Pitch: Personalizing your music experience

Background – What does the media industry currently look like, how is music selected in public places?  

Among  the media industry, music is the second most profitable medium worldwide with a 45% digital revenue share (with Games taking the first place with 60%). Moreover, music is completely going digital as the physical CD shipments  in the US have decreased from 940 million in 1999 to 88 million in 2017.

Consumers are not only shifting to digital music, but every year the time consumers spend listening to digital music increases. In less than 10 years the time spent listening to digital music increased approximately in 6 billion hours, and the gap between music streaming and music downloads has also been widening,  giving an opportunity to music platforms to implement and improve products that better match / suggest music to consumers in real time.

Today music is selected in the following way:

  1. B2B: Stores choose the music they play by taking into consideration the season of the year (holidays for example) and the average demographic of its consumers (age, gender, location, among others).
  2. B2C: At individual parties or reunions, consumers choose the music based on the preferences of the owner of the playlist, device and/or venue. The music is selected based on the assumption that the preference of all the guests is the similar to the preference of the owner.

Opportunity – Where are there pain points in the current system?

There are 3 main pain points in the current way media is selected and played in public spaces:

  1. No feedback loop: Music and media is often selected and curated by an individual and consumed by a user group in a particular setting with a “best-guess” effort to match audience preferences.
  2. Impact of music on shopping habits: The current system is not optimized for the shopping habits of the individuals in the store but is based on broader, generalized rules of music selection such as aggressive rock for Sales or classical for perusing.
  3. Impact of music on bars/ clubs: Bars and nightclubs have disproportionate reliance on good music selection to attract and retain patrons, which is often repetitive or selected in line with staff wishes.

Proposed Solution – How can a AI address those pain-points?

We would hope to use the best practices of recommendation engines and network analysis to generate a group optimum playlist, optimizing on most satisfied customers rather than total social welfare. A playlist would auto-generate based on this predicted best possible sequence of songs and would be provided as an integrated option in traditional music players within the App Stores of Apple or Android. The pipeline steps are displayed below:

Commercial Promise

In terms of B2B the business model could be based on a subscription model application that companies use to gather information about the music preferences of the customers that visit their stores or that they have in their database. This information would bring value by improving the shopping experience, having the possibility to influence shopping habits and also for marketing purposes. This could be applicable to retail stores, bars and clubs. In terms of the B2C channel the technology could be used to improve the experience in private parties or reunions. In the B2C channel the business model would more based on partnerships with existing music streaming application to help them improve the service they offer.


Benefits: Immediate feedback loop in public places to create a more personalized experience for consumers, collection and analysis of data based on music choice effectiveness for various seasons/locations/events/demographics

Challenges: Issues identifying causation vs. correlation for level of music influence, potential negative public perception of using music to influence consumer feelings and actions, difficulty of introducing brand new music with little or no related data points

Potential Competition

The key players in the music streaming industry have the capability to develop a viable product, with music personalization being a key push in recent years. Spotify is the market leader in terms of its influential playlist selections.Rivals such as Pandora, Google Play Music and Apple Music are also investing in personalization of music services.

In the startup space, Muru is a company targeting the B2B space. It offers background music streaming services for venues, allowing the venue to create a music template that will keep playing the full duration that venue is open and the playlist will also evolve on the fly.










fAIrytales – A Pitch for an interactive children’s book App


Over one fifth of US households spending more than 25% of their income on their children, childcare and early child education hurts a lot of household budgets. Within this picture, one important piece is children’s books. The proper development of a child’s reading ability starts at day 1 and has incredible effects on that child’s later ability to succeed in life.

The US children’s book publishing industry is big business, to the tune of $166 million profit on $2.3 billion of revenue annually. Within this, ebooks make up 12% of the industry, or $244 million. Within this industry there are unaddressed problems. A study conducted earlier this year found that there are all too often biases in children’s books. Female characters are grossly underrepresented in children’s books. And when they are, they tend to be the sidekick.

AI of Solution

The way we could build this system would be to incorporate several different machine learning tools to generate and evaluate stories until a threshold of acceptability was reached. We would ingest a large volume of data, likely from open source books to begin with. From there, we use word2vec, sentiment analysis, and topic mapping to generate stories, and use guidelines from the latest research into child development regarding sentence structure and vocabulary to ensure simplicity and age-appropriateness. Using massive training sets, a RNN will use layered outputs that are then reused in the inputs. Another challenge of the model is the short-term memory. The algorithm cannot remember “long-term” and so an architecture being explored now is LSTM and GRU, using gates within the code. Adding additional layers of gates could help to find higher-level interactions but the more layers we choose, the more training data we need to avoid overfitting.

The long-term vision for this product would be an application that allows the user to feed it the reader’s demographic profile. Children become the protagonist in a story of their own creation. In addition, using technology similar to that used in real-time digital advertisement generation, we  include pictures to follow along with the story. We then can build partnerships with various children’s entertainment content providers..


In order to pilot this concept, we would first need to ingest a lot of data. Luckily, there is a wealth of children’s books available within Project Gutenberg. Less luckily, these are mostly written in the 19th century. That means that they contain outdated language and social ideas that are largely shunned in most modern children’s stories. However, for a proof of concept, this is acceptable.

Once we have data, we can create a genetic algorithm to generate simple sentences using the book data we ingested, word mapping algorithms, and a list of age appropriate vocabulary to create sentence. Evaluation includes: 1) Using sentiment analysis, how well does the sentiment of the story follows the sentiment of training stories. 2) using topic mapping and word2vec, the stories transition topics gradually and make logical sense. We use a genetic algorithm to optimize both the story arc and the story coherence thendisplaying it to a young reader.

Commercial Viability

Our app is commercially viable given our initial research and interviews with parents.Based on the feedback of data scientists from the hackathon, our research seems promising and is executable given the fact that children’s stories are simpler than complex stories or writing for adults. Further, we have created a survey for parents to demonstrate the commercial viability of our app.

One parent indicated they spend $500/year on books for their child. Another indicated his interest specifically because he’s from India and the books he buys in the U.S. are more focused on American characters with American names. He is interested because he would have customized stories with Indian demographics that his child does not get living in Chicago. One concern is if he has to input too many things it might become tedious to use our app and he might stop using it after a while. We should aim to provide an easy “default case” that can readily be used for his child.

We also analyzed our competition: Epic! Which is a digital library for kids with a library of over 25,000 books both for kids and for educators. The price is $7.99/month and it has over 44,000 reviews on the app store. While they have an early mover’s advantage they do not have the option to customize based on various factors such as gender, race, citizenship of the parents etc. which will be our unique selling point and provides us with an advantage. We also plan on having a freemium model with up to 3 free stories in a month which will be based on a basic initial input into the app. For anything more interactive or customized, we plan on charging $8/month (similar to our competition) to our users as an upgrade cost. Based on the fact that the children’s story book market in the U.S. is worth $2.3 billion we have a huge potential even if we target the users who only use e-books.


Link to survey sent out to parents:






Most large retailers advertise their products to potential consumers through catalogs, sometimes paper-based but increasingly digital. Augmented Reality (AR) technology has the potential to disrupt the $83 bn field of digital advertising [1], and is already starting to do so, especially through product visualization and placement. Product visualization is an extremely important tool for companies which have a large number of products that often cannot be displayed to customers in an efficient format, potentially leading to lost revenue and increased customer churn. AR technology can allow consumers to interact with products much more seamlessly in multiple ways, which include personalized recommendations and more intuitive search mechanisms. In addition, this is an excellent application of augmented imagination techniques since it relies on machines developing the crucial selection step that can allow them to prioritize creative choices based on criteria. [7]




ViSenze is a Singapore-based startup founded in 2012 that develops AI for use in visual e-commerce. They have a number of solutions built around the ecosystem of image recognition such as image search, similar product recommendation and automated product tagging. While many companies have enabled image-based search on their websites, ViSenze’s smart algorithms can combine image search with text-based modifications. For example, a user might see a dress that they like but want it as a long-sleeve, and by combining text with visual recognition on the ViSenze platform, they would be able to modify their search criteria. This allows creative solutions from various inputs.


The ViSenze technology stack consists of the consumer clicking images and sending it to the ViSenze API. The API analyzes what is in the image, tags and attributes products and then combines with other contextual, search and internal data to also show visually similar recommendations.

Fig. 1 – Potential Use Case of ViSenze technology[2]


Effectiveness & Commercial Promise-


The potential for a company like ViSenze to achieve profitable performance is significant. Per Statista, in 2017, 1.66 billion people worldwide purchased goods online, and global e-retail sales totaled US$2.3 trillion.[8] . The scale of online shopping and its continued growth, especially into mobile devices, ensures a need for companies to continue to emphasize consumer ease of product search, search result satisfaction, and speedy engagement to drive purchase completion.


Based on the promise that consumers can integrate images into their search, both independently and with text, ViSenze has managed to raise $14M, most recently through Series B funding round [4]. Most of their current customers are focused on the retail space, especially in clothing, jewellery, and interior design, since the feature works especially well in industries with large inventories and somewhat homogeneous products. In an experiment conducted by ViSenze, about 150 tech-savvy people were asked to do a keyword search for a garment they were shown. 96.6% of users were not able to find it; they were frustrated and gave up after 90 seconds.  The 3.4% who did find it took 4 – 6 minutes. Visual search managed to beat keywords search by delivering results 9 times quicker than that a keywords search takes. [3]


There are a number of competitors in the space, most focused on either furniture or fashion. Some companies include Slyce and Whodat in addition to image recognition technology from major companies like Amazon and Google. However, while these companies possess similar technological abilities ViSenze is differentiated by focusing on directly adding value to retailers and shopping platforms through offering the function to be tailored to their products. The increased proliferation of companies into the space is based on increased demand for optimized search capabilities by customers.




We see a number of challenges and potential paths ahead for ViSenze:


  • Industry specialization: Most B2C startups in this space have focused on fashion and household products. ViSenze could try to specialize in one of the fields that they already have a foothold in, such as jewellery or other luxury products.
  • Getting retailers on platform: Larger companies like Amazon and Google already have a wealth of products on their platform and can monetize image recognition tech more easily, for example through more accurate recommendations. ViSenze should use machine learning technology trained on data across parameters both related to the product and the context in order to provide a superior recommendation tool.
  • Social media compatibility: Making ViSenze functional on top of social media platforms would be an effective means by which to gather large amounts of data and provide personalized recommendations.
  • Allow customers to co-create new designs with the help of AI (search: picture + Long sleeves, or “make it look 80ties”) If the wished product does not exist, and the customer is willing to pay extra, then have an automated factory produce (3d print?) and ship it to you.
  • Augmented reality: Let smart mirrors display the new outfit on you based on ViSenze selection.






2: https://www.visenze.com/visual-search


3: https://www.visenze.com/why-use-visual-search




5: https://www.crunchbase.com/organization/visenze


6: https://www.technavio.com/blog/top-12-image-recognition-software-companies


7: A Big Data Approach to Computational Creativity, IBM Watson Research Center, arXiv (Class reading)


8: https://www.statista.com/statistics/379046/worldwide-retail-e-commerce-sales/


Team Members:

Mohammed Alrabiah

Tuneer De

Mikhail Uvarov

Colin Ambler

Lindsay Hanson






Advanced materials, commonly segmented as ceramics, glasses, polymers, composites and metals & alloys, are one of the most important inputs in the design and engineering of industrially manufactured products ranging from semiconductors, consumer electronics, automobiles, and aerospace engineering. Advanced materials are essential both to the development of new products that were previously infeasible as well as improved performance of existing products  (e.g. inception of the Gorilla Glass used in Apple’s iPhone screens) and more cost-efficient manufacturing of the same. The advanced materials market was estimated to have annual revenues of $42.8 billion in the US in 2015, and is expected to grow to a size of $102.5 billion by 2024.


However, significant challenges remain in satisfying the growing demand for new industrial engineering applications. Many of the materials currently used – in their natural form – are already either very rare (e.g., platinum) or very costly to extract (e.g., aluminum and titanium). The result is an exceptionally high demand for developing new advanced materials in a variety of different ways. When combined with other elements to create new materials, such as “high-strength, lightweight alloys”, the process requires precision, as even a slight difference in composition can result in performance discrepancies. As such, manufacturing companies, such as 3M and Morgan Advanced Materials, spend on average ~5% of sales on R&D, often conducting “expensive and time-consuming, trial-and-error experiments to find the right combination of components and manufacturing processes”. Given how important breakthrough materials are in enabling many nascent technologies (e.g. solar panels, electric vehicle batteries, etc.) as well as those in which there are likely to be breakthroughs in the near future (e.g. nanotechnology), there is a need for a speedier and more cost-efficient way to discover new materials.


Many people within the field believe that it is likely that big breakthroughs are likely to happen in the near-future, pointing to initiatives such as the Materials Data Repository (MDR) and the Materials Government Initiative (MGI) that have opened up massive datasets containing information about advanced materials development that are only beginning to be explored. In spite of these initiatives, most materials science academic programs do not currently include a strong emphasis on computational materials science development methods, leaving a significant opportunity for groups willing and able to specialize in those methods.



Our solution, materiAI, leverages big data and AI to map out the chemical and physical traits of different materials and forecasts how they can combine with one another. For example, MateriAI can identify properties that allow certain materials to be better at conducting electricity and absorbing energy, or to be stronger but lighter, than others. By establishing patterns and understanding relationships like these, our solution can assist researchers and materials engineers in the discovery and creation of new materials. It can sift through all possible compositions and create “recipes” that optimize for qualities, such as durability, malleability, transparency, scratch-resistance and more. In addition to composition, the algorithm can also learn about the optimal processes (e.g. melting or beating metal, at what temperatures, etc.) for desired results.




Similar to IBM Chef Watson’s training set, we would parse white papers, research and the internet to understand how metals react, which combinations work and how properties are created based on the attributes of the combined materials using an algorithm developed at Google called Word2vec which looks at context in which word occur. Given that material scientists have used computational modeling for years (with varying degrees of success) to suggest properties for different applications, our solution would aggregate and harness this knowledge to make more accurate predictions. To prioritize, we would start by selecting a material, such as ceramics, which has a high potential for disrupting multiple industries. With a multitude of requirements for different types of properties based on the industry our technology would be extremely useful in tailoring the material based on the industry. Automobile manufacturers leverage the thermal and electric properties of ceramics and may not require it to be porous, substituting the porosity for ideally a strengthening component. This may not be the case when used in the medical equipment industry wherein its reaction to water and body fluids is the priority. Starting out with ceramics will provide us with valuable insights across a diverse set of industries giving feedback which can be incorporated when moving into new materials specific to an industry.



To ensure our algorithm works in creating optimal combinations for new material inputs, we would want to run tests on effectiveness and market adoption. For effectiveness, we would want to make sure that our “recipes” create mixes that are at least as good as the input they are trying to replace. This would mean running tests on durability, integration with other parts, speed and ease of us, and overall quality. In terms of market adoption, we would also want to test whether a) the algorithm attracts users and b) whether the “recipes” are utilized. This would require analysis of price (willingness to pay), number of customers who purchase, number of times algorithm is used, how many “recipes” are created, how often these “recipes” are used, and how often the resulting compound is used. These data statistics would also lend itself to a feedback loop so that the algorithm can continue to improve.



Currently, there are a few research labs at MIT and Stanford, as well as startups, such as Citrine Informatics that are taking an AI-driven approach to materials development. However, this so-called “materials informatics” approach represents a very small share of advanced materials research and development. Thus, in spite of these existing competitors, the size of the market opportunity should be large enough for a number of entrants to be successful, especially for those entrants that are able to be early-movers in the space.


Our startup will also differentiate itself from its competitors in two primary ways. First, materiAI will employ an anonymous data-sharing platform through which companies can elect to share historical research data in exchange for a reduction in price or as a means of accessing richer data from other companies that have opted in to the same. Second, Citrine Informatics has concentrated its early efforts on a handful of target verticals, namely aerospace engineering, chemicals, and consumer electronics. By focusing on different verticals such as the automotive, medical, and construction industries, we can capture an entirely different segment of the advanced materials market without significant overlap with our competitors.


















Siddhant Dube

Eileen Feng

Nathan Stornetta

Tiffany Ho

Christina Xiong



The newspaper industry represents a huge opportunity, with 2.7B print readers and 1.3B digital readers worldwide. In the U.S. alone, newspaper publishing was a $24B industry in 2018 and magazine and periodical publishing was a $28B industry in 2017, and this doesn’t even take into account news from other sources such as online-only publications. However, people have been increasingly concerned with bias in the news, with 72% of U.S. adults stating that news organizations “tend to favor one side” when presenting the news on political and social issues. Only 20% and 5% of U.S. adults trust information “a lot” from national news organizations and social media, respectively. The desire for unbiased news carries beyond the United States, as globally a median of 75% across 38 countries state it is “never acceptable” for a news organization to favor one political party over others.



To solve news bias issues, Knowhere uses machine learning and artificial intelligence to ultimately produce news stories written by machines. To do so, Knowhere’s algorithm selects topics based on what is most popular on the internet. The system then reads thousands of articles on the topic (from a combination of left wing and right wing sources) to gather data and writes an article on the topic based on what it finds. The technology strives to removes any bias, and presents information based solely on facts, while still reading like an article written by a journalist. Knowhere also weighs the accuracy of the sources it is using, making sure to weigh more accurate sources more heavily.


For political topics, Knowhere also produces “left” and “right” versions of each article so users can pick which they prefer, as well as compare different tones and viewpoints. Ultimately, two human editors review each article before it is published by Knowhere.


Challenges and Competition:

Most competitor companies (such as Google and Wikipedia) are focusing on mitigating the issue of fake news and increasing accuracy, while Knowhere is focused on the issue of removing bias. One of the challenges Knowhere may face is the fact that people often like to read biased information and fall privy to confirmation bias. They may not be open to trusting sources that offer perspectives different from theirs. Additionally, while Knowhere is taking strides to reduce bias, the company as a whole and its algorithms were created by humans, which means that there is inherently bias in the structure of the system.  Finally, Knowhere may face challenges in respect to making money. Most news sites make money through targeted advertising, but theoretically, Knowhere is looking to reach an audience through unbiased information, which makes targeting toward a specific group less useful.



Knowhere aggregates and filters news but does not create original journalism and relies completely on the work of others. It is currently not clear how Knowhere can best monetize their service. Objective news feed that targets all internet users will not generate individual user data that can be used for advertising targeting. Knowhere can offer paid consulting services that will help other news providers become more accurate or at least make them aware of how they stack in comparison with their competitors. That service may be of interest to chief editors who can benefit from understanding the biases of the journalists who work for them. As such, it can aid hiring decisions in the publishing industry. Finally, publishers can make the service available to their writers to increase their productivity.


Another challenge is to make the Knowhere content more widely available. To increase consumer awareness for their service, Knowhere can publish an index of accuracy for news sources. It can publicize the index through partnering with parties such as Wikipedia, Firefox, Google, or other interest and advocacy groups. Further, Knowhere should develop an app with the ability to set alerts for different topics. One of their areas of focus are complicated issues that are likely to have follow up coverage. The alerts can help retain the attention of users who liked a Knowhere article on the topic. Knowhere can add key word labels and a “Topics” section. Thus, a user who is newly interested in a topic, can easily access the universe of articles that Knowhere has already published on that topic. They can also make sharing easier and add buttons that allow users to email articles (the current options are only to share through Facebook and Twitter). Finally, they can add a “Local News” section that can draw in additional users.  




What’s in your closet? Sophia will tell you.


Consumers often do not know how to create the PERFECT outfit for an important event or they don’t know how to combine the articles of clothing they currently own into a good outfit.

Consumers may forget clothing items they already have hiding in their closet, so these items stay on the hanger underused.  Or, an individual may have 80% of an outfit that would fit their preferences, but they don’t know what to add to complete the ensemble or simply, they’re unsure what’s appropriate for a specific social event coming up on their calendar.


What’s in Your Closet, WIYC, would help these consumers better understand how to maximize the items already in their closet or find pieces to complete a look via relevant offline or online channels.

To use What’s in Your Closet, the customer will first need to add photos of their current closet inventory. This can either be an upload of the customer’s personal photos and/or transfer of items found from online transactions (e.g. receipt from Nordstrom, complete with item photos) to the customer’s account. Once their closet is ‘complete’ in their account, they can add color with style preferences so the algorithm can make the best recommendations in the future.

Then, the customer can come to their WIYC account when they are struggling to find an outfit.  They can input detail on a particular event or occasion (wedding, cocktail event, formal), and then the algorithm will take into consideration what you wore on recent occasions, seasonal and weather details, your budget, and your ideal fashion style.

What’s in Your Closet will then offer three suggestions based on your time constraints and preferences.

  1. 1-2 outfit recommendations based on what’s currently in your closet.
  2. 1-2 outfit recommendations based on what’s available locally in stores, or online, and which you can purchase in time for the event (coupled with items like shoes and accessories that are currently in your closet).
  3. 1-2 outfit recommendations based on items in your friends’ closests, if you elect to participate in social networking through the app and “share” your closet with others (coupled with items like shoes and accessories that are currently in your closet)

Pilot & the future

The solution will be piloted across a target audience of  young, fashion-minded women, potentially college students. This will be ideal as this audience will likely have less items to ‘catalogue’ into their account, so therefore can get up and running more quickly but at the same time have a high need for frequent outfits for social settings.

As the application is rolled out to a wider audience, the algorithm will be able to make even better recommendations as it receives more inputs on what outfits work or don’t work for particular types of customers. For example: by body type, by style, by location, by occasion.

In the future, ideally the application would be able to function as well as or better than a human stylist and personal shopper.  

Risks and Competition

Currently, the typical alternative to this type of product are people (fashion advisors, department store sales reps) that advise customers on their apparel when making a purchase.  Furthermore, online department stores such as Amazon will suggest outfits and articles of clothing based on previous clothing purchases. Because What’s in Your Closet caters to consumers who are looking to create ensembles based on what they currently own, closer competitors are websites and applications that suggest outfits based on what the user types in.  

In addition, there are some technology companies that play in a similar space, but do not tailor based on current items in ones’ closet. Stitch Fix, for example, deploys algorithms to predict which clothing attributes customers will prefer based on a created style profile and then deliver preferred items to the customer’s doorstep to try on.  Other services, like Wantable, also combine predictions based on a personal style profile or quiz.

The key risks to this application involve user engagement. This would require customers to ensure that their items are somehow catalogued into their profile so that the application can share which items look best together based on the event, day, etc.  However, once customers are fully engaged in the application and have items logged in, they will be less likely to leave.

Exit options

Some potential buyers for What’s In Your Closet include major departments stores and retailers like Nordstrom, Lord and Taylor, H&M, and others. This could also appeal to online-only brands such as Cuyana or Nasty Gal.


[1] http://www.vogue.co.uk/article/future-of-fashion-artificial-intelligence-post-material-world

[2] https://www.businessoffashion.com/articles/intelligence/top-industry-trends-2018-7-ai-gets-real

[3] https://www.stitchfix.com/

[4] http://fortune.com/2018/03/15/fashion-ai-artificial-intelligence-future-kim-kardashian/

[5] https://fashionista.com/2017/11/fashion-brands-stylists-ai-artificial-intelligence-chatbots

[6] https://www.businessoffashion.com/articles/opinion/how-fashion-should-and-shouldnt-embrace-artificial-intelligence

Team members:

Pavlina Plasilova, Kelly de Klerk, Yuxiao Zhang, Aziz Munir, Megan McDonald

Pitch: Crowdsourced AI Security


Crime has substantial societal costs: local government resources, medical bills, increased spending by companies and individuals on security, and depressed property values, to name a few. The global security market was valued at $70.02 billion in 2016 and is projected to grow substantially in the next eight years. Rising terrorism and mass shooting concerns in the US and abroad has led to a surge in the adoption of security systems. Additionally, technological proliferation, smart city infrastructure, and advanced analytics are transforming the industry.

The market is sizeable arguably because people have high willingness to pay to protect their loved ones and their property. Robberies, shootings, kidnappings, and even missing pets generate high levels of physical and emotional stress and significant costs for the victims and their communities. Effective use of analytics for crime detection and prevention is thus the key to saving and improving lives.


Solution and Feasibility:

Our proposal is to tackle security challenges by aggregating data from public and private cameras and sourcing image, video, and voice recordings from individuals. Currently people witnessing a crime are often willing to make recordings but there is no centralized platform for affected communities that can aggregate the incoming data. This causes delays in law enforcement reaction and some evidence likely never gets reported. A centralized crowdsourced data platform augmented with data from private surveillance cameras and public cameras such as ATM and supermarket surveillance can be effective in crisis situations such as location a runaway robber, terrorist, or missing people or pets. At later stages, the company will be able to identify high risk areas with limited coverage and augment the system capabilities by installing cameras or sound sensors there. Local governments or businesses interested in the area can subsidize the cost of the new sensors.

Further capabilities for crime prevention that the company will develop is to use AI and machine learning to identify faces of criminal suspects in the data streams. Machine learning can be used to set alerts based on detecting things like people with face masks, firearms, or unattended bags in crowded areas. Similarly, machine learning can be used to scan the data for the faces of missing people or images of runaway pets.

The number of security cameras in North America effectively doubled between 2014 and 2016. Along with an increase in mobile devices, the proliferation of drones and satellites will provide additional data for the platform.



Crime detection and prevention solutions will be of interest to local authorities. We plan to pilot the technology by partnering with local authorities who are interested in using AI to help improve security. For example, in New Orleans, Mayor Mitch Landrieu has proposed a $40 million crime-fighting surveillance plan, which will combine municipal cameras with the live feeds from private webcams operated by businesses and individuals. As such, we plan to team up with New Orleans and other cities to demonstrate our capabilities. Other areas for expansion include airports, crowded public spaces, and college campuses.



There are several companies which are using AI to help detect and predict crime. ShotSpotter, for instance, already uses sensor data to detect and alert authorities in the case of gun fire. ShotSpotter can provide information on the type of weapon and the likely location where it was fired with accuracy within 10 feet. The company had an IPO in 2017 and claims to have presence in over 90 cities in the US. Hikvision, a Chinese security camera producer, uses facial recognition to search for criminals or detect suspicious activity such as unattended bags. This technology is being used in the US in varying situations, such as by the Memphis police (for crime) or by the U.S. Army (to monitor a base). Many of the competitors are focused more on sensors and security cameras, which will have more power and impact if they can be utilized along with image, video, and voice recording from individuals. As such, our centralized crowdsourced data platform will provide a richer and more holistic view of the data. Further, many of our competitors are very specialized and niche. Over time, we plan to expand our capabilities to focus not just on crime but also on good Samaritan acts such as finding missing items or returning lost pets.



In the wrong hands, the technology can be used for the opposite purpose and actually aid crime. Criminals can identify when people leave their homes on vacation or target individuals. This will be mitigated by building in special access rights into the platform so that the general public can upload images and videos but special groups such as law enforcement officers can utilize the data.

Privacy concerns also have to be addressed. These concerns will be mitigated by ensuring transparency about the data sources and using data sources that the police can already get access to. Despite data privacy concerns, in crisis situations people are often very willing to assist authorities as much as they can to speed up the recovery of their loved ones.




Team Members:

Sam Steiny

Rosie Newman

Gergana Kostadinova

Javier Rodriguez

ComPulse – Swarm Decision Making for Localized Communities



Just like larger organizations, those that operate at a local level, such as the Chicago Department of Planning and Development, or the Residents’ Welfare Association, seek to use their budgets to effectively allocate taxpayer/stakeholder money towards the most pressing problems. Stakeholders are often skeptical towards the organization’s choices. In polls conducted by Gallup, about one in every three Americans does not have much confidence that local governments are adept at handling their local problems, and this number has remained roughly consistent since the 1970s. We also think that part of the problem might be attributed to voters or residents feeling isolated from the decision making and that current polls conducted in some places are limited and do not properly address the concerns of the affected stakeholders. By utilizing algorithms and swarm intelligence, a concept borrowed from the behavior of groups of bees or fish in nature, we can utilize the ‘wisdom of the crowds’ and allow stakeholders to collectively reach more efficient and effective budgeting solutions that will provide the most satisfied outcomes for each party.




We are proposing a swarm intelligence-based platform, which is defined as the collective behavior of decentralized and self-organized systems. Swarm intelligence based platforms have shown to make more accurate predictions as it provides the interfaces and algorithms to enable “human swarms” to converge online, combining the knowledge, wisdom, insights, and intuitions of diverse groups into a single emergent intelligence. (One of many examples is the prediction of the winners of the Kentucky derby.) ComPulse can expand the use cases to budgeting decisions, which are partly prediction problems (considerations of which areas require the most care) and partly judgment/opinion problems.

The technology stack consists of a user interface, commonly called the ‘design space,’ which would be intuitive in terms of the options represented. Users would make an initial choice and be able to view the rest of the swarm and their choice distribution. Another key aspect is an ‘egalitarian choice architecture,’ in order to minimize systemic bias in all aspects of the system such as question structure and sequence. ComPulse would be able to help organizations in screening and posting of questions that would be accessible to a large and diverse audience. The back-end of the stack will be powered by AI algorithms that are focused on optimization techniques and are part of a larger family of swarm-based collective decision making algorithms.


Fig. 1 – Illustrative representation of initial design space for candidate prediction (Source: Unanimous.ai)


Empirical Demonstration-


ComPulse would start by gathering the opinions of experts from different bodies or areas whom would collect points on their contributions, which could be utilized as discounts on posting fees when they initiate a budget proposal as an incentive for them, on higher level open-ended questions. These opinions would be aggregated and split into defined choices for each question or problem where the broader crowd who are relevant to these proposals (e.g local residents) would pick their choices through a thumbs up or thumbs down or alternatively make exclusive picks.

There are a few key parts using this technology:

  • Bringing together a diverse group of individuals such as administrators, residents, and development planners who are knowledgeable about the questions at hand.


  • Decreases bias, increases representation of various interests, entices engagement and allows for compromise.


  • For a specific question, each member makes an initial pick at roughly the same time and can see choices for the rest of the swarm (Refer to Fig. 1). Can decrease herd mentality which is often present during an ongoing poll.


  • Users interact within the design space through simple push-pull mechanisms in order to reach consensus on a certain question and are allowed to change their answers (Refer to Fig. 2).


  • In cases of no clear consensus, opportunity for further discussion and debate regarding problems.

Fig.2 – Illustrative representation of final consensus for candidate prediction (Source: Unanimous.ai)

So far, this technology has been used mostly in different applications from predicting sports and financial trends, to assessing the effectiveness of advertisements and movie trailers, real-time swarms have been shown to significantly amplify intelligence in each area. By combining this technology with a situation in which many diverse stakeholders are brought together to determine a solution it optimizes these solutions, reaches decisions and makes more accurate predictions on outcomes.




It is difficult to measure the tangible benefit from such improvements due to the difficulty of quantifying efficient decision making and increased voter trust. However, in order to test a potential local use case, we interviewed some members from the University of Chicago on their thoughts on the student government’s $2.3 mn annual budget that needs to be allocated amongst diverse functions and opinions to support the student life and organizations within the university. Citing a representative in the Student Government “This could be a good opportunity for us to reach out to the larger student body and make them feel adequately represented”

ComPulse could charge a posting fee to the organization for each set of questions as a result of facilitating the definition of the problem. Moreover, contracts with large organizations that have a number of local branches with different sets of problems (e.g., City governments) would be a way to reach a large number of stakeholders and become a key part of the decision-making process for a number of organizations.

ComPulse’s business model would consist of a for-profit side supporting its not-for-profit arm: profitability can be found in the B2B space, contracting within and between large businesses’ employee bases as they seek to maximize the efficiency of departmental or project-based thought processes. The side serving state and local governments would be primarily non-profit.


A platform that could be so central to augmenting decision making in organizations is not without a few risks. One that was pointed out during our interviews was pressure from incumbents who might not want to dilute the budgetary decision-making powers that they hold (Ivy Missen, UChicago Student Government). Another potential risk would be fears of the lack of adequate representation, especially for those without access to technology. ComPulse would work with organizations to ensure reduction of bias in the sampling process.




Gallup Government Trust Polls: http://news.gallup.com/poll/5392/trust-government.aspx


Unanimous Case Studies: https://unanimous.ai/case-studies/


Swarm Intelligence and Democracy: https://eanfar.org/can-swarm-intelligence-save-democracy/


Akay and Karaboga, Algorithms Simulating Bee Swarm Intelligence: https://www.researchgate.net/profile/Dervis_Karaboga/publication/220638051_Akay_B_A_Survey_Algorithms_Simulating_Bee_Swarm_Intelligence_Artificial_Intelligence_Review_31_68-85/links/5666a8c208ae4931cd627ba8/Akay-B-A-Survey-Algorithms-Simulating-Bee-Swarm-Intelligence-Artificial-Intelligence-Review-31-68-85.pdf


UChicago Student Government Budget: http://sg.uchicago.edu/budget/


Team Members:

Mohammed Alrabiah

Tuneer De

Mikhail Uvarov

Colin Ambler

Lindsay Hanson

DevGame: Crowdsourced, developer-selected video game content

Crowdsourced, developer-selected video game content


Companies who develop and publish video games require a continuous source of unique and compelling creative content in the form of new game concepts, new game franchises, new features, and expansions for existing games to remain competitive in the global video game industry.  Given that firm resources are, by nature, limited in scope and scale, the firm’s organic creative resources and development capabilities will always be unable to globally maximize the scope and scale of the firm’s creative ideation. However, simultaneously, there is some level of latent creative ideation resident within the “crowd” of the online gaming community, where a diverse set of creatively-inclined individuals may be interested in providing creative ideas for these purposes in return for some reward.  Yet this “crowd” has neither the programming, development, and distribution resources of a video game company, nor the resolve and inclination of a commercial entity – both of these elements are required to move creative ideas from raw ideation to commercialized content.

Significant value could be derived from a platform that, on one side, provided a low cost means for these companies to acquire curated, crowdsourced creative ideas, and on the other side, provided a potentially rewarding yet near zero cost means for individuals in the “crowd” to supply those creative ideas.


Create a online platform marketplace where players are able to upload ideas for new content.  More specifically, users would have free access to the platform and a list of games by title with the corresponding features on which to submit creative content (e.g. maps, worlds, characters, weapons, storylines, etc).  The user would submit a written proposal with supporting graphic designs; if the user required graphic design support, DevGame would use in-house graphic artists to help generate a minimal amount of design content for a low nominal fee.  Next, DevGame would validate proposals as authentic and complete to ensure quality control and curation. With a sufficient number of proposals, DevGame would make individual content available to game developers, as well as central creative themes from the aggregated ideas.  Game developers would retain the role of selecting the specific content, compensating the selected user for his/her content. Although the marketplace would be initially used for product extension (e.g. in-game add-ins), the marketplace would eventually be expanded to allow for product derivatives or entirely new games.  In the latter case, game developers could use individual submissions or mix and match components from different entries. DevGame would work with both users and developers to assist in naming and licensing requirements to protect generated content.

Design of Demonstration

The demonstration would have to accomplish the two goals of validating the quantity and quality of submissions in order to make the marketplace viable for game developers. In order to achieve this outcome, DevGame would select a popular game title that depends on constantly producing new content in order to maintain its user base (e.g. Fortnite) and list specific features on which users could submit proposals. After an initial trial period of collecting submissions, the game developer would select an idea for implementation. The game would then be updated using the user generated content. Then over a period of several weeks, statistics about player growth, retention and reviews of new content would be collected and then compared against the historic figures. If successful, the results would show that the crowdsourced content would be more in-tune with what users want and would lead to higher participation.

Pilot Program

The pilot program would begin by launching the marketplace across a few specific college campuses that are well-known for producing graduates headed to video game development. We would partner with a well-known developer to ensure that users had access to a viable product, with the first contest beginning with in-game add-ons for the selected title. After a set period of time, the contest would close and the winning submission made public to the online forum, generating the necessary user interest for follow-on contests.

Commercial Viability

The global video game industry is projected to reach $90B in sales by 2020. Activision Blizzard, one of the biggest game developers, had $4.7B in revenue in 2017. Thus, enabling a 1-10% increase in either market size (value creation) or individual company revenue (value capture) represents both an achievable and substantial outcome. As a brief example of the size of development costs, Rockstar Games spent $265 million on the development of Grand Theft Auto V alone and has released several expansions since in order to maintain revenue. Although numerous companies have attempted to crowd-select creative content (i.e. have the crowd decide the next feature of the game), no companies currently act as a marketplace for the aggregation of creative content. Moreover, crowdsourced and selected games, while numerous in quantity, have achieved fairly mundane results. Thus, DevGame would be a free service for the user while charging a fee to game developers to access crowdsourced creative content. Additionally, companies would provide incentives, either monetary or in-game rewards, to players whose content was selected and then developed.  Following the completion of an initial launch phase, DevGame could increase its value capture by implementing some scalable toll or fee on the value of content passing in both directions across the DevGame platform – from company to user, and from user to company.


Activision Blizzard Inc Form 10-K


Ubisoft Earnings


Global Video Games Market from 2011 to 2020


Crowdsourced Video Games are a Terrible Idea


5 of the Best Colleges for Gamers


Team Members

Thomas DeSouza, Matthew Nadherny, Patrick Rice, Samuel Spletzer

Viki: Global TV Powered by Fans


Since the 1990s when the Korean Wave (hallyu) spread throughout the world through today, viewership in Korean dramas has been steadily increasing. With a consumer base comprised mostly of millenials and women, viewers turn to these television shows for a number of reasons, from wanting to experience another culture to desiring to binge-watch stories, to simply wanting to experience a better plot than can be found in other shows. There are hundreds of Korean dramas today, and the quality and quantity of these shows is steadily increasing. Because of these show’s wide potential appeal, there are two primary opportunities: 1) a way of getting these shows onto the screens of everyone outside South Korea in a timely fashion and 2) a way to subtitle and localize these in a viewer’s preferred language.




Viki was founded in 2007 to provide internationals access to Korean dramas. Viki solves the problem of translation and localization by using crowdsourcing to subtitle shows. These community contributions are driven by intrinsic motivation, as users are not paid for their work. There are, however, certain recognition tiers for those who hit a threshold (Qualified Contributor status) granting subtitlers special perks, such as access to exclusive content. Viewers can track the completion of subtitles on a newly released show (unsubtitled shows are available on Viki a few hours after being shown in South Korea), and can either choose to watch the show before all the subtitles are complete, or wait until the shows are 100% subtitled in their language (a process that can take <24 hours from its premiere in South Korea). With close to 200 available language subtitles, Viki can provide any of its show to users in nearly any language, provided the community has contributed to produce its subtitles.

For each show, there is a designated team which includes language moderators who ensure the accuracy of subtitles. When the subtitles meet quality standards, the team leader (called the “Channel manager”) locks them to prevent further editing

Additionally, in 2017, Viki launched “Learn Mode” where viewers can watch shows with two sets of subtitles – their own language and that of another. In addition to helping users learn, this also improves the community’s ability to provide future subtitles. They also have the ability to pause the show at any time and highlight a word in the subtitle to learn pronunciation and spelling.


Effectiveness & Competitive Landscape

Significant market adoption (over 35 million monthly active users) has demonstrated Viki’s success in bringing new sources of content and providing suitable translations of that content. Academic research such as that done by researchers at Carnegie Mellon has likewise shown that the crowdsourcing model on which Viki relies for Active Crowd Translation (ACT) can approach the quality level of an expert in the field, giving its translation model further legitimacy.

In sizing the relevant addressable market for Viki, there are two industries of note: the market for translation services, and the market for television and movie entertainment. The total market size for translation services in the US is estimated at $5 billion per year. Furthermore, TV and Movie content production and distribution are estimated at revenues of $100 billion market per year in the US. Viki operates in both of these industries, and specializes in the intersection between the two, as well as how translation of entertainment content in the former opens up greater market opportunities for the latter. Because of the widely recognized wisdom that “content is king” and the race to find timely and unique streaming content first in a competitive market, that intersection has taken on ever-increasing importance.

In terms of revenue, Viki has three main streams: ads, viewer subscriptions, and licensed syndication of subtitled content to other distributors (e.g., Hulu, Yahoo, and Netflix). Viewers have the option of joining Viki for free, however this will force them to watch many ads throughout the course of a show or limit the number of episodes they can see.

Competitors in the television and movie streaming market, such as Netflix (which started streaming Korean dramas in 2015), Dramafever, and Hulu, have the ability to offer similar entertainment. However, their current Korean drama selection is mostly limited to older, “classic” series and they do not seem to have arrangements to onboard every new drama as quickly onto their platform. Meanwhile, competitors in the traditional translation and localization space lack the technology and distribution capabilities that Viki has.


Improvements & Suggestions

We see a number of opportunities for improvement within Viki’s product through a combination of machine learning, new business partnerships, and expansion into adjacent geographic markets and new content formats.

  • Use of machine learning to assist with crowdsourced translation. Viki can start off by building a simple translation system that can detect translation errors and propose suggestions to users based on language-specific rules. For instance, literal translations often ignore grammar, context and the meaning behind idiomatic expressions, such as “letting the cat out of the bag”. Given the number of companies, such as Google and Facebook, that are invested in neural machine translation, we believe that over time, Viki can partner with them to build a more complex solution that uses a combination of speech and image recognition to improve the accuracy of users’ translations. Such a partnership could be deeply beneficial to a technology company such as Google, because while Google has previously deeply indexed text content such as books, websites, and academic articles, its efforts in entertainment content have been more limited, and could help them compete against rival Amazon, who owns imdb.com. Such an arrangement could also further improve the quality and robustness of their data for machine translation and speech-to-text transcription. Additionally, machine learning could supplement easier, common subtitles (for example, the words “hello” or “watch out!”), reducing the need to translate these easier phrases. Greater reliance on assistive machine learning techniques could also be used to: (1) reduce the typical 1-day delay between when shows air and when subtitles are available on Viki and (2) encourage users who can speak and write in less popular subtitle languages to contribute.
  • Use a translate plugin: Both Microsoft & Google have simple speech to text converters with options to translate content into multiple languages. This could be used as a stop gap arrangement to immediately launch shows when they become available with more detailed subtitles coming in < 24 hours.
  • Expand to more shows and countries by partnering with other content providers. Viki can expand on existing partnerships with production companies to offer a wider selection of dramas. By extending, as well as signing on new, long-term contracts (potentially ones where Viki can co-produce dramas), Viki can increase customer retention and acquire new users.

  • Incentivizing multi-lingual members with more diverse rewards. Currently, volunteer translators are granted a free Viki subscription, badges, and some exclusive promotions. Viki can improve upon their rewards system by offering rewards on their partner platforms, such as Netflix, Samsung, and Fuji TV, and could even consider allowing Gold-status qualified contributors early access to content related to the shows they translate. This could further incentivize users to provide more and higher quality translations.
  • Expanding into adjacent markets for translating popular books, manga, and music. In the same way that television and movies require intensive effort to be translated and localized, books, manga, and music are similar forms of media that could likewise be translated through crowdsourcing. Because these are all forms of content that are already made available through subscription-based models (like Viki), the same model could readily be applied to them as well. A ready first candidate for partnership in this area could be Amazon, which already has millions of subscribers to its Kindle Unlimited service for books, and which already has an extensive presence in many parts of the world with unique content that could appeal to users in other countries.










http://clients1.ibisworld.com.proxy.uchicago.edu/reports/us/industry/keystatistics.aspx?entid=1245 http://clients1.ibisworld.com.proxy.uchicago.edu/reports/us/industry/keystatistics.aspx?entid=1246







Siddhant Dube

Eileen Feng

Nathan Stornetta

Tiffany Ho

Christina Xiong