3. Exponential
computing
What makes revolution?
early adopters: open hackable
platforms with a culture of innovation
free (Internet): innovative apps do not
need financial justification to be
developed
the stupid network
6. Software &
hardware
Software lets you visualise the
physical; takes you above hardware
& infrastructure
This flexibility allows you to delay
decisions (!)
8. Cool new things &
devices
Contact lens that measure concentration of sugar
Google Glass (immediate computing)
Oculus (turned down $1.1B deal without having
product)
Magic leap (600M seed round)
HoloLens
CastAR
10. Networking
Problems with multiple wireless low-power
standards (Zigbee, BLE & etc)
Seems that BLE is winning the race
Nest —> Machine - Machine - Human network
Probably Google bought them for 3B not
because of thermostat, but rather to have
gateway to talk to other smart home
devices!
12. Wearables
Problem: people buy it, but dont wear it
When people are healthy, they dont care
about wearable health devices
Intel Edison & Baby Jumper
application
13. Killer app
Wearables & sensors are waiting for
killer app
Probably it is in the field of
context: putting smaller devices into
environment & sensing what is
happening there
15. Bitcoin &
blockchain
Global & outside of government control
No central bank to meddle with it or fix it
Senders & recipients are anonymous (or
maybe not)
Transactions are irrevocable
Quicker/slower than other instruments
16. Bitcoin & open
hackable culture
Bitcoin lets anybody innovate in
Money, Title, Transactions
Ethereum platform
Research: smart contracts
21. DeepMind Tech
Input: raw pixels
Output: value function estimation of future
rewards
Acquired by Google in 2014
Plays Atari with deep reinforcement learning
Video: learns from game & comes to intelligent
strategy that developers have not envisioned
22. 3 myths
You cant just sprinkle a little bit of AI
Garbage In - Garbage Out (you have to clean &
shape data before using it)
Smart algorithms are not the same as human
level intelligence or sentience
—> start with the problem, not AI algorithm.
Framing the problem is THE important step
23. AI comes with
tradeoffs
One side: job disruption, human
identity, risk amplification
Other side: faster, cheaper, better
problem solving
24. 4 key areas
Verification - meets the spec
Validation - spec meets the
problem/situation
Security
Control
26. AI
Symbols - computer are very good at
Perceptions - just become good at
Concepts - just starting —> why? —>
cant make decisions without data,
need to have mechanical models
about reality
27. Where we are
Vicarious —> next generation AI
algorithm that passed Captcha test
Future states of cow —> AI imagination
Human brain had no upgrade fo 50K
years
Siri & Google Now - long way to
understand our intentions & wishes
28. More on where we
are
VIV - spin-off from Siri
Intelligence become a utility
IBM watson
29. Development
vectors
Deep learning algorithm - champion now
A_PIE - raging Pose illumnation
Expression
Face 2 algorithm —> 99% identifying
faces in different poses, emotions & etc.
Better than human
But there are mistakes
31. Application
oriented AI
Machine Intelligence Landscape
Gecko conference
Attensisty - sentiment analysis
TalentBin - talent search engine for entire web
Brigtherion
Narrative Science
BitSight
Gaggle - solving business challenges through predictive analytics
Experfy - hire data scientists (the problem: you expose your data)
Watson Ecosystem - 100M venture capital, SU may connect with them
Neil’s workshop at Stanford: using AI in decision making
32. Books
Race against the machine
Kasparov - How life imitates chess
AI changes balance of power —>
Radical abundance by Drexler
Abundance by Diamandis
37. Intro
Frequency of earthquakes —> same
exponential behaviour
Inflection point vs critical point
Uncertainty is intrinsic
38. Intro
Identify your biases.. and not all surprises are bad news
Lousy quality is not a bug, a feature (Kodak and
digital cams)
Sacred cows make the best burgers
Forecasting isn't prediction
Actions in the present affect events in future
Quickly come to conclusions - the whole essence of
forecasting
45. Reading
DNA Microrays
23 and me
Cost per genome (biology now falls under Moore’s Law):
2001: 3B
2007: 1M
2013: used car
2015: 1K
2016: pizza
2020: flush of a toilet
Actually DNA sequencing tech is even faster than computation: 5-10x vs 1,5-2x
49. Reading
ApE - plasmid editor (software
development)
DNA 2.0 -> print, build, test
CRISP/Cas9
Drag’n’Drop Engineering
Genome editing with CRISP —> toolkit for
genes
53. 2nd popular FAQ
how did they allow? —> 7 different
levels of government
FBI/BioCurious join conference in 2012
Cow-free Milk & Cheese
Rise of Biotech Garage Inventor
diybiology.org