April Science in the News: Our Friends, Our Foes, Our Machines

“What is the weather like?”

Any person you meet can likely answer this question without batting an eyelash. But what about a machine? Although seemingly simple, understanding and responding to such an inquiry requires a certain amount of intelligence. For instance, a refrigerator might not be the best thing to consult when deciding what to wear in the morning. But Emmett Sprecher began April’s Science in the News (SITN) presentation by asking his phone this very question. A few seconds later, Siri, the personality behind the iPhone 4 spit out New Haven’s weather forecast for the evening. How can your smartphone do what your refrigerator cannot? This was the topic discussed in the basement of the New Haven Public library: artificial intelligence (AI).

Siri, the iPhone’s AI voice recognition system can answer questions about the weather, compose texts for you, play music, and even search the Internet. 
Photo: Ankit Disa.

The three SITN presenters, Emmett, ThaiBinh Luong, and Christopher Bolen -- all recent or current graduates in Yale’s computational biology program -- used an array of futuristic and real-world examples of AI to discuss what AI actually is, how it works, and what the future of AI may hold. Here, I will try to break down their answers for you (spoiler alert: no, robots probably won’t take over the world… according to Christopher, at least).

The April SITN presenters answering questions following their talk at the New Haven Public Library. 

From left to right, Christopher Bolen, ThaiBinh Luong, and Emmett Sprecher. 

Photo: Ankit Disa

Before we can answer how AI operates or how it might affect us in the future, we must answer the first question, what is AI? In the true circular fashion of academia, many computer scientists define artificial intellgence as “the science and engineering of making intelligent machines”.¹Emmett clarified this definition by considering examples that we (whether or not we are aware of them) encounter on a regular basis.  Siri, our iPhone friend, is a recognizable example, perhaps, in part, due to its voice.  However, consider the computer opponent in video games, the Roomba (the robotic vacuum cleaner), and Watson (the robot that recently won a game of Jeopardy); these are all examples of machines that utilize artificial intelligence in some capacity.  As Emmett explained, they all have the ability to perceive their environments and take actions which they determine will lead to the best chance of success.  Success in this case is broadly defined; it could mean winning a video game or sucking up a crumb.  In contrast, non-intelligent machines, like assembly line robots, always take pre-conscribed actions and do not modify their behavior based on their interactions with their environment.

http://upload.wikimedia.org/wikipedia/commons/d/d3/Euclid_flowchart_1.png

http://chzsomuchpun.files.wordpress.com/2010/07/ce822881-9696-4207-8d74-3444a9f49f38.jpg  

                                        

Algorithm vs. Al Gore rhythm.
An algorithm is a set of rules to follow -- for instance, a flow chart (left). Al Gore rhythm (right) is probably something our former vice president doesn't have much of (just a guess!).

So, how does AI work? As ThaiBinh explained, the basic concept is not as complicated as one might think. Often, as in many video game systems, for instance, the decision-making process follows algorithms (not to be confused with any Al Gore rhythms) that are coded into the AI computer. You can think of an algorithm as a recipe or a procedure that tells the machine to take action B when it encounters situation A. The success of such an AI, however, depends on the quality of these rules. Instead of fixed rules, however, one could build an AI to find patterns in the incoming data and make predictions about future outcomes. This is how Netflix is able to make suggestions about movies you may like -- it compares what movies other people like who have the same interests as you.

AI can help us clean our carpets or figure out the weather, but how far can it go? Google’s self-driving car, which uses Google Street View information and visual sensors to navigate, has already logged over 200,000 miles, and driverless car licenses have been approved by the State of Nevada. In the medical field, in addition to the growth of robot-assisted surgeries, IBM has agreed to use Watson to help doctors make medical diagnoses.

Truly, it seems AI is and will continue to revolutionize our everyday lives for the better. But we are often presented with futuristic possibilities in movies and sci-fi books of robots taking over the world and revolting against their creators (us). Don’t worry, Christopher assures, this scenario is very unlikely. Why? Because AI is, in the end, a set of rules programmed in by a person, so if robots were to take over the world… we’d have to tell it to!

 http://images.wikia.com/terminator/images/1/19/Terminator_robot.jpg

Looks like we won't have to worry about a Terminator doomsday scenario ... unless we decide we want to!

- Ankit Disa
3rd year, Applied Physics PhD

Joining the disease detective club

In March’s Science in the News, Tiffany, Lindsey and Jon used a hypothetical example of some sick students at a school to discuss what outbreaks are, what can cause outbreaks and how we can protect ourselves. In real life, disease detectives – more formally known as epidemiologists – do many things to help detect, track and prevent disease outbreaks. Some work directly with patients or communities, collecting samples for analysis and helping administer health policies (such as vaccination campaigns). Others are based at organizations such as the Centers for Disease Control and Prevention (CDC) or pharmaceutical companies. They may work as administrators or work in labs.

   

John Snow, the "father of epidemiology" (not to be confused with Jon Snow from Game of Thrones!).

Sources: http://en.wikipedia.org/wiki/File:John_Snow.jpg

 http://sharetv.org/shows/game_of_thrones/cast/jon_snow

When did epidemiology get started? The “father of epidemiology” is often considered to be Dr. John Snow, who determined the cause of a cholera epidemic in London in 1854. Cholera is caused by the bacteria Vibrio cholera, which infects the small intestine, leading to watery diarrhea and vomiting. Without proper treatment, it is often fatal. Historically, cholera epidemics have been both frequent and devastating due to poor sanitation and they continue to occur in less developed parts of the world (in the wake of the 2010 earthquake, Haiti experienced a cholera outbreak that resulted in over four thousand deaths). During the 1854 cholera epidemic, Snow interviewed people to determine where cholera cases had occurred and found that many of them clustered around one water pump. He was further able to demonstrate that areas in the city that exclusively relied on this pump had the most cholera cases. His breakthrough was particularly impressive given that disease transmission was still poorly understood – most people, including most scientists, believed that people got sick through breathing “bad air.” Nobody knew that bacteria could cause diseases. Snow, perhaps unintentionally, set a precedent for evaluating future outbreaks although his belief that disease was not caused by the air was dismissed in the immediate aftermath of the 1854 epidemic.

This map shows how the cholera cases were clustered in the 1854 epidemic.

Source: http://en.wikipedia.org/wiki/File:Snow-cholera-map.jpg

If you’re interested in learning more about the history of epidemics, you might be interested in the following course “Epidemics in Western Society since 1600” (offered here, free, for either video or podcast downloading).

Given the broad range of activities epidemiologists carry out, how does one become an epidemiologist? Epidemiologists are often, but not always, doctors. Non-physicians generally have graduate degrees in epidemiology or public health. Yale has one of the oldest schools of public health in the country (founded in 1915). To find out more about epidemiology and public health research at Yale, check out the School of Public Health’s website.

Changing topics a little, I wanted to expand a bit on two types of tests that Jon mentioned can be used to detect norovirus and rotavirus. Viruses contain DNA or RNA cores protected by a protein shell. Viruses are super tiny and therefore epidemiologists need to use indirect methods to detect their presence. One test is polymerase chain reaction (commonly abbreviated as PCR). In this test, short pieces of DNA (called primers) are used to amplify a known viral gene. If viral DNA is present, the primers will stick to the gene of interest. By using a special enzyme, this gene can then be copied. This stick and copy reaction will be repeated many times. In each repeat there will be exponentially more copies of the gene available for amplification. In this way, you can generate a sufficient amount of DNA to detect and measure.

Source: Saheli Sadanand

Another test that can be used to indirectly assess the presence of virus particles is an ELISA, which stands for Enzyme-Linked Immunosorbent Assay. Although the name makes it sound complicated, it’s actually quite straightforward. In this assay, you are trying to detect antibodies, proteins that our immune system makes to bind to and get rid of pathogens such as viruses and bacteria. To do this, you basically construct a sandwich. In the norovirus and rotavirus example, the bottom layer of the sandwich are virus particles. You then load your sample of interest (for example, blood from the sick patients). Then you add a detecting reagent, which is another antibody – but specific for the non-virus-specific portion of the antibodies. Finally, you add a special chemical that reacts with the detecting reagent to produce a color. The brighter the color, the more virus-specific antibodies that are present in the sample. With the appropriate controls (for example, a sample that comes from a healthy person), epidemiologists can determine whether the amount of virus-specific antibody is abnormal – which would imply that the patient is in fact sick with the virus.

-Saheli Sadanand
5th year Immunobiology graduate student