Opportunity

Mammography, the first line of defense in detecting breast cancer, is a high-volume area of medical practice: roughly 39 million mammograms are conducted in the US alone annually.[1] Mammograms are performed using a low-dose x-ray machine that sends ionizing radiation through the body structure. The product is the black and white x-ray image with which most patients are familiar. While medical X-ray technology as a whole underwent digitization at the turn of the 21st century, the newly digitized visual output remained black and white, simply a high-resolution version of the familiar image on which thousands of medical practitioners had been trained. Thus the medical science of reading a diagnostic mammogram remained essentially unchanged.

 

We believe this digitization without transformation was a dramatic missed opportunity. First, while digitized diagnostic x-rays do capture a rich range of grayscale (anywhere from over 4,000 to nearly 66,000 shades, or 12-16 bits/pixel), their usefulness is constrained by the number of gray shades a computer monitor can display (most of which have caught up to current imaging machines, but which still remain a funnel point as diagnostic machinery improves).[2] Second, and more importantly, the human eye can only distinguish approximately 30 of these many thousands of shades of gray.[3] Thus, even the highest-trained radiologists have a naturally limited capacity to distinguish normal and abnormal structures within the breast when represented in grayscale.

 

A commonly used tool today is Computer-Aided Detection (CAD), which reviews the digital x-ray, searches it for abnormal areas of density, and displays a visual marker (such as a yellow triangle or red square) near these places for the radiologists to carefully visually review. CAD encourages more careful image review by the radiologist, and its use resulted in a 6% increase in early detection over non-CAD image reads.[4] Yet CAD does little to address the difficulty of reading the x-ray in the first place, particularly as relates to the 40% of women with dense breast tissue. It fails to apply the fullness of computational power to analyzing the x-ray.

 

Consumers (both patients and radiologists) have a clear interest in reducing the instances in which an x-ray because it provides insufficient information due to lack of visual clarity, necessitates additional diagnostic imaging. Both parties also have an interest in reducing false negatives (times when cancers are present but visually undetectable within the x-ray) as well as false positives (When normal structures or non-cancerous growths are detected yet cannot be distinguished from cancerous growths). All of these contribute to increased health costs, pressures on patient mental health, and burdens on time and talent within the medical practice. The challenge is even greater in countries like India, where only about 10,000 trained radiologists exist, woefully undeserving a nation of 1.2 billion.[5]

 

Solution

Our company, Radiotek, will apply a smart algorithm to grayscale x-rays, analyzing groupings of pixels that relate to one another in grayscale intensity to apply distinct color sets to the previously undetectable groupings. We would apply supervised machine learning techniques and utilize training data from thousands of patients. Along with the images/scans from these patients, we would also need the outcomes. The quality of our technology would vary depending on the data on outcomes that we can collect, ranging from biopsy results to a much simpler scale for categorization of outcomes.

 

Our concept hopes to make a leap forward in breast cancer detection by building beyond the static algorithmic approach to bring the power of deep neural networks to abnormality detection in medical imaging. We aim to utilize convolutional neural networks for image processing in order to allow for filter-based detection of interesting features in mammograms. The end result for the clinician is a conversion from grayscale to color in an analytical manner, not merely one-to-one, revealing structures and their relationships to one another in ways the grayscale previously obscured. [6]

Fig. 1 – A high level view of convolutional neural network architecture used in face detection

The end goal is to produce a tool that will support the radiologist in speedy decisions, and perhaps even provide preliminary diagnostics in cases of extremely high confidence. We would also need to ensure that the algorithm is capable of reinforcement learning, in order to get feedback, learn from mistakes and ultimately move towards goals set by the radiologist or hospital in order to provide the most efficient assistance.

 

Empirical Demonstration

 

There are a number of competitors in the US, among which one, called Imago Systems, has demonstrated early success with the static version of this algorithm. Their patent-pending Image Characterization Engine (or ICE) has already launched clinical trials and is currently in the FDA review process. Below is a depiction of the ICE at work on a breast cancer case: [7]

Fig. 2- Imago Solution Implementation7

 

With confidence in the potential of a static algorithm, we can begin building the deep neural network. With insights from experts in diagnostics and from our research of the Indian market, we believe that in order to penetrate the market we would need to partner up with large private hospitals or research institutions in Tier I or Tier II Indian cities. Due to a more robust infrastructure and their interest in specialization, we plan to readily obtain historical de-identified clinical data on patients confirmed with or without cancers, calcifications, and other abnormalities. A partnership with these organizations will also provide a natural entry point into the market for building proof of concept.

 

Another competitor, Zebra Medical Vision, has tested their machine learning based algorithm and has gotten 91% sensitivity and 80% specificity – as good a performance as any published results for radiologists working with or without computer-aided detection (CAD).[8] Additionally, artificial neural networks are already in use in other diagnostic areas.[9] Given their established presence in the medical arena, we believe the time to functionality should be reasonable. Upon achieving this, we envision giving the radiologist a front-end device such as an iPad which will run the Radiotek API. The API will be communicating with multiple AWS servers (which can respond to API, run model etc.). This will ensure low initial costs for us as well as a seamless experience for the user.   

 

Commercial promise

Both the Indian market for healthcare in general, as well as the market for radiology have been experiencing around 15-20% CAGR in recent years and are projected to continue at the same rate. The current market size of radiology is estimated at ~$1.2 bn, with high concentration in major metropolitan areas.[10] There are a number of unique benefits for Radiotek here:

 

• High growth sectors: Medical infrastructure market growing at 15%. Hospital services market currently valued at $80 billion and accounts for 71% of the industry revenues. Steady population growth and increasing insurance coverage. [11]

 

• Business-Friendly government: Low customs duty rates on medical service (9-15%) Government of India has permitted 100% FDI for all health-related services under the automatic route.

 

• First-mover advantage: Machine learning based diagnostics market not as competitive, and ability to capture unique market. In addition, our business model has a stronger customer stickiness element with higher switching costs for our customers.

Fig. 3- Indian radiology market geographic breakdown10

 

Ultimately our product would be sold under a license in which a medical center would pay for access to the algorithm, would house and store the image output within its own data centers (to conform with user privacy regulations), and would receive regular updates as the program continues to learn and improve. Users would pay the initial installation and licensing fee, then would pay monthly fees for both updates and volume usage. We would incentivize high volume usage to increase the volume of cancer detection data informing our system, and thus would provide tiered discounts to high volume users. Moreover, our product would initially be a diagnostic tool that would be validated by radiologists for every scan, but as we scale, we would allow for the tool to clear healthy patients with a high confidence level, therefore allowing other medical personnel such as nurses or technologists to read the scans and make decisions on non-complicated results. This would be truly disruptive in changing the landscape of the Indian radiology market and making it more accessible.

 

Financials

Lower wages in the Indian market will alleviate our labor expenses. We expect to hire two well-experienced data scientists, who would be able handle the development of the algorithm, a sales representative to be able to initiate engagements with potential customers and management at a total of labor wages of $150,000 in the first year. In addition, we expect to incur storage and computing costs through AWS EC2 at a total of ~$5500 in the first year, $20,000 for market research (industry dynamics and sales process in India), and $10,000 for  misc expenses (computers, travel etc.)

 

	Year 1	Year 2	Year 3
Revenue	$54,000	$282,000	$1,002,000
Operating costs	$156,000	$315,820	$365,700
SG&A	$14,500	$20,000	$25,000
Other costs	$29,500	$20,000	$30,000
Total Cost	$200,000	$355,000	$420,700
EBITDA	-$146,000	-$73,820	$581,300

 

If we are able to secure our ask of $200k we expect to be able to cover our expenses in the first year and be able to apply for a $100k loan at an interest rate of 8% that will be covering our expenses in the second year. We will then be able to produce sufficient sales beyond that with 3 license sales in our first year, 15 in our second year and 50 in our third year (break-even point) at a $12,000 installation fee and $500 usage fee per month for each license.

Fig. 4- five-year revenues & net income projections

 

Challenges

• • Regulatory challenges: Murky regulatory landscape with bodies such as Indian Medical Association and Ministry of Health and Family Welfare (MOHFW). Less stringent but also less clear cut than U.S. FDA.

• • Competition: International competitors like Imago and Deepmind as well as Indian home-grown startups such as Niramai. But different approaches/shortcomings like static algorithm, less focus on building neural network, and alternative approach to invasive radio scans, respectively.  

• • Data volume: Needing sufficient clinical data. Could work with de-identified historical data and relevant health information to provide full picture of cancer screening.

• • Data storage: Need to find secure way of storing patient data, need to be careful about sharing data between hospital networks. Upon talking to data scientists, suggested ‘secure enclave’ on hospital servers housing data within which neural network can be trained and then removed, leaving data inside.

 

 

Sources

 

1. JoNel Aleccia, High tech mammogram tool doesn’t boost detection, study shows (Seattle Times, September 28, 2015) https://www.seattletimes.com/seattle-news/health/high-tech-mammogram-tool-doesnt-boost-cancer-detection-study-shows/
2. Tom Kimpe & Tom Tuytschaever, Increasing the number of gray shades in medical display systems — how much is enough? (J Digit Imaging, 2007). https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3043920/
3. Francie Diep, Humans can only distinguish between about 30 shades of gray, (Popular Science Online, February 19, 2015) https://www.popsci.com/humans-can-only-distinguish-between-about-30-shades-gray?dom=tw&src=SOC
4. BreastCancer.org, Computer aided detection mammograms finds cancers earlier but increases risk of false positives. http://www.breastcancer.org/research-news/20130425-4
5. Dr. Arjun Kalyanpur, The Teleradiology Opportunity in India. https://www.entrepreneur.com/article/280262
6. Lawrence, Steve et. al., Face Recognition: A Convolutional Neural Network Approach. (IEEE Transactions on Neural Network, 1997) http://www.cs.cmu.edu/~bhiksha/courses/deeplearning/Fall.2016/pdfs/Lawrence_et_al.pdf
7. Imago Systems website, http://imagosystems.com/
8. Ridley, Erik. AI Algorithm uses color to better detect breast cancer, 2018.
9. Artificial neural networks in medical diagnosis, Journal of Applied Biomedicine, 2013. https://www.researchgate.net/profile/Eladia_Pena-Mendez/publication/250310836_Artificial_neural_networks_in_medical_diagnosis/links/5698d76608aea2d743771eef/Artificial-neural-networks-in-medical-diagnosis.pdf
10. Redseer Consulting, 2018: http://redseer.com/articles/indian-diagnostic-market-shifting-to-preventive-care/
11. U.S. Government Export Website, Indian healthcare Industry: https://www.export.gov/article?id=India-Healthcare-and-Medical-Equipment

 

Team Members:

Mohammed Alrabiah

Tuneer De

Mikhail Uvarov

Colin Ambler

Lindsay Hanson

Opportunity-

 

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]

 

Solution-

 

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.

 

Alterations-

 

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.

 

Sources-

 

1:http://www.adweek.com/digital/u-s-digital-advertising-will-make-83-billion-this-year-says-emarketer/

 

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

 

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

 

4:https://techcrunch.com/2016/09/15/singapores-visenze-raises-10-5m-to-bring-the-benefits-of-ai-to-e-commerce/

 

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

Opportunity-

 

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.

 

Solution-

 

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.

 

Pilot

 

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.

Risks:

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.

 

Sources:

 

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

Opportunity

The surveillance and security services industry in North America recorded revenues around $34 bn in 2017 and is constantly growing. A significant part of this comes from security services in private enterprises, especially for small and medium sized enterprises, where shoplifting is a large concern. For an average store for each item stolen they have to sell 50 more of the same product in order to make up for the loss.  Machine learning techniques and feedback loops can help human-machine ensembles solve this problem. Smart sensors can alert human surveillance centers, allowing organisations to devote less resources towards security. Moreover, analyzing behaviour using these mechanisms will allow security companies and businesses to reduce costs incurred through thefts.

 

Proposal

We will provide a minimal number of video cameras to each store. The cameras will be embedded with video recognition software that can process and analyze movements, postures and behavior patterns. Smart sensors will complement our cameras by being able to record ambient information such as the number of people in store and temperature.

 

This will in part be in areas where behavior is necessitated to be monitored. The cameras will be able to provide real-time alerts when suspicious behavior is observed, which can be integrated with a smartphone to be able to swiftly respond. The system will also be able to monitor safety related behaviors and send alerts to law enforcers. For example the system will immediately alert the police and send them footage if it detects an individual who is unlawfully carrying a weapon.

 

Another potential future use case that be leveraged in addition to the security feature would be to roll an autonomous retail store experience for customers visiting the store. To facilitate this service, retailers could create smart contracts that would authorize automated payments, discounts and enforce rules against returns or disputes depending on the customer’s attributes (e.g loyal customers gets rewards and more wiggle room etc..). This would eliminate the need for a human at the retail store and eliminate wait times at check-outs and dramatically mitigate a major felon amongst shoplifters — “sweethearting” — the collusion of a shoplifter and the checkout clerk.

 

Empirical Demonstration of Commercial value and promise

The cost savings that will occur from this product will be immense for businesses that regularly experience shoplifting or errors to their products. This would eliminate the need for a human or a conventional sensor whom would not have the same accuracy as our product in capturing the people with unpaid items leaving their store. In addition, this would make it harder for the employee to manipulate for his own self.

 

Furthermore, the product does not violate the privacy of the customers visiting the store as it is merely a surveillance camera that records the movement and behavior of the individuals in the store. This could expand to monitor other unwanted behaviors such as terrorism, reckless driving and any other behavior that is valuable to be reported to the business or organization to enable them to take action. The real value comes from the ability to report in real-time rather than the retrospective method used from a conventional camera’s past records especially for behaviors that need a swift response.

Perhaps the one added-value feature that would be revolutionary for law enforcers would be the ability to have superior facial recognition through these cameras. This would help law enforcers identify people in complicated situations where there is a large crowd and save them the time to repetitively watch the video for clues, saving them enormous resources that could be utilized to other areas.  

 

Challenges

We think that it might be difficult to convince small businesses to equip themselves with our ZeroChain cameras and sensors when they do not realize the cost savings incurred by our product. Moreover, they would need to be convinced that our product actually works better than conventional tools already available in the market.

 

Our mitigation for this case would be that we could charge these businesses our equipment at cost and only charge them a marginal subscription fee. We would then charge them each time our camera successfully detects an individual that could have potentially incurred losses to the retailer. This would make our customers feel comfortable knowing that our interests are aligned with theirs. Additionally, big retailers such as Target have accelerators focused on retail technologies, and others such as Walmart have shown interested in investing in an autonomous and seamless customer experience through acquisitions by Walmart Labs.

 

Sources:

http://www.wpri.com/target-12/local-stores-use-artificial-intelligence-to-catch-shoplifters/1136106605

 

https://www.fastcompany.com/3064771/can-this-ai-powered-security-camera-learn-to-spot-fishy-behavior-as-it-happ

 

https://www.technologyreview.com/s/608765/i-tried-shoplifting-in-a-store-without-cashiers-and-heres-what-happened/

 

https://venturebeat.com/2017/11/22/ai-could-make-video-surveillance-a-proactive-crime-fighting-tool/

 

Team Members:

Mohammed Alrabiah

Tuneer De

Mikhail Uvarov

Colin Ambler

Lindsay Hanson

New Opportunity:

International spending by Chinese tourists has skyrocketed in the past 8 years, showing increases of over 150% from $38B to $261B with the growth of the Chinese middle class.  The tourism market is currently very fragmented, as would be expected due to the number of countries and people involved, there is an opportunity to integrate. Traditionally, Chinese tourists travel with their families or through travel agencies that offer group tours. Due to the data capabilities and market share of major Chinese companies such as Tencent and Baidu, we believe there is a chance for them to quickly disrupt the tourism market by offering personalised recommendations and trip planners using prediction models as well as matching algorithms.  

 

AI Solution:

AI and machine learning techniques will be used to create a powerful trip planner and recommendation tool that is personalized to individuals. Building off the kind of matching techniques that are used by platforms such as say, Netflix, machine learning can play a large part in the recommendations section. The AI solutions will be powered by continuous real-time data through surveys, reviews and location tracking. Therefore, the application will identify travel-companions and restaurant recommendations by overlaying these techniques on top of traditional filtration tools specified by the users. The solution required must be able to combine different kinds of data types, for example users may have an image of a tourist spot with a comment about how much they loved it, therefore the company would need to apply image recognition techniques. This could be combined with multiple searches in the week prior to travelling on the Baidu search engine, for best restaurants in a certain district of a city. Thus, combining and profiling using such data would create an optimal route and itinerary for the traveller.

 

Design of Demonstration:

 

To understand the commercial promise, we must focus on the pain points of customers and see how the product addresses these. One of them is logistics and excessive paperwork to track all kinds of reservations, compounded for families. With the centralized dashboard (pictured below), these can be compiled in one place, and this can be done through automatically pulling data from e-mail and other communications. A proof-of-concept has already been offered by Google on its Trips app.

Another pain point can often be the recommendation system. Younger travellers often want to skip the touristy spots and go somewhere authentic. This is often not possible due to the language barrier, but by utilizing crowdsourced recommendations (already a part of these companies’ infrastructure) and filters, it will be easy to save time and money finding suitable places.

Pilot:

We have talked about many features, so a pilot would focus on rolling out some core features in a few cities. Say, we pick Paris, a popular tourist destination with a lot to see and a language barrier for most. We would beta test with young travellers going to Paris, and initially propose itineraries and recommendations based heavily on crowd-sourced data and location tracking. In this service, the main user is not the main source of revenue, and thus we would have to focus on user experience. As more users are attracted to the platform, there would be various ways to monetize, the most popular of which we foresee as being a B2B model, where businesses are charged for giving deals to travellers and advertising on the platform. Moreover, as travellers take multiple trips, machine learning techniques can be employed more heavily to solve prediction problems related to planning itineraries and suggesting establishments.

 

Sources:

https://www.emarketer.com/Article/WeChat-Users-China-Will-Surpass-490-Million-This-Year/1016125

10 Most Popular Social Media Sites in China (2018 Updated)

https://www.wsj.com/articles/internet-tightens-popular-chinese-wechat-app-to-become-official-id-1514541980

https://www.chinadaily.com.cn/specials/0711mafengwo.pdf

Google launches a personalized travel planner, Google Trips

 

Team Members:

Mohammed Alrabiah

Tuneer De

Mikhail Uvarov

Colin Ambler

Lindsay Hanson

 

Opportunity:

 

The Pennsylvania judicial system is one of many state prison systems that have been considering adopting statistically driven tools to determine how much prison time should be sentenced to individuals found guilty of committing crimes. The state spends $2 billion a year on its corrections system — more than 7 percent of the total state budget, up from less than 2 percent 30 years ago. Further, recidivism rates (the tendency of a convicted criminal to re-offend) remain high: 1 in 3 inmates is arrested again or re-incarcerated within a year of being released.  By properly identifying and distinguishing high / medium / low-risk offenders, the system has the opportunity to calibrate its sentencing accuracy to optimize the operations of its correctional facilities. In theory, risk assessment tools could lead to both less incarceration and less crime.

 

Solution:

 

The available risk assessment tools assign points to certain variables (such as age, gender, income, drug use, previous convictions, etc.) that have demonstrated to be strong indicators of criminal behavior in historical data. Social scientists have followed former prisoners and examined the facts of their life and monitored their lives for a number of years to develop an understanding of their propensity for repeated criminal activity. Many court systems use the tools to guide decisions about which prisoners to release on parole, for example, and risk assessments are becoming increasingly popular as a way to help set bail for inmates awaiting trial. This will ultimately help them save on their costs by providing better data-driven judgments towards criminal sentencing.

 

Commercial promise and challenges:

 

 

The main value proposition is that having an algorithm-based component to the judicial decision-making process helps many stakeholders through the value chain-

 

• Reduced risk of individual bias affecting judgments
• Increased efficiency reducing trial and bail time. Good for judges and defendants.
• Reduced costs which will inhibit better allocation of tax-payer money

 

While humans inherently rely on biased personal experience to guide their judgments, empirically grounded questions of predictive risk analysis play to the strengths of machine learning, automated reasoning and other forms of AI. One machine-learning policy simulation concluded that such programs could be used to cut crime up to 24.8 percent with no change in jailing rates or reduce jail populations by up to 42 percent with no increase in crime rates. Importantly, these gains can be made across the board, including for underrepresented groups like Hispanics and African-Americans.

 

On the other side, this approach faces challenges on an individual level especially because the system is based on a probability factored from similar offenders in the past that will influence the offender’s sentence despite that his future could be different or in other words an outlier to the statistics. To minimize these errors we will need to know whether the system will have enough variables to most accurately assess the individual as much as possible and whether these tools will supplement the judge’s decision rather than depend on it. 

 

Competition:

Even though a sizable amount of the agencies and organizations using AI systems in criminal justice reform are governmental bodies, the algorithms and software they use are privately owned. Due to the nascent nature of this industry, competition may either be among private companies trying to develop more efficient and fair algorithms, or there may be competition from an altogether different process, such as a community-based open-source AI project. A report from the Brookings Institute highlights the success of programs such as Google’s Tensorflow and Microsoft’s DMTK as proof.    

 

Proposed alteration:

 

The possible risk-assessment tools should be highly integrable with the existing software and processes in use in the justice system. While companies will want to utilize the ‘black box’ model that allows them to keep their algorithms confidential, it may lead to legal challenges such as in the ‘Loomis v. Wisconsin’ case (Wired). Thus, we would emphasize an open-source based solution with data security being prioritized.

 

Another difficult question in building the model is to tease out factors that are strong indicators in the prediction model without regressing to biases based on race and SES that are prevalent in the current judicial system and are socially deemed unfair.

 

Lastly, these tools could be enhanced by factoring in an inmate’s behavior in jail for their next trial to mitigate mistakes when they happen. If data shows that an inmate will likely not repeat a crime when they show good behavior during their sentence then this will provide us to have a further efficient system that will be as fair as possible while maintaining our goal of reducing costs on correction systems.

Sources:

 

https://www.themarshallproject.org/2015/08/04/the-new-science-of-sentencing#.wewudvPdi

 

https://www.brookings.edu/blog/techtank/2017/07/20/its-time-for-our-justice-system-to-embrace-artificial-intelligence/

 

https://www.wired.com/2017/04/courts-using-ai-sentence-criminals-must-stop-now/

 

https://www.thoughtworks.com/insights/blog/how-artificial-intelligence-transforming-criminal-justice-system

 

https://www.cs.cornell.edu/home/kleinber/w23180.pdf

 

https://www.uaa.alaska.edu/academics/college-of-health/departments/justice-center/alaska-justice-forum/34/3winter2018/a.pretrial-risk-assessment.cshtml

 

Team Members:

Mohammed Alrabiah

Tuneer De

Mikhail Uvarov

Colin Ambler

Lindsay Hanson

Opportunity:

 

Fender Musical Instruments Corporation is a manufacturer of guitars, basses, amplifiers and auxiliary equipment. They target all levels of players from beginners to experts, across musical genres. The opportunity has arisen from their struggles in retaining their customers since many initial learners drop out due to the difficulty of subject and no clear direction of learning. They realized that a person who goes past the first year of training buys equipment worth thousands of dollars over their lifetime (Business Insider). Thus, the opportunity is to use digital tools and augmented intelligence to retain customers.

 

Solution:

Fender utilized sensors on their devices, big data analytics and cloud computing to launch Fender Play, a digital service that could generate revenue. The basic advantages to their solution can be broken down into several points:

• Track users’ progress, and give them access to systematic modules across a variety of ‘paths’ so that they can learn according to their own style. This solution is also driven by a demographic shift where the majority of novice guitar players are trying to learn songs that they like, not necessarily learn techniques emphasized in traditional classes or play in a band.
• In addition to providing users the ability to skip certain lessons, allow them to go at their own pace to save time and money compared to traditional instructors.
•  Inter-device connectivity (guitars, amps, phone), with automatic updates and reminders, causing higher brand loyalty due to Fender owning the entire stack of guitar-related services.

 

Commercial promise:

Promise #1: Introducing a digital interface between a player and a guitar by adding sensors is a key enabler of the solution. Other musical instruments have interfaces that are easier to digitize (electric pianos). Digital pianos created a niche and made self-learning easier. We can expect a similar effect with digital guitars. An integration with social media and smartphones creates a collaboration platform that may, for example, allow remote band recitals, further reducing quitting rates.

 

Promise #2: Introduction of the smart guitar may do to a traditional guitar what electric piano did to an “acoustic” piano: displace it. Fender can benefit from a first mover advantage if the guitar market shifts towards digital.

 

Promise #3: Fender has considered the instructors segment by creating a tool that allows them to monitor and follow-up with students post-lesson during at-home practice.  By doing so, they are showing that the platform was built to be accepted by all customers in the market without exclusion or resistance.

Because of the lack of publicly available data, we have found it difficult to directly measure the effectiveness of the program. Thus, we cannot predict the potential revenue and growth from this segment.

Competition

 

Fender has traditionally competed with large instrument and especially guitar manufacturers. Gibson is the most prominent competitor in this space. Fender could change their business model and reposition themselves as a digital services company, they would find themselves going against competitors in the digital content industry (e.g popular channels on Youtube) or even community and collaboration-based music platforms such as Soundcloud or Bandcamp if it allows users to post their own music on FenderPlay. These competitors might be equipped with much better analytic tools and capabilities, but there might be certain niche markets where Fender has better brand loyalty and can attract people (guitar-heavy music like rock).

 

Proposed alteration   

 

Fender Play is priced at $19.99 a month for beginners. While this subscription price might seem reasonable, offering a free trial for customers who have purchased a Fender guitar would attract them to subscribe by trying it out before spending their money.

Fender should look into exploring the data collected and observe which lessons have users actually found helpful based on their level. By looking back at the data and observing which patterns were most likely for their users to stay as a loyal customer, Fender should then create clusters based on the data collected of each segment and create suggestions that were most likely to be helpful based on historical data and correlations. This strategy should start with casual users who typically dropped guitar lessons as part of their frustration because it narrows the suite of offerings while the digital ecosystem is built. With this increased intelligence and cloud storage capabilities, they might then be able to provide more personalized offerings in their ‘paths’.

A different strategy would be to link to popular music-tech companies like Spotify or Soundcloud to approach the target market that learns guitar to create their own music or collaborate with other artists.

Lastly, in order to keep their customers engaged so that they do not give up, Fender might incorporate a competition element into the platform, in which users would be able to see a scoreboard with their rankings against other peers. If there are rewards, badges, levels and an ability to compare progress with peers, people may have an increased incentive to stay involved than otherwise.

 

Sources:

 

https://blogs.wsj.com/cio/2017/07/25/the-morning-download-fender-launches-internet-of-the-guitar/

 

https://blogs.wsj.com/cio/2017/07/24/fender-amps-up-its-digital-play/

 

https://www.reuters.com/article/us-fender-musical-software/electric-guitar-maker-fender-jumps-into-online-learning-idUSKBN19R145

 

http://www.businessinsider.com/fender-play-review-2017-10

 

Team Members:

Mohammed Alrabiah

Tuneer De

Mikhail Uvarov

Colin Ambler

Lindsay Hanson