Loading presentation...

Present Remotely

Send the link below via email or IM

Copy

Present to your audience

Start remote presentation

  • Invited audience members will follow you as you navigate and present
  • People invited to a presentation do not need a Prezi account
  • This link expires 10 minutes after you close the presentation
  • A maximum of 30 users can follow your presentation
  • Learn more about this feature in our knowledge base article

Do you really want to delete this prezi?

Neither you, nor the coeditors you shared it with will be able to recover it again.

DeleteCancel

Make your likes visible on Facebook?

Connect your Facebook account to Prezi and let your likes appear on your timeline.
You can change this under Settings & Account at any time.

No, thanks

T-Systems Exponentials

No description
by

Brad Templeton

on 19 October 2016

Comments (0)

Please log in to add your comment.

Report abuse

Transcript of T-Systems Exponentials


Free Space Optical (FSO) at 160 gigabits (Pisa)
THZ radio link at 3 Gbps, some day 30 Gbps
Time for “Glass Roots” to enable real competition
Fiber now 255 Tbits
per second with
DWDM





25 billion
calls at once!









Single
multi-core fiber
LAN Wireless
Picocell approach offers unlimited bandwidth
Haas: Wireless through LED bulbs
Reardon claims “DIDO” system offers full bandwidth to every device in the room
Low power “personal area networks” coming
Spectrum allocation becoming obsolete
But others want to auction and own it
Cognitive Radio
Whitespace
Network “management” & neutrality
Battles more political than technical

GPUs generating too much heat
Portability requires low power
Contact lens projector the holy grail
Registration on real world difficult
VR sports, gaming & shopping
Augmented/Alternate reality games
In Graphics
Looking to
bridge the
"Uncanny Valley"
User Interface Breakthroughs
Replacement for the keyboard needed
Speech Recognition,, natural language & translate
Siri, Watson still have far to go
Gesture interfaces from 3-D cameras
Reading
Lips
Emotions
Gaze
Haptic feedback
Displays & touch
everywhere
All while small & efficient
The ball-n-chain: System Administration
Is the Apple “walled garden” the only answer?
Challenges:
Improve performance in face of complexity
Provide real time operations
Virtual machines
Cloud computing and apps
Software development needs better re-use
Ambient Intelligence
"The Internet of Things"
Brad Templeton, Computing Track Chair
Chip-Multiprocessors: “Single-Chip Cloud” Datacenter on a Chip
6/29/2010
48 (45 nm)
Chip-Multiprocessors: High Performance Technical Computing
6/29/2010
http://download.intel.com/pressroom/images/Aubrey_Isle_die.jpg
32
32 cores X 16-wide-vector ≈ 1 TFLOPS peak
Chip-Multiprocessors: Laptop
6/29/2010
http://download.intel.com/pressroom/images/corefamily/Westmere%20Die%20Flat.jpg
2 (32 nm)
Chip-Multiprocessors: Server
6/29/2010
http://www.intel.com/pressroom/archive/releases/2010/20100330comp_sm.htm
8 (32 nm)
COLOR SPECTRUM

DEVICE
CONTROL

DIMMER

Images courtesy of Cisco
End of rapid scaling of single threaded code
At least it’s still scaling up – 10-15%/year is great
40%/year speedups for parallel software on parallel hardware through 2020
Assuming energy/computation falls 40%/year?!

Great disruptions in status quo bring great opportunities

Moore’s Law Forecast

Automatic parallelization: not any time soon
Legacy code “renovation” is very challenging
Serial assumptions are baked throughout many abstraction layers / programming interfaces
E.g. Windows’ GUI thread “pulls you in”
Backwards compatibility
Prefer to chase new rainbows

New experiences and new scalable features achieved by composition of old with new
Prospects for Parallelization of Legacy Software?

What else can we do with billions of transistors?
Idea: chip-multiprocessors (CMP)
Tile die with 2, 4, … CPU cores – 2X cores/2y
Simpler cores -> even more cores
Finesse power with lower V, lower freq, sleep
Finesse memory wall with memory parallelism

“We can’t think of anything better, so let’s see if programmers can make these things work!”

Towards the Next 100X:
Explicit Parallelism – Multi-Core

[The Manycore Revolution, SciDAC Review, Fall 2009]

Serif's and scattering bars

Context
Cheaper, Smaller, Lower Power
1,000 distinct apps
Medical:
EEG
Blood
BP
Proteins
Alcohol
Fields:
Radio
Magnetic
Capacitance
Pressure
Location
Ambient:
Sound
Light
Humidity
Temperature
Barometer
Beaglebone

of
Bandwidth
Cool New
Devices
'n Things
Moore's Law on Everything


What makes an exponential revolution?
Insatiable market demand for doubling, little demand for incremental improvement
Open,
Hackable
Platforms with a
Culture of Innovation
"Software is eating the world"
- Marc Andreessen
Software lets you virtualize the physical

Takes you above hardware and infrastructure





The flexibility allows you to delay decisions
What fraction of the Fortune 500 would be off the list in 10 years, according to Babson Olin School of Business?

10%? 25%? 40% 60%?
But 15 steps -- halfway -- is only 20 miles!
(2014: back from bankruptcy,
13K employees, $770M cap)
Exponentials
At
Robocars
Saving Lives
Energy
Innovation
Policy
Parking
Human Drivers
3 parking spaces per car, 60% of land in LA
Over 25% of USA energy budget, more of the CO2
1.7 light years driving!
The Reality
They are coming
Cars galore with:
Self parking
Lane keeping warning/assist
Automatic Cruise Control
Pedestrian and collision detection/warning/braking
Programmers can save the world
What car do I need
today?
"The right vehicle for the trip"
Mostly an ultralight
Buy what you use, not what you might need
Your car rents itself out
Or don't buy at all
All the economics of cars change
Greener than transit…
Time for Apollo-like resolve
Robots don't care how convenient it is to refuel
Enables all alternative fuels
200 million tons CO2
Ending overseas oil imports
Political issues
Legal, but...
Must not be over-regulated
Jurisdictions will compete
Possibly Singapore, China, India, Israel, Japan
Children, the aged, the Drunk
They don’t park they stand.
Can stand in driveways, in front of hydrants, double and triple parked
The more cars on road, the more road capacity
PDLBKBR
Brad Templeton, Singularity University
So NOT speaking for Google
Moore's Law
Comes to transportation

Competing Innovators vs. 19th century approaches
Mike & Maaike
Mobility
on
Demand
A pleasant work environment, face to face
No central control or new infrastructure
The "poor man's teleporter"
What car do I need for my life?
Plus 100 cars in Gothenberg in 2017
iPhone Speed or Industrial Speed?
Drive
Deliver
Refuel
Park
One Vital Italian Product
Disruption:
The existing big players rarely survive.
Wither Ontario?
40% of fatalities due to drinking
Netherlands: 640 deaths
and 162 cyclists
Germany: 3,400 deaths
(about 65% of USA per km)

2.4M accidents, 377K injuries, 67K serious
167 deaths, down from 400 10 years ago
(About 30% better per km than USA)



3,800 injuries
3650 Deaths
(down 50% from 2001)

Double UK, 4x Netherlands, 30% better than USA

260K Injuries

Most dangerous city in Europe?
1750 Dead
From 3500, 10 years ago

190,000 injured
~40% of USA per km or per car
7,168 Injuries, 160 fatalities
Almost double US rate per vehicle
62,000 Fatalities
Some studies say 275,000
many times rate per car than USA
190 deaths
5,000
injured

Half of US
per km
US: Kill 34,000 people each year - #3 in years of life lost

NHTSA: $871 billion cost of accidents ($600B suffering) - 9% of GDP, 29 cents/mile (0,16 €/km)

50 billion hours spent driving - 8% of GDP
Canada: 2000 deaths, 150K injuries (Just under USA per km)
India: 142,000 people -- worst in the world.
Similar per capita, but 7x worse, per car, than the USA. 5L seriously injured.
1.2 million killed per year
Four Models
1. Retain Traditional Cars
2. Buy private robocars
3. Buy a better car, hire it out
4. Summon Robotaxi
(at first instead of 2nd car)
Two Cultures
A car with a computer
A computer with wheels
$70B from Hyundai
5,800 dead
344,000 injured
11.4 deaths/100K,
Same as USA
2.5x the rate per km!
Industries

Automotive (duh!)
Energy
(all kinds)
Real Estate
Insurance
Parking
Auto Loans
Retail
Delivery/Shipping
Medical
Repairs & Service
Transit
Food
Ownership
Radio
Construction
Robotics
24,000 deaths
4 million accidents

Going the
wrong way in
8% of accidents


4x USA death rate
per car
1.7m km real driving
16,000 deaths/year!
3x USA rate per person
11x USA rate per car
270,000 dead/year
712 deaths
2x USA per car
Much more per km
3250 morts
1/2 per person
Just under USA per km
765 Deaths
almost 2x USA per car
Full transcript