Welcome to Etherbox!
This pad text is synchronized as you type, so that everyone viewing this page sees the same text. Add __NOPUBLISH__ to prevent this pad from being archived.
Nokia
Bell Labs research lab
'Developing disruptive research for the next phase of human existence'
Marthe &
Claes found a backdoor. You can go out on both sides.
Koen Koen Koen Tom
----------------
Koen from bell labs:
-------------
100 years of innovation, 17 countries, 9 nobel prizes
Transforming the world through machine learning. Telstar and the problem of low contrast of football.
Miniaturizing antennas.
- Claude Shannon, mouse, ML
- transistor
- bell/Delt? star, sattelite
(Telstar)
- dsl modem 'blue fish' as design
designed 'here' in antwerpen
"Today we focus on the network that connects
What will the network of tomorrow look like?
everyone and everything
"
"The future X network"
'core cloud'
'edge cloud'
'access networks'
"we all thalk about the cloud"
storage is increasing
price decreases 'we have to find an anwer to that for our research'
people + devices are put at the same level
(from) People/Devices --> (to) Things
past ---> future
core cloud > computech ip networks broadband access
past ---> things
core
c
loud
---
> ip networks > broad access compute tech
edge/access networks : much more datacenters
infrastructure gets distributed
more complex to manage
17 countries with research departments
US, EU, China + Israel
no locally organised topics
teams are across different countries
(interesting global distribution with Tel Aviv)
https://
bell-labs
.com
/programs/experiments-art-and-technology
Jasna Rok works with the Antwerp based lab
fashion
jasna rok = artist in residence in antwerp
"getting inspired in two directions"
Smart zones in Antwerp
, Sint-Andriesplein
"to interact with the people"
Themes:
broadband access
internet of things
custom networks
indoor networks
augmented intelligence
(it is a broad topic)
-
How to make programming more intelligent, but in two directions.
-
Programmers > Program
-
Program > Programmers
-
(?)
hc = headcount
biggest are in Finland
(history Nokia)
, Murray Hill, Paris (original Alcatell) and Stuttgart
"i
n
telligence"
Murray Hill - original lab ?
HC = head count
"there's a style guide for making slides"
-----------
Koen
-----------
The Future of Parallel Internets (= the internet is dead?)
form a technological point of view :
you develop technology and you don't know how it will be used
For example SMS, wasn't expected to be so popular
evolution for economical reasons,
now the question is how will the intera
c
tion be with the people who will use it
development:
a lot of parralell internets /
Int
e
rnet in the 'past':
connecting every* to every
*
Every*
Everything
Everyone
Everybody
Everywhere
Everytime
Everyone that can afford it
IP
= common language to communicate with each other
CEE = connecting every* to every*
compromise made to serve every*
"
you can't make everyb
o
dy happy
"
changes /more bandwith, less latency
if we want tomake everybody happy we need different internets
"if we want to make everybody happy"
The becoming
(R)evolution
--> slicing
'
Network Slicing
'
because of
5G Hype
(5th generation broadband)
"cut the network in different slices you cando different things in each slice
economics and technogies
what are the driving forces ? "
slicing is important for technologists (flexibility) and economy
needs a lot of small antennas:
On
e
antenna per street light
radio waves are radar, they cannot have obstacles in between
Slicing
-
>economics
-
>sharing equipment (one antenna per street light)
-
>flexibility
-
> Cloud Virtualization
-
> Customized behavior per slice
one antenna per streetlight
you are in teh shadow and you don't see anything anymore
different operators would install different antennas ?
smartphones, selfdriving cars need different slice of the broadband
economical angle: 25 antennas
on each streetlight is not possible
/practical
, one device will be sliced / mangaed
we make 1 antenna and make it sliceable
(sliceability vs scaleability?)
and each part of the device can be used differently
flexibility as
an
opportunity
'everything is in the cloud'
each program connected to slice of antenna, you have multiple costumized slices, calculate radio waves differently
"every slice can be different, that is the key"
"Opportunity or threat?"
"different" ?? Q: what does that mean ?
'Google slice, Constant slice
:-) (transversal slice?
intersectional slice?
)
, Netflix slice' :-
(
'
"So that people can develop their own language on their own slice", who are the people in this case?
chopping spectrum or time in parts
imagine 'colour spectrum', red is for this operator, green for the other, can also depend on time
'everything is going to be sliceable and controllable in a different way'
(controllable by who/what?)
(!!)
"It's a multiplex"
What protocol is deciding who can send and when?
- interslice protocol
-
inside a slice you can speak a different language
'you can send your packets in another language, but the packets will be dropped'
You send your content in a language, but with a header. This is send in packages.
"if you are too agressive you push away others, if you are not agressive enough you get pushed away by others"
Is this letting go of network neutrality?
"Do we need to be network neutral?"
Everybody, in the sense of everybody that can afford it to use a slice.
Can you rent a slice? It is like renting a server.
Koen:
we don't know who is deciding on the slicing ?
"We just create the technology".
who is the slicer?
"the big players" "is proximus here?"
"Amazon?"
Threat = further monopolisation. Of course the internet as imagined never really happened, but this might make it even worse?
only a few key players will dominate it
here's an opportunity for 'everyone' to use it but actually it will be controlled by a few
'
everyone
'
as in everyone who can afford it
Q: who is 'everyone' ?
'the infrastructure is already there. if
'
everyone
'
agrees, then it's only about making the programme'
If you look at cloud-computing ... it is not a monopoly ... because of concentration it has driven down the prizes.
(discussion)
'you don't even have to pay in advance, you can pay as you go. economics of scale has made sure that everyone can run a server in the cloud' (Tom?)
Q: is pay as you go as a concept valid? why not 'pay as you can' ? or pay as you want ?
Tom(?): Open competitive market, will make the rent of a server cheaper and available for more people
Jogi: Fascinating technologies developed in labs vs. the capitalist economy that breaks good ideas. Will there be room for smaller parties? The public internet is dying/fading.
technology is one thing, what capitalism makes out of technology is another thing
Technology drives the economy today
we all want global latency
Making latency lower
depends on what is the round trip time
Google put datacenters close to each other, have larger bandwith
concept of queueing
now making cues always smaller, total time you have to wait is 'access to Google datacenter'
from "world wide web" to "an access t
hrough
the Google datacenter"
(they are the winners)
behaviour of TCP could not change
"The one that could make a lot of datacenters neirby, is actually the winner"
difference between evolution and revolution
evolution: continuing from current situation (Google's position for example)
revolution: being aware of what's going on, and taking action if it's necessary
Wendy: are there already regulations for slicing?
standards are being developed
to get around network neutrality
, 3GPP (?)
network neutrality is always part of the discussion
separate private network next to others will now be possible
Roel:
Who is taking care of the standardization?
3GDP?
5g
Pieter: difference
b
etween the basestation and the endpoint?
Koen 1:
The base stations are coordinating the package that can be send.
If it is very crowded, a coordinating system is more efficient ...
What will be the impact on the density of antennas?
Koen 1: T
hey
(the antennas)
are nearby and don't need a lot of power
'it is reducing the radiation'
'we never know what unexpected impact new technology might have'
Femke: what is the difference between every slice?
Koen
1
:
selfdriving cars need to be very reliable, and very low latency
. Sending more things in parallel is a way to make sure that packages arrive.
This is different from a network that prefers a lot of "
throughput
", where it does not matter to miss a few packages.
(but is that really the only difference that can be made?)
(yes indeed this is very practical/problem-solution thinking)
--> would be interesting to hear more about their concept of 'difference' and the potential for interaction between slices.
---------
Koen
Hooghe
3
: Turning
old copper into new gold
fiber/coax/wieless are different physical layers
Cooper based transmission
copper based access
connecting homes/industry to the internet & access to datacenters
Smart Digital Network "Fabric"
big cables telephone, connecting every home to telecenter, orginally 4Kb
UTP Unshielded Twited Pair > voice >>> 4kHz (telephone services)
Digital Subscriber Line (DSL)
1st generation: voiceban
d
technology
1000 users per node
1-2 MHz
max 10km
2nd generation: Asymmetric Digital Subscriber Line (ASDL)
Slidequote: "NOT FOR DISTRIBUTION SOLELY FOR (...) Nokia"
spectrum upt 2 ( dmt > modulation)
dasl - has history in technologyy for video on demand in hotel rooms
max 5km
Asymmetric
/ Asynchronic
(more download than upload)
(time-space)
1-2 MHz
1000 users/node
sufficient to stream video
-
ISDN (1st generation)
-
ADSL (2nd generation) -- connections not through a grey box anymore, but a blue appealing stringray modem
-
VDSL (3rd generation)
-
G.fast (4th generation)
-
MGfast (5th generation) -- MultiGigaBit fast
"not anymore these grey boxes, we developped a bit more fancy modem, what we called the stingray modem"
"of course people want always more and more"
3rd generation:
VDSL: very high bit rate Digital Suscriber Line
10-30MHz
max 2km
500-1000 users/node
Need to bring equipment closer by the houses
because you have a lot of noise
street cabinets in every street in Belgium
PON = point to point
"high bandwith backhaul" (??)
this is most common used access technology in BE
The next generation is:
"vectoring"
VSDL2 Vec
FTTH (?) (is it: Fibre to the home?)
twisted pairs > cross talk >
"mature technology"
eg Proximus
The Netflix effect?
massive calculations in central office to predict optimal wanted signal at the other side: from 50Mbps to 150Mbps
predistorting the signal
Next step:
new technology, first deployments
today
.
4th generation
G.fast
FTTP
212MHz
max 300m
500-2000Mbps
10-100 users/node
"We bring fiber very close to the user, but not yet into the users home." The last part of the cable is connected by copper.
Fibre To The
Curb!
FTT
F -- Front door
5
th gen
. MGFast
FTTFrontage
20-300m
1-10 users/node
424 MHz
2-10 Gbps
"almost termination the single user network node"
want to avoid to dig up all cables to all homes and replace them
can be built on top of the copper network
(DSL = ?
Digital Subscriber Line?
)
thanks to improvement on processing power
'by distributing
the access node
into smaller nodes
, we increase the value of the available copper'
equipment comes always closer to the home, still needs to come closer
"increasing fiber penetration"
MGfast will simultaneously transmit in both direction on all frequencies, doubling the aggregate spectral efficiency
(discussion)
Keith?:
async
h
ronous
/assymmetric character of ADSL
was in my opinion the moment that the internet started to die
it killed the possibility for one to self-host
Koen
3
:
'most people are just consumers
. for them it is not an issue at all
'
Keith: it can be reversed technologically, but not economically
Can you can get a synchronous uplink? Apparently, in Belgium you can.
"symmetrical"
is at a business price.
"we have the solution and we have it working"
Koen 1: "But, also IP-addresses are changed actively, which is making it hard anyway to self-host"
Koen
2: "it is all business"
Jogi: it is has never been symmetrical.
big companies Instagram youtube actually depend on substantial upload capacity
"you want to be reliable" (?)
These days you can rent runtime.
lambda time.
a reason to self host is that somethi
ng
is mine
I have
/own
a small part in the internet
.
'it is my own'
it is not necessar
i
ly for selling something
I should be able to share it and not paying more for it because I am connected to the network
uplink is problematic when you make your home device accessible f
or
externals
Wendy: Cloudifying uses more energy? Koen 3: not always.
depends where you do it.
Datacenter in Iceland is more interesting: thermal energy, no cooling necessary
Koen 3: In general, scaling up is a way to use less energy, as you can use one bigger cooling system.
"So you are saying we better melt the glaciers in the North than using a fan"?
Roel:
How many years do you foresee copper still being used?
Koen 3:
"we are predict
ing
the death of copper since forever."
Since
'
95
(30 years?)
we have said this will be the last thing we do with copper and then it will be fibre.
most economical and energetic way to do
case of Sweden: removing copper to install fiber because of the efficiency it will bring to the users
In Belgium the copper is still there, caused by a monopoly situation:
cable operator / telecom operator
trying to maintain their copper network
(how come Claes' jumper blends so perfectly with the wall?)
haha, yes!
"fixed line will disappear with our kids"
kids grow up with a mobile idea of connectivity.
"the connection will be better in your garden"
fixed line doesn't mean anything to them, except when they want to serve (koen X)
An: Cutting down trees in North-NL is shocking, because they give too much shadow (?). So no more large trees?
(How does this work with the shadow?)
---
-------
TOM
------
Tom van Cutsem (and Bart Theesem who is not here)
Augmenting software development through machine learning
"
Future of software development:
augmenting softw
ar
e development through machine learning"
Beside network technologies, Bell Labs also works on: augmented intelligence
'make experts more productive at things'
Who of you writes code?
"50% is fine"
Coding the tip of the iceberg -- re-using tons of sourcecode (ahh the full appropriation of F/LOSS)
Inner Source (!?) vs Open Source
2 bil lines of code vs 85 mil code repos
Also at Nokia we use a lot of software to make our own software.
Exponential growth, the iceberg of coded lines becoming bigger
Tom uses the metaphor of the iceberg, that is questionable: for him the tip represents the amount of visible
(newly written?)
code versus the tons of code that are invisible
(already written)
under the surface. Saying 'the iceberg will grow' when referring to the use of computers sounds a b
i
t cinical, denying ecological impacts.
Yes, also in the sense of thinking about F/LOSS as an ecosystem. The language of efficiency and productivity projected on the iceberg...
'how to effectively search all the code that is there out in the "open"
'manually searching through code is inefficient'
Most of the time in the work of software developers, is in searching for usable code
"IDE is the tool that developers use to develop code"
AI Co-pilot code editor
-- code compa
ss
"oh!!
sci-kit learn
is my default example"
"other developers also used ..."
recommendation system
scraping/crawling the Open Source ecosystem
www.code-compass.com
Cute code names as soft-pets (?): nupic, pandas, numpy, matlplotlib, pybrain, tensorflow, scipy, keras, pypi
^ images of O'Reilly book covers come to mind
re-mapping disciplinary separations back onto code / divisions of work / ways of doing things
"if you are a developer, you like to stay within your editor"
"the developer does not even have to type anymore"
open source python projects as a universe of dots
networks of software projects, clustering
"every dot is a python project"
clusters represent a subpart of the python ecosystem
How different are all these software projects actually?
Wendy: What are the sources that you searched in?
sources: Github + package managers (eg pypi, npm) + metadata from LibHunt
"quality control is through the community"
the text editor Tom uses is Visual Studio Code
= Open Source from Microsoft
but it could work with other editors too, such as Atom
analytics and pattern matching used to determine what is 'relevant'
(thinking about crawlers and what they do to fish population)
Nearest neighbours and multidimensional vectors.
'Unsupervised learning from Big Code'
extract library dependency data from source code > learn from vector representation > compute nearest neighbours
Femke: in your talk you mention code as ecosystems, how is this tool contributing to the ecosystem?
the result that the search engine produces is tailored to the domain of the user
(is he saying that this tool contributes to the F/LOSS ecosystem by promoting sameness, zoning, not doing things differently?)
he assumes programmers are more interested in sub-ecosystems
recommendations through homophily
"I'm already here, what else is there?"
yogi: when you are computing nearest neighbors do you run into hubs?
overheard keywords: storm, spring nature, ....
---
Koen 1
:
Future of Data Streams Platforms
"Supporting citizens in their daily activities"
constant feeds of information
almost immediately consumed data
Big data analysis: stored information, analytics later
Other situation: active analytics, in the moment
Typical image of what we call a smart city
Picture with straight lines connecting houses, boats?, cars, weather, windmills, restaurants, bike parking places
Trying to discuss with IMEC
"to help them realize it"
Experiment at the St. Andriesplein
Connection between lights and what happens in the street (basketball activities)
"The question is ... What kind of software do you want to use for this project?"
discussions about this leads to piles of different choices
you need a mix
[A diagram of a network, with the following nodes
/edges (interesting it is more about zones?)
:]
far edge
metro edge
core
public
core cloud
edge cloud
slice
You might want to have things running
close to the sensors
....
in the cloud
then you have a little bit of blueprint
"You want all these technologies to be just like lego blocks"
"like a puzzle"
'plumb it your own way'
decompose it and distribute it
How to turn this into reality?
stream broker: software intelligence that makes sure that technolgy is connected (?)
haptic
filter
sensing
orchestration
stream broker
transform
AI/ML
analytics
Store
merge
Stream Broker
world wide streams
sources/sinks
an [umbrella] library
stitching software together
writing a data flow
what if the pieces do not fit together?
library language within typescript
typescript, a variant of javascript
a function that transforms the output of one think into another
you can write easily plumbing functions, forwarding output and inputs
(plumbing --> link to piping?)
YES. But are they promiscuous?
This type of piping is not really placed in between? But as an umbrella covering the other things...
connecting video streams, input video, import functions (for ex tensorflow, gstreamer) pipe out
cam.pipe
You author it centrally, to make it easy
But the way you compose it as a data flow, using our own compiling techniques ....
guarded area, people in danger (person moving close to robot arm)
input video [far edge] >>> video analytics (foreground, contours, fence) [metro edge] [core] >>> SDK (=?)
"And then, when the blobs enter the zone ..."
"object recognition of tensor flow will tell us if it is a human or not"
At the St. Andriesplein
Dashboard (weather, temperature, cameras, sound)
To control the lights on the square.
green ghosts: motion overlayed on still images of the square.
"the reason why we do this is privacy. We want to explain what we are doing but we are not allowed to use the images from the camera."
as a privacy tool
At 4am in the night, we keep the lights off, as it is not a correct time to play basketball.
"person: 94%"
probability that this is a human
Tensorflow out of the box is used.
[what datasets are behind this? Who's bodies are used to identify others on the street]
"let's detect what they are playing with, is this really a basket ball?"
Smart Lighting - Under the Hood
how 'smart' cities need copper being turned into gold, because we need 'smart' cities so that need more copper turned into gold
An: The example case is a square and playing basketball. Is there an ethical commission linked to these developments?
Knowing there is a rght wing mayor in Antwerp, is there also an ethical commission i
n
volved in this? what is the border between light for playing and ...... ?
IMEC? What organisation is this?
(a consortium of university and commercial labs)
"There are big boards on the square, where the project is announced as a playful use case. I have no idea what implementations are planned."
We as a technological company are not involved in the ethical discussion.
We only need to follow the privacy rules. We're not allowed to do post-analysis and store the video's without the green ghosting.
(all about innovative technology but not a lot of imagination when it comes to adverserial use)
TESTCASE: make motion / heat blobs that are not recognised as humans by Tensorflow. Geometric fashion, crippling devices, coupling bodies into 1,
oh, Tom is addressed as Bart (bug?)
Ah, now he is addressed as Tom again
Is the consumer allowed to decide when to turn the light on?
Could you not have put a switch on the square? It could have been more effective/efficient?
Aha. Surprise:
"We didn't make this technology in response to a request of the city of Antwerp. "
"We saw a parallel between different requests and technologies before, where you always want to combine softwares."
Q: so what is the technology they developed that he is talking about? making an urban-political switch board?
This project is a starting case..
"You want to avoid that everybody just can turn the light on and off."
Oh. not for Every* I guess.
the main switch :-)
Q: The technology is so attractive/interesting (for many reasons) to develop ... use cases come afterwards?
Also, there is a lot of work put into the technology, seperate from the use cases. As a standalone project.
Is this a way to remove/ignore questions around responsibility?
Remark: In an example in france, before starting the project, ethics is discussed with a citisens group.
also, there is a sensor streetlights that only switches on when people pass, but they never thought about the darkness this creates around it: a pedestrian can not switch on the other lights for their own security, to check t
h
e area for danger.
How do we keep ethics in the pipe :-)
We as a company are more interested in developing these techniques in the context of industrial production, robot arms, etc.
"there is always a difference between technology and how it is being used"
Wendy: "always take the worse use case into account"
Anne: ethical co-pilots
Eodie: ethics outside of engineering