(First appeared in Issue 23 of the CS4FN magazine “The women are (still) here”)
The stereotype of a computer scientist is someone who doesn’t understand people. For many, how people behave is exactly what they are experts in. Kavin Narasimhan is one. When a student at QMUL she studied how people move and form groups at parties, creating realistic computer models of what is going on.
We humans are very good at subtle behaviour, and do much of it without even realising it. One example is the way we stand when we form small groups to talk. We naturally adjust our positions and the way we face each other so we can see and hear clearly, while not making others feel uncomfortable by getting too close. The positions we take as we stand to talk are fairly universal. If we understand what is going on we can create computational models that behave the same way. Most previous models simulated the way we adjust positions as others arrive or leave by assuming everyone tries to both face, and keep the same distance from, the midpoint of the group. However, there is no evidence that that is what we actually do. There are several alternatives. Rather than pointing ourselves at some invisible centre point, we could be subconsciously maximising our view of the people around. We could be adjusting our positions and the direction we face based on the position only of the people next to us, or instead based on the positions of everyone in the group.
Kavin videoed real parties where lots of people formed small groups to find out more of the precise detail of how we position and reposition ourselves. This gave her a bird’s eye view of the positions people actually took. She also created simulations with virtual 2D characters that move around, forming groups then moving on to join other groups. This allowed her to try out different rules of how the characters behaved, and compare them to the real party situations.
She found that her alternate rules were more realistic than rules based on facing a central point. For example, the latter generates regular shapes like triangular and square formations, but the positions real humans take are less regular. They are better modelled by assuming people focus on getting the best view of others. The simulations showed that this was also a more accurate way to predict the sizes of groups that formed, how long they formed for, and how they were spread across the room. Kavin’s rules therefore appear to give a realistic way to describe how we form groups.
Being able to create models like this has all sorts of applications. It is useful for controlling the precise movement of avatars, whether in virtual worlds or teleconferencing. They can be used to control how computer-generated (CGI) characters in films behave, without needing to copy the movements from actors first. It can make the characters in computer games more realistic as they react to whatever movements the real people, and each other, make. In the future we are likely to interact more and more with robots in everyday life, and it will be important that they follow appropriate rules too, so as not to seem alien.
So you shouldn’t assume computer scientists don’t understand people. Many understand them far better than the average person. That is how they are able to create avatars, robots and CGI characters that behave exactly like real people. Virtual parties are set to be that little bit more realistic.
ChatGPT is an AI chatbot which can have a conversation with you using everyday (natural) language. It has access to the huge amounts of digital information found on the internet and its ability to use language to write good answers in response to questions has been tested and improved by giving it feedback. It learns, and refines its output.
You don’t need to programme it, you can just ask it a question (giving it a prompt) and, after a moment’s ‘thinking’ it will produce an answer. Lots of people have been testing ChatGPT by asking it questions that we already know the answer to, to see how well it can do. It’s actually quite impressive! If you think it’s made a mistake or misunderstood your prompt you can reword your question or give it more information to help it improve its answer.
Can ChatGPT write an article that’s suitable for our magazine or website?
I gave ChatGPT the following prompt: “Please write a 200 to 500 word article suitable for the CS4FN magazine” and moments later it generated a completely new article about machine learning. That told me it knew (or at least could find out) that CS4FN has something to do with computer science. I’m not surprised that it picked machine learning as a topic given that it is a machine that can learn. I didn’t tell it what to write about, it came up with that idea itself.
The article it gave me contained some technical words and not all of them were explained, so I asked it to rewrite the article for a younger audience (who might not have come across some of the words yet).
The result is below – what do you think?
2. ChatGPT wrote this…
From Logic to Learning: The Incredible Advancements of Artificial Intelligence
Artificial Intelligence (AI) has come a long way since it was first invented in the 1950s. At the beginning, AI was based on logic and rules. Experts in a specific field would give the computer a set of instructions to help it make decisions. But, these rules were limited and couldn’t handle complex situations.
Today, AI has seen a huge shift towards machine learning. This means that instead of following rules, AI learns from data. This shift happened because computers got stronger, and more data became available. The breakthroughs in machine learning algorithms helped too.
Machine learning has two categories: supervised learning and unsupervised learning. In supervised learning, the computer learns from labeled data. This means that the correct answer is given to the computer for each input. Once the computer has learned enough, it can predict the answer for new data. Supervised learning is useful for applications like image and speech recognition.
In unsupervised learning, the computer learns from unlabeled data. This means that there is no correct answer provided. The goal of unsupervised learning is to find patterns in the data. This type of learning is useful in applications like anomaly detection and clustering.
Deep learning is a special type of machine learning that uses neural networks. Neural networks are inspired by the human brain and have layers of connected nodes. Each node applies a mathematical function to its inputs and sends the result to the next layer of nodes. Deep learning has revolutionized AI, leading to breakthroughs in areas like natural language processing, computer vision, and game playing.
Even though machine learning has achieved so much, there are still challenges to overcome. One challenge is the lack of transparency in machine learning models. Unlike rule-based systems, it can be hard to understand how a machine learning model made a decision. This can be a problem, especially in areas like healthcare where the decision-making process needs to be trustworthy and explainable.
Another challenge is the need for a lot of data to train machine learning models. While data is easy to find in some areas, like social media, it can be hard to come by in other areas like medical imaging. This lack of data can make machine learning less effective in some applications.
In conclusion, AI has come a long way since the 1950s. We’ve gone from rule-based systems to machine learning, which has led to breakthroughs in many areas. Although there are still challenges to overcome, the future of AI is exciting!
The first attempt is below…
The Evolution of Artificial Intelligence: From Rule-Based to Machine Learning
Artificial Intelligence (AI) has come a long way since its inception in the 1950s. Initially, AI was developed based on rule-based systems, where experts in a specific domain would provide rules to the system to help it make decisions. While these systems worked well in some areas, they were limited in their ability to handle complex problems and adapt to new situations.
In recent years, AI has seen a significant shift towards machine learning, where systems learn from data rather than relying on pre-defined rules. This shift has been driven by advances in computer hardware, the availability of large amounts of data, and breakthroughs in machine learning algorithms.
Machine learning has two main categories: supervised learning and unsupervised learning. In supervised learning, the system is trained on labeled data, where the correct answer is provided for each input. The system uses this labeled data to learn a function that maps inputs to outputs. Once trained, the system can make predictions on new, unseen data. This approach has been successful in a wide range of applications, such as image and speech recognition.
In unsupervised learning, the system is trained on unlabeled data, where no correct answer is provided. The goal of unsupervised learning is to discover patterns and structure in the data. This approach has been useful in areas such as anomaly detection and clustering.
Deep learning is a subset of machine learning that uses neural networks to learn from data. Neural networks are inspired by the structure of the human brain and consist of layers of interconnected nodes. Each node applies a mathematical function to its inputs and passes the result to the next layer of nodes. Deep learning has revolutionized the field of AI, enabling breakthroughs in areas such as natural language processing, computer vision, and game playing.
Despite the many successes of machine learning, there are still limitations and challenges to overcome. One challenge is the lack of interpretability of machine learning models. Unlike rule-based systems, it can be difficult to understand how a machine learning model arrived at a particular decision. This lack of transparency can be problematic in areas such as healthcare, where the decision-making process needs to be explainable and trustworthy.
Another challenge is the need for large amounts of data to train machine learning models. While data is abundant in some domains, such as social media, it can be scarce in others, such as medical imaging. This scarcity of data can limit the effectiveness of machine learning approaches in certain applications.
In conclusion, the evolution of AI from rule-based systems to machine learning has enabled breakthroughs in many areas and holds tremendous potential for the future. However, there are still challenges to overcome, particularly in the areas of interpretability and data scarcity. Addressing these challenges will be critical to realizing the full potential of AI in the years to come.
When we watch a film, it’s not just the pictures that make the experience, it’s the soundtrack too. The music and sound effects play a big part in setting the mood of a film. They matter. If you are to get the sinking feeling in your stomach or feel the shivers down your spine, it’s probably the music. QMUL’s Antonella Mazzoni wondered if other senses could contribute too … and designed a Mood Glove to find out.
Vibrations
We use touch as well as sight and sound to sense the world. This kind of ‘haptic feedback’ is used, for example, in phones that vibrate to tell us someone is calling. Antonella wondered if haptic feedback could heighten our mood while watching films in the way sounds do. To test her ideas she created a series of gloves. They had simple electronics built in to them that caused small pads to vibrate against the hand. She could control the order they vibrated and also the strength and frequency of the vibration. Early experiments showed it was best to make the pads vibrate on the back of the hand: when placed on the palm they tended to tickle too much. She also found that the positions of the vibrations did not make a big difference to moods, so she placed them in a simple circle.
Moods
Our moods and emotion can be broken into two parts: our levels of ‘arousal’ and of ‘valence’. Arousal is to do with the intensity of the mood. Being angry, delighted, alarmed and excited are all high arousal moods, whereas being bored, tired, sleepy and calm are low arousal ones. Valence is instead about the level of pleasure involved. High valence moods are pleasant and include being delighted, happy or calm, whereas low valence moods are unpleasant, such as being afraid, annoyed, depressed or bored. Together they give a standard way to rate mood.
Antonella next collected lots of film clips for use in her experiments. A series of volunteers watched the clips while wearing the glove and rated the experience in terms of their arousal and valence while watching them. Using these ratings as a baseline, she then ran experiments to explore if, and how, different kinds of vibration in the glove changed the wearer’s mood while watching the clips.
Suspense
In one experiment, she investigated suspense. Suspense is where the audience knows something about the plot that the characters don’t, leading to a gradual buildup of tension or expectation. Suspense can be linked to both positive and negative feelings so is not specifically about valence. It involves gradually increasing arousal. It is something that the score of a film can make a big difference to: transforming a clip with little suspense to one full of it. Antonella wondered if our sense of touch, through her Mood Glove, could deliver a similar enhancement? Perhaps, for example, a gradually building pattern of vibration on our hand could increase the build-up of arousal and so suspense. To find out, she chose 60 film clips that involved suspense. Volunteers rated them in terms of valence and arousal, and she used the 16 with most agreement. These final choices included clips from films like Inception, North by Northwest and Gravity.
Effects
Volunteers experienced heightened levels of suspense
Next she designed some simple effects to test. In her ‘buildup’ effect there was a gradual increase of both the strength and frequency of the vibration. The ‘fade in’ effect just increased the strength of the vibrations, starting from nothing and building to a peak. She also created an illusion that the effect moved across the hand, using the different vibration pads. A new set of volunteers watched the chosen film clips while wearing the glove. It gave different vibration patterns in time to each film. They rated their mood while watching the clips and Antonella also interviewed them about the experience afterwards. She found that the volunteers did experience heightened levels of suspense from certain kinds of vibration patterns for some clips. What worked differed for different clips suggesting a need to design the effect to fit the film.
Jobs
New technology creates new jobs that didn’t previously exist. You can see this in the ever increasing lengths of the credits of films, as new kinds of special effects lead to new jobs. Perhaps in future there will be a new career to follow as a `haptic composer’ for films, just as there are currently jobs composing soundtracks.
Originally Published in the CS4FN “Women are Here” special
Rebecca Stewart CC BY 2.0 Thomas Bonte
Some (female) computer scientists and electronic engineers were asked what they most liked about their job and the subject. Each quote is given with the job they had at the time of the quote. many have moved on or upwards since.
Here is what the real Pros think …
“Building software that protects billions of people from online abuse … I find it tremendously rewarding…Every code change I make is a puzzle: exciting to solve and exhilarating to crack; I love doing this all day, every day.”
Despoina Magka, Software engineer, Facebook
“Taking on new challenges and overcoming my limitations with every program I write, every bug I fix, and every application I create. It has and continues to inspire me to grow, both professionally and personally.”
Kavin Narasimhan, Researcher, University of Surrey
“Because computer science skills are useful in nearly every part of our lives, I get to work with biologists, mathematicians, artists, designers, educators and lately a whole colony of naked mole-rats! I love the diversity.”
Julie Freeman, artist and PhD student, QMUL
“The flexibility of working from any place at any time. It offers many opportunities to collaborate with, and learn from, brilliant people from all over the world.”
Greta Yorsh, Lecturer QMUL, former software engineer, ARM.
Shauna, Kavin and Greta
“Possibilities! When you try to do something that seems crazy or impossible and it works, it opens up new possibilities… I enjoy being surrounded by creative people.”
Justyna Petke, Researcher, UCL
“That we get to study the deep characteristics of the human mind and yet we are so close to advances in technology and get to use them in our research.” – Mehrnoosh Sadrzadeh, Senior Lecturer, QMUL
“I get the opportunity to understand what both business people and technologists are thinking about, their ideas and their priorities and I have the opportunity to bring these ideas to fruition. I feel very special being able to do this! I also like that it is a creative subject – elegant coding can be so beautiful!”
Jill Hamilton, Vice President, Morgan Stanley
“You never know what research area the solution to your problem will come from, so every conversation is valuable.”
Vanessa Pope, PhD student, QMUL
“I get to ask questions about people, and set about answering them in an empirical way. computer science can lead you in a variety of unexpected directions”
Shauna Concannon, Researcher, QMUL
“It is fascinating to be able to provide simpler solutions to challenging requirements faced by the business.”
Emanuela Lins, Vice President, Morgan Stanley
“I think the best thing is how you can apply it to so many different topics. If you are interested in biology, music, literature, sport or just about anything else you can think of, then there’s a problem that you can tackle using computer science or electronic engineering…I like writing code, but I enjoy making things even more.”
Becky Stewart, Lecturer, QMUL
“… you get to be both a thinker and a creator. You get to think logically and mathematically, be creative in the way you write and design systems and you can be artistic in the way you display things to users. …you’re always learning something new.”
Yasaman Sepanj, Associate, Morgan Stanley
“Creating the initial ideas, forming the game, making the story… Being part of the creative process and having a hands on approach“,
Nana Louise Nielsen, Senior Game Designer, Sumo Digital
“Working with customers to solve their problems. The best feeling in the world is when you leave … knowing you’ve just made a huge difference.“
Hannah Parker, IT Consultant, IBM
“It changes so often… I am not always sure what the day will be like“
Madleina Scheidegger, Software Engineer, Google.
“I enjoy being able to work from home“
Megan Beynon, Software Engineer, IBM
“I love to see our plans come together with another service going live and the first positive user feedback coming in“
Kerstin Kleese van Dam, Head of Data Management, CCLRC
“…a good experienced team around me focused on delivering results”
Anita King, Senior Project Manager, Metropolitan Police Service
“I get to work with literally every single department in the organisation.”
Pick up any computer or smart gadget and you’ll find small, colourful pictures on the screen. These ‘icons’ tell you which app is which. You can just touch them or click on them to open the app. It’s quick and easy, but it wasn’t always like that.
Up until the 1980s if you wanted to run a program you had to type a written command to tell the device what to do. This made things slow and hard. You had to remember all the different commands to type. It meant that only people who felt quite confident with computers were able to play with them.
Computer scientists wanted everyone to be able to join in (they wanted to sell more computers too!) so they developed a visual, picture-based way of letting people tell their computers what to do, instead of typing in commands. It’s called a ‘Graphical User Interface’ or GUI.
An artist, Susan Kare, was asked to design some very simple pictures – icons – that would make using computers easier. If people wanted to delete a file they would click on an icon with her drawing of a little dustbin. If people wanted to edit a letter they were writing they could click on the icon showing a pair of scissors to cut out a bit of text. She originally designed them on squared paper, with each square representing a pixel on the screen. Over the years the pictures have become more sophisticated (and sometimes more confusing) but in the early days they were both simple and clear thanks to Susan’s skill.
Try our pixel puzzles which use the same idea. Then invent your own icons or pixel puzzles. Can you come up with your own easily recognizable pictures using as few lines as possible?
Jean Bartik (born Betty Jean Jennings) was one of six women who programmed “ENIAC” (the Electronic Numerical Integrator and Computer), one of the earliest electronic programmable computers. The work she and her colleagues did in the 1940s had a huge impact on computer science however their contribution went largely unrecognised for 40 years.
Jean Bartik – born 27 December 1924; died on this day, 23 March 2011
Born in Missouri USA in December 1924 to a family of teachers in Betty (as she was then known) showed promise in Mathematics, graduating from her high school in the summer of 1941 aged 16 with the highest marks in maths ever seen at her school. She began her degree in Maths and English at her local teachers’ college (which is now Northwest Missouri State University) but everything changed dramatically a few months in when the US became involved in the Second World War. The men (teachers and students) were called up for war service leaving a dwindling department and her studies were paused, resuming only in 1943 when retired professors were brought in to teach; she graduated in January 1945, the only person in her year to graduate in Maths.
Although her family encouraged her to become a local maths teacher she decided to seek more distant adventures. The University of Pennsylvania in Philadelphia (~1,000 miles away) had put out a call for people with maths skills to help with the war effort, she applied and was accepted. Along with over 80 other women she was employed to calculate, using advanced maths including differential calculus equations, accurate trajectories of bullets and bombs (ballistics) for the military. She and her colleagues were ‘human computers’ (people who did calculations before the word meant what it does today) creating range tables, columns of information that told the US army where they should point their guns to be sure of hitting their targets. This was complex work that had to take account of weather conditions as well as more obvious things like distance and size of the gun barrel.
Even with 80-100 women working on every possible combination of gun size and angle it still took over a week to generate one data table so the US Army was obviously keen to speed things up as much as possible. They had previously given funding in 1943 to John Mauchly (a physicist) and John Presper Eckert (an electrical engineer) to build a programmable electronic calculator – ENIAC – which would automate the calculations and give them a huge speed advantage. By 1945 the enormous new machine, which took up a room (as computers tended to do in those days) consisted of several thousand vacuum tubes, weighed 30 tonnes and was held together with several million soldered joints. It would be programmed with punched cards with holes punched at different positions in each card allowing a current to pass (or not pass, if no hole present) through a particular set of cables connected through a plugboard (like old-fashioned telephone exchanges).
From the now 100 women working as human computers in the department six were selected to become the machine’s operators – a role that was exceptional. There were no manuals available and ‘programming’, as we know it today, didn’t yet exist – it was much more physical. Not only did the ‘ENIAC six’ have to correctly wire each cable they had to fully understand the machine’s underlying blueprints and electronic circuits to make it work as expected. Repairs could involve crawling into the machine to fix a broken wire or vacuum tube.
Two of the ENIAC programmers, are preparing the computer for Demonstration Day in February 1946. “U.S. Army Photo” from the archives of the ARL Technical Library. Left: Betty Jennings (later Bartik), right: Frances Bilas (Spence) – via Wikipedia.
World War 2 actually ended in September 1945 before ENIAC was brought into full service, but being programmable (which meant rewiring the cables) it would soon be put to other uses. Jean really enjoyed her time working on ENIAC and said later that she’d “never since been in as exciting an environment. We knew we were pushing back frontiers” but she was working at a time when men’s jobs and achievements were given more credit than women’s.
In February 1946 ENIAC was unveiled to the press with its (male) inventors demonstrating its impressive calculating speeds and how much time could be saved compared with people performing the calculations with mechanical desk calculators. While Jean and some of the other women were in attendance (and appear in press photographs of the time) the women were not introduced, their work wasn’t celebrated, they were not always correctly identified in the photographs and were even not invited to the celebratory dinner after the event – as Jean said in a later interview (see the second video (YouTube) below) “We were sort of horrified!”.
In December 1946 she married William Bartik (an engineer) and over the next few years was instrumental in the programming and development of other early computers. She also taught others how to program them (an early computer science teacher!). She often worked with her husband too, following him to different cities for work. However her husband took on a new role in 1951 and the company’s policy was that wives were not allowed to work in the same place. Frustrated, Jean left computing for a while and also took a career break to raise her family.
In the late 1960s she returned to the field of computer science and for several years she blended her background in Maths and English, writing technical reports on the newer ‘minicomputers’ (still quite large compared to modern computers but you could fit more of them in a room). However the company she worked for was sold off and she was made redundant in 1985 at the age of 60. She couldn’t find another job in the industry which she put down to age discrimination and she spent her remaining career working in real estate (selling property or land). She died, aged 86 on 23 March 2011.
Jean’s contribution to computer science remained largely unknown to the wider world until 1986 when Kathy Kleinman (an author, law professor and programmer) decided to find out who the women in these photographs were and rediscovered the pioneering work of the ENIAC six.
Vimeo trailer for Kathy Kleinman’s book and documentary
YouTube video from the Computer History Museum
The ENIAC six women were Kathleen McNulty Mauchly Antonelli, Jean Jennings Bartik, Frances (Betty) Snyder Holberton, Marlyn Wescoff Meltzer, Frances Bilas Spence, and Ruth Lichterman Teitelbaum.
Could your smartphone automatically tell you what species of bird is singing outside your window? If so how?
Mobile phones contain microphones to pick up your voice. That means they should be able to pick up the sound of birds singing too, right? And maybe even decide which bird is which?
Smartphone apps exist that promise to do just this. They record a sound, analyse it, and tell you which species of bird they think it is most likely to be. But a smartphone doesn’t have the sophisticated brain that we have, evolved over millions of years to understand the world around us. A smartphone has to be programmed by someone to do everything it does. So if you had to program an app to recognise bird sounds, how would you do it? There are two very different ways computer scientists have devised to do this kind of decision making and they are used by researchers for all sorts of applications from diagnosing medical problems to recognising suspicious behaviour in CCTV images. Both ways are used by phone apps to recognise bird song that you can already buy.
If you ask a birdwatcher how to identify a blackbird’s sound, they will tell you specific rules. “It’s high-pitched, not low-pitched.” “It lasts a few seconds and then there’s a silent gap before it does it again.” “It’s twittery and complex, not just a single note.” So if we wrote down all those rules in a recipe for the machine to follow, each rule a little program that could say “Yes, I’m true for that sound”, an app combining them could decide when a sound matches all the rules and when it doesn’t.
Young blackbird in Oxfordshire, from WikipediaThe sound of a European blackbird (Turdus merula) singing merrily in Finland, from Wikipedia (song 1).
This is called an ‘expert system’ approach. One difficulty is that it can take a lot of time and effort to actually write down enough rules for enough birds: there are hundreds of bird species in the UK alone! Each would need lots of rules to be hand crafted. It also needs lots of input from bird experts to get the rules exactly right. Even then it’s not always possible for people to put into words what makes a sound special. Could you write down exactly what makes you recognise your friends’ voices, and what makes them different from everyone else’s? Probably not! However, this approach can be good because you know exactly what reasons the computer is using when it makes decisions.
This is very different from the other approach which is…
Show it lots of examples
A lot of modern systems use the idea of ‘machine learning’, which means that instead of writing rules down, we create a system that can somehow ‘learn’ what the correct answer should be. We just give it lots of different examples to learn from, telling it what each one is. Once it has seen enough examples to get it right often enough, we let it loose on things we don’t know in advance. This approach is inspired by how the brain works. We know that brains are good at learning, so why not do what they do!
One difficulty with this is that you can’t always be sure how the machine comes up with its decisions. Often the software is a ‘black box’ that gives you an answer but doesn’t tell you what justifies that answer. Is it really listening to the same aspects of the sound as we do? How would we know?
On the other hand, perhaps that’s the great thing about this approach: a computer might be able to give you the right answer without you having to tell it exactly how to do that!
It means we don’t need to write down a ‘recipe’ for every sound we want to detect. If it can learn from examples, and get the answer right when it hears new examples, isn’t that all we need?
Which way is best?
There are hundreds of bird species that you might hear in the UK alone, and many more in tropical countries. Human experts take many years to learn which sound means which bird. It’s a difficult thing to do!
So which approach should your smartphone use if you want it to help identify birds around you? You can find phone apps that use one approach or another. It’s very hard to measure exactly which approach is best, because the conditions change so much. Which one works best when there’s noisy traffic in the background? Which one works best when lots of birds sing together? Which one works best if the bird is singing in a different ‘dialect’ from the examples we used when we created the system?
One way to answer the question is to provide phone apps to people and to see which apps they find most useful. So companies and researchers are creating apps using the ways they hope will work best. The market may well then make the decision. How would you decide?
This article was originally published on the CS4FN website and can also be found on pages 10 and 11 of Issue 21 of the CS4FN magazine ‘Computing sounds wild’. You can download a free PDF copy of the magazine (below), or any of our other free material at our downloads site.
Queen Mary researcher, Milan Verma used his AI program that modelled the way human brains see the world (our ‘vision system’) to change the details of some pictures in places where the program predicted changes should be easy to spot. Other pictures were changed in places where the AI predicted we would struggle to see even when big areas were changed.
The images flash back and forth between the original and the changed version. How quickly can you see the difference between the two versions.
Spot the Difference: Challenge 1
As this image flashes, something changes. This one is predicted by our AI to be quite hard to see. How quickly can you see the difference between the two versions?
Challenge 1 – can you see which part of the image is visible or obscured?
Once you spot it (or give up) see both the answer (linked below) and the ‘saliency maps’ which show where the model predicts where your attention will be drawn to and away from.
You can also try our second challenge that is predicted to be easier.
Spot the Difference: Challenge 2
As this image flashes, something changes. This one is predicted by our AI to be easier to see. How quickly can you see the difference between the two versions?
This article is inspired by a keynote talk Wendy Hall gave at the ITiCSE conference in Madrid, 2008.
What inspires researchers to dedicate their lives to study one area? In the case of computer scientist Dame Wendy Hall it was a TV programme called Hyperland starring former Dr Who Tom Baker and writer Douglas Adams of Hitchhiker’s Guide to the Galaxy fame that inspired her to become one of the most influential researchers of her area.
A pioneer and visionary in the area of web science, many of Dame Wendy’s ideas have started to appear in the next generation web: the ‘great web that is yet to come’ (as Douglas Adams might put it), otherwise known as the semantic web. She has stacked up a whole bunch of accolades for her work. She is a Professor at the University of Southampton, a former President of the British Computer Society and now the first non-US President of the most influential body in computer science, the Association for Computing Machinery. She is also a Fellow of the Royal Academy of Engineering and this year she topped it all and gaining her most impressive sounding title for sure by being made a Dame Commander of the British Empire.
So how did that TV programme set her going?
Douglas Adams and Tom Baker acted out a vision of the future, a vision of how TV was going to change. At the time the web didn’t exist and TV was just something you sat in front of and passively watched. The future they imagined was interactive TV. TV that was personal. TV that did more than just entertain but served all your information needs.
In the programme Douglas Adams was watching TV, vegetating in front of it…and then Tom Baker appeared on Douglas’s screen. He started asking him questions…and then he stepped out of the TV screen. He introduced himself as a software agent, someone who had all the information ever put into digital format at his fingertips. More than that he was Douglas’s personal agent. He would use that information to answer any questions Douglas had. Not just to bring back documents (Google-style) that had something to do with the question and leave you to work out what to do with it all, but actually answer the question. He was an agent that was servant and friend, an agent whose character could even be changed to fit his master’s mood.
Wendy was inspired…so inspired that she decided she was going to make that improbable vision a reality. Reality hasn’t quite caught up yet, but she is getting there.
Most people who think about it at all believe that Tim Berners-Lee invented the idea of the web and of hypertext, the links that connect web pages together. He was the one that kick-started it into being a global reality, making it happen, but actually lots of people had been working in research labs round the world on the same ideas for years before, Wendy included, with her Microcosm hypermedia system. Tim’s version of hypermedia – interactive information – was a simple version, one simple enough to get the idea off the ground. Its time is coming to an end now though.
What is coming next? The semantic web: and it will be much more powerful. It is a version of the web much closer to that TV program, a version where the web’s data is not just linked to other data but where words, images, pictures, videos are all tagged with meaning: tags that the software agents of the future can use to understand.
The structure is now there for it to happen. What is needed is for people to start to use it, to write their web pages that way, to actually make it everyday reality. Then the web programmers will be able to start innovating with new ideas, new applications that use it, and the web scientists like Wendy will be able to study it: to work out what works for people, what doesn’t and why.
Then maybe it’s your turn to be inspired and drive the next leap forward.
You may not think of computers as argumentative, but some of them do actually bicker quite a lot, and for good reason. Many hi-tech systems we depend on rely on the right computers getting their way, and the problems involved are surprisingly tricky to get right. Barbara Liskov’s contributions to this fiendishly difficult problem of making sure the right computers win their arguments helped her scoop the world’s top computing prize in 2009.
The ACM Turing Award, which Barbara won, is the computing equivalent of a Nobel Prize. She was awarded it for more than 40 years’ work. Her early work on programming languages has been used in every significant new language that has been invented since. More recently she has moved on to problems about ‘distributed systems’: computer systems involving lots of separate computers that have to work together. That’s where the arguing comes in.
Barbara has been working on an area of distributed computing called ‘Byzantine fault tolerance’. It’s all about providing a good service despite the fact that just about anything can go wrong to the computers or the network connecting them. It’s so important because those computers could be doing anything from running an ecommerce site over the Internet to keeping an airliner in the air.
Happy birthday
Here’s a thought experiment to show you how tricky distributed computing problems can be. Alice is a cellist in an orchestra and it turns out that the birthday of the conductor, Cassie, is on the day they are playing a concert. Jo, the conductor’s partner, has asked Alice to arrange for the orchestra to play Happy Birthday mid-concert as a surprise. The rest of the orchestra were enthusiastic. The only trouble is they didn’t have the chance to agree which break to do it in. In fact, no one’s sure they’re definitely doing it at all, and now they are now all sitting in their places ready to start.
For it to work they all have to start precisely together or it will just sound horrible. If anyone starts on their own they will just look and sound silly. Can it be done, or should they all just forget the whole thing?
Alice decides to go ahead. She can tell the others it’s on by whispering to those next to her and telling them to pass the message on. As long as her message gets to enough people across the different sections it will be ok. Won’t it?
Actually no.
The problem is: how will she know enough people did get the message? It has to be passed when people aren’t playing, so some could presumably not get it in time. How will she know? If the whispers didn’t get through and she starts to play, she will be the embarrassed one.
Are you in?
That seems an easy thing to solve though – when each person gets the message they just send one back saying, “I’m in”. If she gets enough back, she knows it’s on, doesn’t she? Ahh! There the problem lies. She knows, but no one else can be sure she knows. If any are in doubt they won’t play and it will still go horribly wrong. How does she let everyone know that enough people are willing to go for it? Alice is in the same situation she was at the start! She doesn’t know it will happen for sure and neither does anyone else.
She can start whispering messages again saying that enough people have agreed but that won’t help in the end either. How does she know all the new messages get through?
Change the problem
A computer scientist might have a solution for Alice – change the problem. Following this advice, she starts by whispering to everyone that she will stand up and conduct at an appointed time. Are they in? Now all she needs to be sure of is that when she stands up, enough people have agreed to play so that she won’t look silly. The others send back their message saying ‘I’m in’, but no one else needs to know in advance whether the song is definitely going ahead. If she doesn’t stand up they won’t play. If she does, they go ahead.
Delivering a good service
General knowledge It’s called Byzantine Fault Tolerance after some imaginary generals in the ancient days of the Eastern Roman Empire, whose capital was Byzantium. The original problem was about how two generals could know to attack a city at the same time.
Byzantine fault tolerance is about designing this kind of system: one that involves lots of ‘agents’ (people or computers) that have to come to an agreement about what they know and will do. The aim is for them to work together to deliver a service. That service might be for an orchestra to play Happy Birthday, but is more likely to be something like taking airline bookings over the Internet, or even deciding on action to take to keep the airliner they are flying in the air. The separate computers have to agree as a group even when some could stop working, make mistakes due to software bugs or even behave maliciously due to a virus at any point. Can a way be engineered that allows the system as a whole to still behave correctly and deliver that service? This is the problem Barbara Liskov has been working on with Miguel Castro at MIT. Of course they weren’t thinking of birthdays and orchestras. They were interested in providing the service of looking after documents so they can be accessed anytime, anywhere. A simple computer does this with a file system. It keeps track of the names of your documents and where it has stored them in its memory. When you open a document it uses the records it has kept to go and fetch it from wherever it was put. With this kind of file system, though, if something goes wrong with your machine you could lose your documents forever.
Spread it around
A way to guard against this is to create a file system that distributes copies of each file to different computers around the Internet. When you make changes, those changes are also sent to all the other computers with copies. Then if the copy on one machine is corrupted, perhaps by a hacker or just by a disk crash, the file system as a whole can still give you the correct document. That is where the computers need to start arguing. When you ask for your document back how do all the computers with (possibly different) copies decide which is the correct, uncorrupted version? That sounds easy, but as the orchestra example shows, as soon as you create a situation where the different agents (whether human or computer) are distributed, and worse you can’t trust anyone or anything for sure, there are lots of subtle ways it can all go horribly wrong.
The way Barbara and Miguel’s solution to this problem works is similar to what Alice was doing. One computer acts as what is called the ‘primary’ (Alice played this role). It is where the request from the client (Jo) goes. The primary sends out a request to all the backup machines for the document, just like Alice’s whispered messages. All the backups reply with their copy of the document. As soon as more than some predetermined number come back with the same document, then that is the good copy.
Not so simple
Of course the detail of Barbara and Miguel’s method is a lot trickier than that. They’ve had to figure out how to cope if something goes wrong with the primary (Alice herself) to ensure that the client still gets their document. Their version also works without any synchronisation to make things happen in lockstep (suppose Alice is at the back so can’t stand up and conduct to keep things in time). There are lots of other details in Barbara and Miguel’s version too. Messages are timestamped, for example, so that the recipients can tell if a message is new or just a copy of an old one.
Practically perfect
The particularly special thing about Barbara and Miguel’s way of providing fault tolerance, though, is that it doesn’t take an excessively long time. Various people had come up with solutions before, but they were so slow no one could really use them. The new method is so fast it’s almost as if you weren’t bothering about fault tolerance at all. Better still, the fact that it doesn’t need synchronisation – no conducting – means it also works when the replicated services are on the Internet where the computers act independently and there is no way to keep them in lockstep.
Barbara’s work might never actually help with an orchestral surprise like Alice’s, However, because of it future computers will be in a better position to shrug off their own kind turning rogue due to hackers, cosmic rays or even just plain old breakdowns. Not a bad reason to have a byzantine argument.
cs4fn 