Here are some notes from the first day of the Technology Review Emerging Technology conference at MIT – I’ll probably add a bit of commentary later, but here they are raw.
Let me say, it’s a pretty stunning lineup, and some of the things they are working on are truly world changing – not just “hitting keys on the monkeybox” as Joshua framed all our social computing stuff.
———
Nicholas Negroponte
the $100 laptop
solutions to most problems include education – or may just be education
particularly primary and secondary schools
in emerging nations, the problem isn’t connectivity – it’s not a solved problem, but there are enough people working on it, regulatory regimes are changing
for education the roadblock is the laptop
people often start working for the emerging world as charity or similar, this is fine, but maybe not always the best reason
[weirdly go into Powerpoint rant]
made his first powerpoint for this!
Logo, computing for children at MIT (Seymour Papert), 1980s
Costa Rica – best, longest use of computers in development
built two schools in Cambodia, with computers – 1999 – no water, electricity, even roads
told the kids to take home the laptop to use
no kids were allowed to open the laptops – parents thought it would break
parents loved it – the computer was the brightest light source in the house!
in 4 years, 1 laptop out of 50 broke
100% of the AC adapters broke
why? ownership. proud to own and use.
families made cases for them.
in Maine, kids get a laptop during their education
“one laptop per child” – olpc
non-profit – means price will go down
scale – important for mindset. tech cos focus on bigger, better, smaller, not cheaper. sclae gets you strategy in these companies.
countries—free
corporate partners – needed to get this moving
25% MIT funded, 75% others
countries pay in advance for a million units
“impossible” – means MIT can do it
50% of the price of a laptop is sales/marketing/distribution
display is biggest cost – $35
75% of the hardware is to support the weight of the OS and software – obese and unreliable
7.5 inch screen dual-mode, one transmissive, other b&w 4x resolution sunlight-proof display (ebook!)
has to be wind-up
Open Source
Mesh network
2Mbit can serve 1000 kids
Grey market – trying to make them hard to sell, and solve the need for cheap laptops other ways. Make them distinctive, like a post office truck or an Army jeep – unsellable
parallel commercial market – maybe $200
design is important – not cheap, not a toy
shows the latest picture of it!
personalisable! – case schemes – maybe engrave the kids name on each laptop
AC cord is the strap, several power modules
looking for a 3rd party accessory market to grow
rollout – 5 countries plus MA – china brazil thailand south afica (malaysia?)
beta is 15 million
china and india – half of all primary and secondary schoolkids in the world
Nov 17th – launch – tethered prototypes
year 2 – 07? 100-150 million units
3 times the current world production of laptops
haven’t talked about the content side
part wikipedia-style
also have Squeak and Scratch and other projects – tools to be creative
will move to flexable and printble displays, e-ink
10:1 ratio for handcranking – can do it for ebook mode, at least
eink would give 100:1
****
questions
what about China, and the crackdown on Internet and censorship?
he is clear with heads of state – we are selling you a trojan horse
china spends $19 a year per child on textbooks
he avoids the question
what’s it like being on the other side of the fence – making real products rather than just researching for others?
On-the-fence – it’s not a commercial machine
only way to be able to stand next to government
how do you break through the education mafia and the dept of education?
Maine was key
no one went from being enthusiastic to the other way, but the others have migrated to being enthusiastic
truancy plummeted
PTA meeting attendence skyrocketed
kids participated more in classrooms
is now the time to rethink the entire teaching process?
this is not teaching as we know it
only a small part of learning comes from teaching
this is a tool to make it more continuous and seemless
it’s going to take decades
is this a Marxist thing? Well, education is always mainly state funded
curve downwards, $100 is still to expensive, gonna have less memory etc. if possible
how to stop pornography?
experience in Guatamala has been opposite to the example given
Gutenburg doesn’t get flack for it, cell phone industry don’t get flack for terrorism (!)
finding ways to block things and look for best way forward
*********************************************
Jeff Hawkins
runs a nueroscience institute(!) as well as starting Palm,Compass etc.
can a new theory of neocortex lead to the creation of truly intelligent machines?
Hierarchical Temporal Memory (HTM)
published a book called On Intelligence
a biological theory
now have a mathematic formalism
formed a company called Numenta
will describe the technology, not to prove his claims (only got 45 mins)
[gets his brain out]
neocortex – 1000 square cm if ironed out
has differentiated areas
about half neuroscientists believe this principle
senses are a filter (million of senses parallely processing)
world – is causes
through senses
to cortex/HTM – “beliefs”
causes are persistant, inc. words, ideas, all patterns
want to build a representation
what does an HTM do? – discovers causes in the world – infer causes of novel input – predict future causes and input – creates motor behaviour
you can determine the cause of a picture in about 300ms
going to talk just about the first
discovering and inferring causes has proven to be very difficult
– machine learning, visual pattern recognitionHTMS use a hierarchy of connected nodes
each node discovers causes, passes beliefs up and predictions down
why does a hierarchy make a difference?
belief propogation – quickly have entire system reach best overall consensus
shared representations – give generalisation and efficiency (seems similar to OO reuse)
affords a mechanism for attention
matches hierarchical causes in real world
each node stores sequences of patterns – changing sense data leads to stable beliefs at top, which lead to changing predictions and behavious at the bottom
How does each node discover causes?
Nodes learn common spatial patterns, learn common sequences
things that happen at the same time are likely to have the same cause
[this fits with Seth Roberts’ theory on diet]
can assign common causes to patterns, and ignore uncommon patterns
Use context from above in hierarchy
bottom-up doesn’t account for pictures of both apples and bananas meaning fruit
Do they really work?
no time for a live demo
built a simple visual system, 32 by 32, 3 levels of HTML, bottom level each node looks at 4 pixels
trained by moving it around in the display. over time is very important.
Numenta’s goals
maximise postive impact of HTM technology
build a community to develop HTM and HTM applications
product mid end of next year
think these can get very very big – 300 million nodes in the human brain – can go further
what’s it for?
go beyond biology – go faster, larger and give exotic senses
this is not about being human like things at all – no emotions
what humans find easy that computers find hard
vision, language, robotics
security to self-driving cars
add IR, sonar, radar
discovering causes in exotic worlds
geology, markets, weather, physics, genetics
www.onintelligence.org
www.numenta.com
***
questions
howcome the brain isn’t actually connected like this?
actually axioms are large but sparse
which is more important – discovery, inference or prediction?
too hard at the moment to see – and they’re inseperable
most excited in the discovery
example of exotic sense
brains are a 2d surface, not an efficient way of doing this, just what nature stumbled on – so can design computer systems in 3d and 4d systems
how does it stop erroneous discovery?
beliefs are just beliefs
we have many false beliefs
an inherrent problem
has to do with quality of data, especially primary data
can only correct by showing good data and conflicting evidence
any insights into how humans and groups learn?
(humans as nodes)
complex systems – markets, ant colonies, brains
maybe a general theory of complexity
**************
panel – Nolan Bushell, Dean Kamen et al
Nolan Bushnell, uWink
The quest & curses of the games industry
empty demographics – 15% play games
portable connected experience
arcades are dead
consumers will have
terabytes of data
gigabytes of bandwidth
secure permission only access
invisible technology
with you
high speed connectivity
geo information
bio sensors
do we really want photo realism???
diminish the potential for product differentiation
next round of consoles will be the last that are commercially successful
increasing software costs
drops in software innovation
blood ‘n’ gore loses cartoon defence
1982 – 44 million players
currently 18 million – what happened?
big bucks from a few people
uWink is trying to create stuff for other demographics
gaming so unpleasant on the small screen – portable is overhyped
other areas
megagames
machinma
game technology enhanced sports
immersive movies
micro music
self contained games
a million channels of entertainment
uWink Media Bistro
games used to be for parties
a framework for people to have a good time
most games today are one on one with the screen
even online, not a social experience
people want a social interactive experience
games to increase the talk
can play games on touch screens, and also order food and drink
chuckecheese pizza is basically horrible
electronic cocktail makers
party table – 6 players console, classic games and new
6 people is unstable, most parties are 2 or 4
will need people to mix
uWink will launch these in the fall
then working on a high school
school is obsolete
need to destroy the classroom
edutainment will become a reality
everyone believes in innovation, until they see it
not really interested in in-home entertainment – as this doesn’t provide a social experience (???)
women want a different kind of game – simple, knowledge based, word games, stimulate conversation, gives insight into themselves or life
wine tasting game
****
Dean Kamen
DEKA – safe water for the developing world
business and goverment far more difficult that the technology
80% of disease caused by water, or water bourne pathogens
came from working on ways to make water clear enough for injection
1.1 billion lack safe water
correlation between bad water and lack of money
needs to be 50 times more efficient than distillation to be economic
need to get to 500w for a village of people – 100 people
1-2p a gallon
experts say there are ways of doing it cheaper, but each for specific problems – heavy metals, bacteria etc.
need a one-shot method – no need to decide what is wrong with the water to treat it
put any water in – jsut add water!
no consumables, just electricity
about $1000 a box
local entrepreneur model
******
Robert Langer
most paradigms in media materials come from doctors going into their house and finding things similar to properties of the body
artificial heart – flexible like ladies’ girdles, so tried polyether urethane
hard to change once this has been made
now relooking at materials for medical use
looking at polymer erosion – changing the way things dissolve
reviews make it difficult to change things
materials for minimally invasive surgery – strings that go into the body, and then change shape at body temperature water
making and growing any kind of human cells – currently FDA-approved for skin grafts
*******
Charles Liebert
uniqueness in biotechnology: the bottom up paradigm
dna, proteins go into cells, which go into big systems like humans
nanowires and nanoclusters go into device and circuits, to go into large integrated nanosystems – electronics, photonics or biology and medicine
limited set of building blocks
need to encode information during synthesis not by lithography – key differentiator to current engineering
nanophotonics – bottom up assembly of optically active building blocks – subphoton size building blocks
handheld systems for detecting anything biological
could be used to build a natural interface wtih the brain neurons
contact
email:
chris is at anti-mega.com
MSN:
chris_heathcote is at hotmail.com
IRC:
ChrisDodo
iChat/AIM:
antimega77