Human-Machine Engagement
📧 Alexander O'Connor
Computers are not just at desks any more
## 1
The computer is disappearing. It's gone from the early days of multi-tonne adding
machine to being small enough to embed in a pair of wireless earphones. Along
the way they've become so pervasive as to be invisible. We stop recognising
them because they are ubiquitous.
## 2
The presence of a computer in every device has contributed to the notion that
software is eating the world: more and more tasks that were undertaken in the
real world with atoms are achieved by switching digital bits. They have become
easier to use and more powerful, but how far can we take it?
## 4
Xerox PARC is a landmark in the history of computing. PARC labs is where the
mouse-and-window concept came from, they were working on pad-type devices in
the early 90s, and they invented key networking technologies. Steve Jobs
famously visited their lab one day, and some months later the Apple Lisa,
the Mac's ancestor, was born.
1. The user’s attention to the technology must reside mainly in the periphery . This means that either the technology can easily shift between the center of attention and the periphery or that much of the information conveyed by the technology is present in the periphery rather than the center.
## 5
One of PARC's researchers was Mark Weiser, and he was the one to coin the
concepts of 'calm technology' and 'ubiquitous' computing. It was a compelling
vision: the computers as part of the fabric of the world needed to compete
differently for our attention, they needed to instill calm rather than provoke
alarm. I think the current model is far from that.
2. The technology increases a user’s use of his or her periphery. This creates a pleasant user experience by not overburdening the user with information .
## 6
I know it's a problem for me, I am constantly on twitter, facebook, email,
messaging services. My phone buzzes with alerts and more. I want this to be
better, I would like a calmer environment: I want to see machine intelligence
embedded and fit our needs. In short, it needs to be humane.
3. The technology relays a sense of familiarity to the user and allows awareness of the user’s surroundings in the past, present, and future.
## 7
We started by calling these types of systems 'intelligent', but that might have
been a bit of a reach. Then we started calling them adaptive, personalised, and
contextual. that made more sense, they respond to individuals based on the
situation. But
they probably increased the sense of drowning rather than decreasing it.
I want calm, I want to see humanity in my computing.
## 8
For my work, I've been dealing for a while with the questions of how to support
different kinds of users. Experts are different from novices, but expertise is
a many-sided thing. I often help family memebrs and friends fix their laptops,
and solve other problems with their devices. I often don't know what the answer
is, but I do know how to google. It's been shown that skill with the information
system can outperform domain knowledge.
How many ways can you read a sentence?
## 9
I apply my work to questions about __content__, especially text. There are rules
of thumb that guess about 4/5ths of human knowledge is embedded in what we call
unstructued information. I am interested in how to help experts get information
out of text, and how to get information into textual form.
## 10
I've looked a lot recently at an area called 'topic modelling'. It's been very
popular in the natural language, search, translation, history and even literary
areas. The problem is a simple one: computers are dumb, and they don't
understand language. What tricks can use use about what they are good at
counting mostly to help us gain insight?
For remember that in general we don't use language according to strict rules – it hasn't been taught us by means of strict rules, either. — Ludwig Wittgenstein
## 11
Topic modelling works on the principles of the distributional hypothesis:
words of a feather flock together.
You can get a computer to count how words arise together in a number of ways.
Different models give you different insights. This, for examples, shows how
you can learn underlying concepts such as gender from sufficiently large word
collections
What's in your toolkit?
## 12
I'm also interested in looking at how people use their tools, and manage their
knowledge.When you try and build software you're balancing the practical and
interesting on both sides. What can I do with a computer that's interesting to
me and useful for you? I still run into this problem, I use paper notes but have
not found a workable way to digitise them.
How do we search for people when their names and details change?
One example of this kind of interest is frame by google in these terms:
people think in terms of 'things' not 'strings'. String is a term of a
sequence of letters as represented in a computer program. The fact that we use
labels to name things means they often are confusing, collide, or fragment.
Think of all the different ways we can process a simple image.
## 15
I've worked in how historians store their knowledge about these things in
semantic storage. This can be rather complicated, but the value can be large
when those semantic stores are connected to each other, or to other databases.
It's part of the idea that you should be able to ask questions of the internet,
not merely request documents.
I'd like to work on more ways to embed machine intelligence into content. I'm
especially interested in how content remixing, authoring, and collaboration can
move towards a more humane, contextual way of doing things. How can we make
content that is appropriate for long-form Sunday afternoon reading as well as
a quick video update?
More importantly, how can we do it so that the content consumer can properly
inquire about where the content came from, what the basis for its claims are,
and how to share it for their own network? While I do think that citizens and
journalists and academics are all contributing increasingly in the same content
spaces, I don't think that means that every article should be 140 characters.
I work as part of a bigger team. The ADAPT Centre has a hundred researchers
ranging from linguists to video analysis experts. We've people building smart
search engines, tools for reducing the torrent of notifications, translating
ironic tweets, and answering non-factual questions automatically.
We work with industry, and we're keen to work with other academics. Half of our
funding comes from industry engagements, and we have an engineering team to
build proof of concept systems, and to support the commercial application of
research.
I like to talk about it being the sweet spot of collaboration. The idea is to
automate the things that are hard for humans but easy for computers, while
helping to find the things that are hard for computers, or hard for both.
Computers can help by speeding things up, by repeating things lots of times, and
more.
“It takes these very simple-minded instructions—‘Go fetch a number, add it to this number, put the result there, perceive if it’s greater than this other number’—but executes them at a rate of, let’s say, 1,000,000 per second. At 1,000,000 per second, the results appear to be magic.” —Steve Jobs
So come and have a chat, see if any of that is of interest, and perhaps we can
work together?
