Building Invisible Interfaces
Mark Weiser Computer Science Lab Xerox PARC November 2, 1994
My goals for this talk
That you will come to know
the "ubicomp" perspective of new UI research some deep problems with VR, agents, GUIs, voice, magic, seamlessness, . . . some of our work at PARC some new systems challenges of the ubicomp perspective
That some of you will think deeply and then
choose to work from a ubicomp perspective, or attack the root ideas of ubicomp
[Link]
Weiser, November 1, 1994 10:30 am PST
[Link]
Weiser, November 1, 1994 10:30 am PST
PARC and Xerox
Xerox Palo Alto Research Center is 24 years old
one of four research centers, PARCs focus is "systems" PARCs research in every Xerox product, including toner
PARC atoms
EML
advanced materials
to
EIL CSL
hardware and software systems
culture
ISTL
humans and technology
PARC-generated product technology
laser printing, multiprocessors, quad-spot diodes, ...
SPL
social dimension
PARC-generated Xerox startup companies
synoptics, spectra-diode, parcplace, U.S. display consortium, ...
components and subsystems
laser diodes
LCD displays
networks architecture operating systems
linguistics graphics hypermedia
collaboration anthropology
[Link]
Weiser, November 1, 1994 10:30 am PST
[Link]
Weiser, November 1, 1994 10:30 am PST
Three revolutions in computing
1950, mainframe computing:
one computer used by many people
1975, PC computing:
one computer used by one person
10 9 10 8 10 10 6 10 5 10 4 10 3 10 2 10 1 10
7
10
2000, ubiquitous computing:
many computers used by one person they had better be nearly invisible
m
1940
r ainf
am
e
PC
1960 1970 1980
UB
1950
1990
2000
2010
machines sold per year
[Link]
Weiser, November 1, 1994 10:30 am PST
[Link]
Weiser, November 1, 1994 10:30 am PST
10 9 8 7 6 5 4 3 2 1 1920
Good technology is invisible
"invisible" technology stays out of the way of the task
like a good pencil stays out of the way of the writing like a good car stays out of the way of the driving
bad technology draws attention to itself, not the task
like a broken, or skipping, or dull pencil like a car that needs a tune-up
computers are mostly not invisible
m ram ainf e
PC
1940 1950 1960
UB
I
1980 1990
you can hardly fail to notice when you use a computer they dominate interaction with them
ubiquitous computing is about "invisible" computers
1930
1970
number of interesting research ideas
[Link]
Weiser, November 1, 1994 10:30 am PST
[Link]
Weiser, November 1, 1994 10:30 am PST
Examples of Invisible Technologies
Electricity
Electric motors hidden everywhere (20-30 per car) Wired and wirelessly ubiquitous
How to Build Invisible Technologies?
start from arts and humanties: Philosophy, Phenomenology,
Anthropology, Psychology, Post-Modernism, Sociology of Science, Feminist Criticism, Your own experience... This is the most important part of the talk. You may not get it on first hearing. Patience. When I am done youll know what is wrong with: creating an entertaining and dramatic user interface computers magically meeting our desires a computer idealized as an assistant virtual reality as the ultimate user interface
Literary Technology
Continuously surrounding us at many scales Used trivially and profoundly Literally visible, effectively invisible
[Link]
Weiser, November 1, 1994 10:30 am PST
[Link]
10
Weiser, November 1, 1994 10:30 am PST
Your Personal Experience:
the flow state
Remember the last time you spent several productive hours. It had some of the following characteristics:
- time passed unnoticed - you were unaware of your surroundings - it took you about half an hour to get engaged - consciously you focused on a goal - unconsciously you drew on tacit skills and knowledge - the situation was very rich with details and nuances that you unconsciously took into account The things you did not think about -- the tacit, the context, the world -- made you smart.
Philosophy of Polanyi: "The Tacit Dimension"
perception and thinking are always involve both "distal" and "proximal" elements
distal is the current surface of thought proximal is closer to us, and tacitly focuses on the distal
for instance:
distal: words coming out of my mouth proximal: my muscle and brain activity when speaking
the "tacit dimension" is easily overlooked, and more important to understand, than the more available distal
[Link]
11
Weiser, November 1, 1994 10:30 am PST
[Link]
12
Weiser, November 1, 1994 10:30 am PST
Post-Modernism (Toulmins Cosmopolis)
Modernisms origins in the 1600s
vision of human rationality as adequate to all things a particular historical response to chaos in European affairs
Example
Example: Taylorism vs. Empowerment
Taylorism is modernist ideal of measured time and motion Empowerment is the post-modernist letting be of a whole person in a complete situation for action.
Modernism the major Western world-view 1600-2000
fueled advances in all fields of human endeaver is breaking down as omnisciently adequate
A return to the whole person
rational analysis is only a small part of the successful person planning is only a small part of thinking and work
Post-Modernism
human nature is not fully explained by logical categories idealized logic has distracted us from real life and nature
[Link]
13
Weiser, November 1, 1994 10:30 am PST
[Link]
14
Weiser, November 1, 1994 10:30 am PST
Feminist Deconstructionism
Standpoint theory
full history of a person in a culture and world is their standpoint standpoints influence our perception of the world around us every "objective" view also has a standpoint
Example
Old business attitude of hiring women and minorites
its the law its the right thing
New business attitude of hiring women and minorites
diversity of perspective is valued different lives provide the most difference in perspective diverse workforce provides business advantage
Standpoints are enacted through peoples lives
standpoints are not reducible to a plan or analysis method standpoints accessed only via the individuals embodying them Example Consequence: Strong Objectivity (Harding) Physics is only weakly objective because it presumes knowledge to be independent of knower Strong objectivity recognizes that knowledge from any one standpoint is limited
(this does not mean you should start trusting businesspeople)
[Link]
15
Weiser, November 1, 1994 10:30 am PST
[Link]
16
Weiser, November 1, 1994 10:30 am PST
Examples: whats wrong with:
creating an entertaining and dramatic user interface --
What we tried to do: unique features of ubiquitous computing
start from social science insights
- radically reinvents technology to fit people - aims for true human effectiveness
draws attention to the machine itself; confuses entertainment with use computers magically meeting our desires -assumes everything important can be articulated; ignores tacit and standpoint a computer idealized as an assistant -idealizes the anthropomorphic ; insults human assistants; misplaces trust;
virtual reality as the ultimate user interface
avoid personal computer - make computers "invisible"
- no thing in the office humming on the desk - ease of use so effective you dont notice the computer
many, many "displays"
- including audio, visual, environmental - including electronic postit notes stuck to things
reduces the human being to just a bundle of senses ubicomp is the inverse of virtual reality
casual, low-intensity computer use
- displays for menus, for icons, for each window - displays for background attention
[Link]
Weiser, November 1, 1994 10:30 am PST
[Link]
17
Weiser, November 1, 1994 10:30 am PST
18
Two misleading concepts
Seamless
is a seamless building one in which you never notice as you move from place to place? making everything the same is easy; hard is letting everything be itself, with other things goal: seamful systems, with beautiful seams
Where is the talk going
So far we have discussed:
- Xerox PARC, and its context for this work - The foundations of new thinking about Invisible Technology, Post-Modernism, etc. - some goals of ubicomp design
Coming up:
- A walk through some new technology - A taxonomy and description of new research challenges
Interface
implies boundary, difference, transduction but the unit of design should be social people, in their environment, plus your device talking "interface" trivializes the problem to the boundary (but there is no better word, so I too say "user interface")
[Link]
19
Weiser, November 1, 1994 10:30 am PST
[Link]
20
Weiser, November 1, 1994 10:30 am PST
Ubiquitous Computing Phase I devices
three sizes of devices; fill everyday space
tabs -- tiny wireless handhelds - hundreds/person pads -- booksized X terminals - tens/person boards -- whiteboard computers - one/person each device is a doorway to networked information
videotape
mobile infrastructure to make them useful
capable of 100s of devices per person capable of self-configuration maintenance both radio and infrared, in-building only
mobility is nearer-term focus, prior to true ubiquity
[Link]
21
Weiser, November 1, 1994 10:30 am PST
[Link]
22
Weiser, November 1, 1994 10:30 am PST
our attempt at ethics
Demonstrate technical feasibilty of non-invasive design
enable individual control over privacy
Some dimensions of our research
applications toolkits real-time programming techniques operating system protocol cellular handoff bandwidth physical wireless layer wired connection physical device
Try to model ethical behavior
get human subjects committee approval for trying new technology
Communicate to public about dangers of ones work
grit teeth, cooperate with media "scare" stories dont leave it to others to say the negative
[Link]
23
Weiser, November 1, 1994 10:30 am PST
[Link]
24
Weiser, November 1, 1994 10:30 am PST
physical device
TAB 2x3x0.5" 2 week battery life 7 oz. 8051 processor 128k NVram I2C external bus resistive touch screen 128x64 mono display PAD 9x11x1" 4 hour battery life 5 pounds 683xx processor 4 MB ram PCMCIA card electro-pen sense 640x480 4 level display
Systems Research Problems
Research driven by new figures of merit
Bits/Sec/M3 Instructions/joule
Only two things are different in ubiquitous systems
traditionally variable things are now static traditionally static things are now variable
coming up: UI implications of new systems problems
the problem of bandwidth the problem of energy
[Link]
25
Weiser, November 1, 1994 10:30 am PST
[Link]
26
Weiser, November 1, 1994 10:30 am PST
The problem of bandwidth: Bits/sec/M3
On Earth Today: 1010 people, 1012 M2 land. Assume 1012 Ghz total radio spectrum, 1 bit/sec/hz. A - 100 bits/sec/person worldwide B - 1 terabit/sec/person within one meter
requires new radios: "nearfield"
4 meter range, 250k/bps bandwidth r6 power drop, excellent cell boundaries 5 Mhz carrier, excellent in-room penetration very simple construction low cost (< $10 transceiver and digital interface) maximize bits/sec/M3 use microwatts of power locate users
A wired
distance
wireless B bits/sec
[Link]
27
Weiser, November 1, 1994 10:30 am PST
[Link]
28
Weiser, November 1, 1994 10:30 am PST
bandwidth assumptions of wireless
old: bandwidth is constant for a given connection new: bandwidth varies widely during use
because different cells have different congestions because mobile-IP is hiding your real network topology because sometimes you plug in
The problem of energy
low energy computing is good
less disposing of batteries less power generation needed easier to keep using
low energy computing is hard
instructions/joule little better today than ten years ago display and backlight and wireless link use most power
UI research problem: how much to hide from the user
applications with feedback on current bandwidth? separate control panel for user monitoring of bandwidth? indirect feedback through performance?
possibilities:
design for better instructions/joule reinvent the display create "low power" UI
[Link]
29
Weiser, November 1, 1994 10:30 am PST
[Link]
30
Weiser, November 1, 1994 10:30 am PST
Joule measures energy: 216 joules in a D battery I/J measures computation per battery things that dont help I/J
turn machine off: I/J = 0. slow machine down: I/J unchanged.
Instructions per Joule
reinvent the "display"
nondisplay displays
break out of TV model -- instead concentrate on interacting with a human somehow blow the drapes, jingle, tingle, get warm, use subtle change
displays themselves
repaint very slowly? be microscopic? use non-symbols? new-symbols? sacrifice everything for contrast throw-away displays: 1 oz, 2mm thick, survive 4 ft. drop, 25 milliwatt
something that can help
slow down while lowering voltage do this continuously, millisecond by millisecond
UI research problems:
how to keep the user in the loop? "MPG" meter? black tailpipe smoke?
[Link]
31
Weiser, November 1, 1994 10:30 am PST
[Link]
32
Weiser, November 1, 1994 10:30 am PST
New User Interface?
keep doing something as long as possible pushback to user what is hard or easy "colored button" menus
color every button according to power-cost of pressing it
Some new directions in ubicomp UI
Ubivid
tens of video streams per person per office create small town community
Project Jupiter
on-line community (MUD=multi-user-dimension) virtual community "chat" rooms, with multimedia some virtual rooms are also real rooms
will users stand for this?
like greyed buttons like real world
how to predict?
static? learned? dynamic by system state?
Casual I/O
"dangling string" ethernet monitor
[Link]
33
Weiser, November 1, 1994 10:30 am PST
[Link]
34
Weiser, November 1, 1994 10:30 am PST
motor
Dangling String Interface
freely hangs from ceiling direct analog connection to ethernet 0.1 turn per packet spins madly when busy wiggles gently most of the time can be heard, peripherally
LambdaMoo and Jupiter
LambdaMoo: self-programmable online community
7,000 users, 5,000 user-built rooms, 50,000 active objects surface UI is that of textual games: "throw knife at dwarf" deep UI is people working/playing/living together
Jupiter: LambdaMoo + multimedia + real rooms
AstroVR: astronomers sharing data and tools in rooms Jupiter prototype: multicast video and audio with rooms glass TTY UI can be ubicomp
8 feet
gives body to something virtual is part of the environment, like a breeze
created by Natalie Jeremijenko
[Link]
35
Weiser, November 1, 1994 10:30 am PST
[Link]
36
Weiser, November 1, 1994 10:30 am PST
Conclusion
Long term: a new world view of what a computer can be. Short term: fun new science to do, many hard problems for more info: "The world is not a desktop." Marc Mark Weiser. ACM Interactions, V1N1. January 1994.
traditionally variable things are now static
Traditionally, CS research tries to time-travel the curves:
- CPU speed doubles every year. - memory cost halves every 1.5 years. - other: display quality, headmount tech, ...
Old research method: aimed at when X is Y
X cpu speed memory networks Y 1000 MIPS on a desktop practically free are a gigabit/sec
[Link] [Link] telnet://[Link]
[Link]
But in mobile infrastructure, it aint so
- For the next ten years, all improvements go into size - Same memory, cpu, and bandwidth as workstations today.
37
Weiser, November 1, 1994 10:30 am PST
[Link]
38
Weiser, November 1, 1994 10:30 am PST
traditionally static things are now variable
bandwidth -- TCP assumptions break CPU performance (temporally and spatially) -- timing end-to-end network characteristics -- end-to-end flows connectivity -- yes? no? maybe? output: screen bits/pixel, size, number, stability input: type (pens?, kbd-simulations?), number physical network: RS-232, cellphone, CDPD, ethernet, ATM physical location: nearest printer, file server, router physical security: identity, network access, services
[Link]
39
Weiser, November 1, 1994 10:30 am PST