face-recognition AI could only “see” Joy
Buolamwini when she wore a white mask
Google & the Chimps
https://www.newscientist.com/article/mg24732951-400-coded-bias-review-an-eye-opening-account-of-the-dangers-of-ai/
Read more: https://www.newscientist.com/article/mg24732951-400-coded-bias-review-an-eye-opening-account-of-the-dangers-of-ai/#ixzz6aymi5nAd
weakness in the training sets:
medical trials
WEIRD — Basically white, west coast, male, upper middle class, college educated, …
27. http://timeandquantummechanics.comArtificial Intelligence/John Ashmead Philcon 2020
I do not think that picture
shows what you think it shows
Preying
Mantis
Shih TzuTemple
School
Bus
ostrich,
Struthio camelus
ostrich,
Struthio camelus
ostrich,
Struthio camelus
ostrich,
Struthio camelus
Szegedy et al — https://arxiv.org/abs/1312.6199
Texture versus shape
Szegedy et al — https://arxiv.org/abs/1312.6199
28. http://timeandquantummechanics.comArtificial Intelligence/John Ashmead Philcon 2020
Avoiding Nuclear War
• 1983/Intl Tensions high
• 5 US missiles reported
• Too few for a strike
• Detection system new
• Report came in too
quickly
• No ground radar
confirmation
Lt. Col Petrov
No common sense
In the aftermath of the incident, the Soviet government investigated the incident and determined that Petrov had insufficiently documented his actions during the crisis.
He explained it as "Because I had a phone in one hand and the intercom in the other, and I don't have a third hand"; nevertheless, Petrov received a reprimand.[4]
30. http://timeandquantummechanics.comArtificial Intelligence/John Ashmead Philcon 2020
Trending Topics:
When Bubbles Collide
Curiously, the way to actually succeed at internet argument is to listen!
Like Traffic Jams — actually this was predicted by Vernor Vinge, the net of a thousand lies or some such
Try DuckDuckGo if you would rather not be bubbled
32. http://timeandquantummechanics.comArtificial Intelligence/John Ashmead Philcon 2020
Weapons of Math
Destruction
• Good teacher fired for
not meeting numbers
• Appears that previous
grades for class had
been up-revved (faked)
• “Hey, what can you
do?”
dangers of algorithms
applied blindly but fiercely
paperclip problems (soviet nail problem)
33. http://timeandquantummechanics.comArtificial Intelligence/John Ashmead Philcon 2020
If you are not paying for the
product, you are the product
• Monitor every click &
keystroke
• Use AI to assess current
emotional state
• Use AI to target with
appropriate ads for max sales
• Deny all the above
https://www.nytimes.com/2020/01/24/opinion/sunday/surveillance-capitalism.html
“By monitoring posts, pictures, interactions and internet activity in real time,” the executives wrote, “Facebook can work out when young people feel ‘stressed,’
‘defeated,’ ‘overwhelmed,’ ‘anxious,’ ‘nervous,’ ‘stupid,’ ‘silly,’ ‘useless’ and a ‘failure.’” This depth of information, they explained, allows Facebook to pinpoint the time
frame during which a young person needs a “confidence boost” and is most vulnerable to a specific configuration of subliminal cues and triggers. The data are then used
to match each emotional phase with appropriate ad messaging for the maximum probability of guaranteed sales.
One of the most famous when Target
35. http://timeandquantummechanics.comArtificial Intelligence/John Ashmead Philcon 2020
• Chess — easy for AI,
hard for humans
• Perception/movement
— easy even for
infants, hard for AI
Moravec’s
Paradox
New Yorker Sep 28th, 2020. Liana Finck, p59
"it is comparatively easy to make computers exhibit adult level performance on intelligence tests or playing checkers, and difficult or impossible to give them the skills of
a one-year-old when it comes to perception and mobility”
As Moravec writes:
Ch
Encoded in the large, highly evolved sensory and motor portions of the human brain is a billion years of experience about the nature of the world and how to survive in it.
The deliberate process we call reasoning is, I believe, the thinnest veneer of human thought, effective only because it is supported by this much older and much more
powerful, though usually unconscious, sensorimotor knowledge. We are all prodigious olympians in perceptual and motor areas, so good that we make the difficult look
easy. Abstract thought, though, is a new trick, perhaps less than 100 thousand years old. We have not yet mastered it. It is not all that intrinsically difficult; it just seems
so when we do it.[3]
A compact way to express this argument would be:
We should expect the difficulty of reverse-engineering any human skill to be roughly proportional to the amount of time that skill has been evolving in animals.
The oldest human skills are largely unconscious and so appear to us to be effortless.
Therefore, we should expect skills that appear effortless to be difficult to reverse-engineer, but skills that require effort may not necessarily be difficult to engineer at all.
36. http://timeandquantummechanics.comArtificial Intelligence/John Ashmead Philcon 2020
Never trust anything that can think
for itself if you can't see where it
keeps its brain? - Mr. Weasley
• AI should not be a
chamber of secrets
• “DeepDream” lets you
see intermediate AI
images
• & we can use the tech
used to probe human
brains to probe AI!
38. http://timeandquantummechanics.comArtificial Intelligence/John Ashmead Philcon 2020
Humanity ascendent?
• Use the AI for the many
things it is good for
• Jobs will be lost, but
jobs will also be created
• Letting us do more of
the fun stuff & less of
the other stuff
• And some stuff we just
couldn’t do before
• If we pay attention
40. http://timeandquantummechanics.comArtificial Intelligence/John Ashmead Philcon 2020
References
• Scharre 2018 - Army of None
• Pickover 2019 - Artificial Intelligence
• Mitchell 2019 - Artificial Intelligence
• Miller 2019 - The Artist in the
Machine
• Shane 2019 - You look like a thing and
I love you
Miller 2019 - The Artist in the Machine
Mitchell 2019 - Artificial Intelligence
O'Neil 2016 - Weapons of Math Destruction
Pickover 2019 - Artificial Intelligence
Scharre 2018 - Army of None
Shane 2019 - You look like a thing and I love you
Tamboli 2019 - Keeping Your AI under Control
Trask 2019 - Grokking Algorithms