What are some innovative and practical applications of AI that have the potential to revolutionize industries or improve our daily lives?

[tanvipeopleplusai]

Can you share an innovative application of artificial intelligence that has the potential to transform a particular industry or domain, ultimately improving people's lives? Tell us more about it in this discussion thread.

If available, please provide specific examples of companies or organizations pioneering this technology, along with online resources where interested individuals can read more about it. Additionally, if you have any personal reactions or insights regarding the potential impact of AI in that industry, please feel free to share.

Tag a team member if you think this might be of interest to them.

[harsha-peopleplusai]

Here's a list of use-cases that I had made with Tanuj for AWS specifically on an Indic-LLM but it applies to AI in general:

Justice

• Supreme courts in India operate in English or Hindi, whereas High Courts and local languages may operate on local language. Eg: Anuvaad, already in use by the Supreme Court of India, would highly benefit from a more accurate model that captures nuances of indic languages used in a legal context.

Education

• In India, the medium of instruction is often a local language. Even in “English medium” schools, classrooms operate in a local language in the classroom by the same teachers as other mediums of instruction. Only the language of the exam is english. All core use cases of AI in education from teachers to students will need high accuracy, high guardrail/safety alignment use cases and hence translating on top of existing LLMs will not cut it. The increased “code-mixing” proficiency of an Indic LLM would further allow for AI to more naturally and effectively fit into existing education workflows of teachers and students.
• The Ministry of Education and DIKSHA already have started using Personalised Adaptive Learning (PAL) and using an Indic LLM would increase accessibility to students not comfortable with a certain language.
  High Accuracy translation of educational content from one language to another helps close the gap.

Business Education

• Blue-collar workers comprise a large percentage of India’s workforce (more than 250 million in number). Massive training efficiency can be unlocked for all corporations who hire seasonal or blue-collar workers with an Indic LLM because of a nuanced understanding of their dialects and language, which would greatly reduce training cost and increase productivity by reducing time to train. And reducing onboarding time and costs with localised learning materials would greatly increase productivity in manufacturing, construction, etc.
• This is equally true for India’s large market of grey collar jobs. Companies such as Quess Corporation are in conversation with people+ai about using AI to help with recruiting, staffing and retraining across 300+ companies. They are a significant player in the Indian job market, hiring up to 20,000 new employees every month and an Indic LLM use-case would greatly increase their efficiency.
• Startups like Apna.co, a job portal for blue and grey-collar jobs, can leverage Indic LLMs to enhance their platform's language capabilities, making job matching and skilling resources more accessible to a linguistically diverse user base

Copilots for Workers

• Gig worker co-pilots. Organisations Namma Yatri & Juspay which could employ Indic LLMs to offer more personalised and culturally relevant services to gig workers, answering a variety of service-related queries and also extend to financial inclusion initiatives. The number of gig workers employed is 7.7million and expected to grow to 24 million by 2030 and an Indic LLM would greatly expand this kind of workers via Skill Development platforms.

• Government worker co-pilots. An Indic LLM can assist with administrative work such as form-filling while retaining cultural context. Even say a 5% increased efficiency at the scale of 5 lakhs+ ASHA workers for example would greatly improve the productivity of delivering such services.

Government’s Administration (Across Health, Agriculture, & Other Schemes)

• PM-KISAN chat-bot by the Ministry of Agriculture would greatly be enabled by an Indic LLM that has support for more languages and is able to recognise nuanced differences in words and cultures across different regions for farmers. Given that more than 210 million workers in India would be in Agriculture and allied sectors, even a slightly faster resolution to their concerns would lead to much greater productivity.

• In programs like MGNREGA, even a 1% increase in communication efficiency due to the high-fidelity of Indic LLMs could translate to millions in saved costs, and given that it impacts more than 19 million households its implications would be profound. It can help in use-cases of improved localised communication, training and skill development, automation of administrative tasks, community engagement etc. for the Government.

• The 104 Government helpline (ArogyaVani) will greatly benefit from a multi-modal Indic LLM (future scope) that can retain cultural nuances and interpret dialectic differences to deliver medical services. This and other Govt. medical helplines will require accurate understanding of local language and dialect context etc.

Media & Entertainment

• The example of Kuku FM shows that content in local languages is doing significantly better with a wider reach. Translation of high fidelity content opens up English-language content to the rest of non-English speakers of India. Any high quality Indic LLMs will enable audiobooks for Bharat.
  Amazon Prime Video, Netflix, etc. will also be able to leverage this for subtitles that retain cultural context or in the future even speech translation with a multi-modal Indic LLM more accurately and cheaply.

Financial Technology

• In the fintech sector, Small Finance Banks (SFBs) like Ujjivan and Fincare, which cater to the unbanked and underbanked population, could greatly benefit from Indic LLMs. By providing multilingual support, these institutions can offer more personalised and understandable financial services to a customer base that speaks a variety of regional languages. This multilingual capability can enhance customer engagement, reduce misunderstandings, and potentially expand the reach of financial services to an additional millions of Indians, aligning with the government's financial inclusion agenda.

[TechKemon]

I'm sharing some specific initiatives in various domains using AI for Public Good. I have also tried to include links wherever possible.

• Sustainability -

Among urban sustainability initiatives, I stumbled upon an interesting initiative today called the 'Green light' project which aims at optimizing traffic lights to reduce emissions. Here's the link - Green light
Additionally, Oizom, an environmental AI company, provides solutions for monitoring and analyzing environmental data, vital for urban planning and pollution management in cities.

• Governance -

There are actually many great ideas here. e.g For instance, quadratic funding and quadratic voting (DemNext), Plural Property (RadicalxChange), and The Computational Democracy Project which maintains Polis, an open-source real-time system for gauging the consensus among large groups of people on various topics (CompDemocracy).(Basically they send a set of questions randomly to users which are voted upon as agreed /disagreed/pass by users)

• Health -

In the healthcare sector, Niramai employs an AI-powered system capable of detecting breast cancer in mammograms with superior accuracy compared to human radiologists, showcasing AI's potential in medical imaging (Niramai).
Artivatic is another notable startup focusing on streamlining insurance and healthcare processes through AI, enhancing efficiency, transparency, and personalization throughout the insurance lifecycle (Artivatic).

• Agriculture -

Cropin- based in Bengaluru, uses AI and big data to do climate smart farming, optimising agricultural practices in tune with weather patterns (Cropin)

• Language Barrier -

Addressing India's linguistic diversity, Reverie concentrates on AI-driven local language translation and processing, alleviating communication hurdles prevalent across the nation (Reverie)

• Education -

In education, Embibe integrates AI with AR/VR to furnish personalized learning experiences for students, making education both accessible and effective (Embibe)

• Overall -

Google's initiative, showcased on their responsibility page, mentions numerous impactful projects like Read Along, in collaboration with partners like UP Govt, Pratham, and CSF, aiming to bolster reading fluency (Google AI for Social Good).

[tanvipeopleplusai]

Use cases for People+AI with eGov Foundation

[Theme 1: First touchpoint of interaction (including localisation)]

• Citizen’s first touchpoint for logging complaints, interacting with government, etc
• First touchpoint for DIGIT community / System Integrators / Digital public good communities to
• Auto answer queries on forums from a large DB of existing information
• User training / Capacity building

[Theme 2: Image processing and actions: Capturing images and performing certain set of actions]

• Capturing proof of work for contractors and automating / automatically triggering the payment process
• Capturing attendance in health campaigns, health facilities, schools etc.
• Performing first level disease surveillance
• Understanding coverage of a health campaign
• Capturing attendance of gov. work beneficiary at worksite ex: MGNREGA

Theme 3: Participatory Budgeting

[Theme 4: Using AI in public works / Public Finance Management space]

[Theme 5: Data analytics on existing datasets, personal assistant for employees and intuitive query builder]

• Monitoring health campaigns
• For employees, query large databases and respond to specific queries for government employees

Theme 6: Revenue/ Expenditure Analysis to improve Financial Management of state/ Union government or department.

---

Theme 1 : First touchpoint of interaction

Use case 1 : First touchpoint for citizen interaction
Voice based complaints registration and voice based responses to the complainant (AI chatbot as first touchpoint for the citizen) for low bandwidth regions

Context: While registering a complaint on the public grievance redressal portal (DIGIT), currently, citizens have to go through 7-8 steps.

Field observation: We have observed officers sending voice messages to the citizens mobile number for the complaint. Officers too find it easy to explain the issue over voice instead of sending messages. Example, we cleaned the bin in the morning at 10.00 and at 12.00 you raised a complaint for overflow of garbage ,we have a routine of vehicles reaching the bins at your area at 10 AM , 2 PM and 6 PM. Your problem will be solved in the next trip for disposal. It becomes tedious to type and since the information is not shared real time, citizens end up reopening the complaints. This reflects negatively on the performance of the ULB as well.

Proposed solution: It will be easier for citizens to take a photo and speak about the complaint instead of writing it down - speaking is a more natural way of articulating the problem. Application should be able to listen to the voice message and categorise the complaint types and departments automatically.

A clear and prompt messaging to citizens would help citizens to understand that the authorities are working on it, they are hearing their queries and doing something about it. We can also work on translation into local languages, and can explore integration with Bhashini in India.

Implementation Scope: This can be adopted in Punjab, here we have seen a drop in complaints being raised as they have other whatsapp channels to resolve this complaint which the officers find a better mode of managing the complaints.

Use case 2: DIGIT chatbot to help answer queries on DIGIT community posts
Context: System Integrator(SI) teams implementing DIGIT(UPYOG) have team members with varying degrees of technical skills. We offer technical support on regular calls but that's for troubleshooting and the support team doesn't have bandwidth to cover all the basic technical questions. Currently, we have 4-5 teams which are implementing UPYOG. Later on, this number can increase to 20+ for (Urban only) which will be working on implementation.

Data sources: We also have a discussion board where queries can be raised by tech teams (https://github.com/orgs/upyog/discussions). We also have a JIRA ticketing process where major issues are logged and support is provided.

Proposed use case: DIGIT Chatbot can serve as the training assistant cum troubleshooter for SI partners. It will help in following 4 ways:
i) Gaps in the documentation can be covered by AI
ii) User can go to any depth and cover any topic
iii) Support capacity of central teams will be augmented
iv) Deriving insights from support calls and enriching documentation with its help

Implementation Scope: The solution may be tested in Urban to begin with but will be useful for any other mission of eGov.

Use case 3: Voice enabled / intuitive training systems for users
Context: Need for multilingual, voice enabled and multi-modal interfaces that enable trusted interactions and access to government services at speed and scale. We can leverage AI to design innovative service interfaces that bridge the digital divide, making it easier for users who may not be digitally savvy to interact naturally with government services.

Implementation scope: AI/voice enabled interface for Health Front Line workers in the Salama mobile app. We can build this as an add-on (optional) feature to the existing interface. It can help reduce training time significantly, can accommodate diverse health care workers and easy deployment.

mGramSeva mobile app that is getting rolled out through Jal Jeevan Mission (JJM) across all villages in India. It can help to have increased reach and effectiveness in remote villages.

Theme 2: Image processing and actions : Capturing images and performing certain set of actions

Use case 1: Tracking proof of work & identifying vector type in health campaigns
Indoor Residual Spray (IRS) involves coating the walls and other surfaces of a house with a residual insecticide. For several months, the insecticide will kill mosquitoes and other insects that come in contact with these surfaces. The IRS does not directly prevent people from being bitten by mosquitoes. In this campaign, the proof of work can be assessed through image scanning and assessment of the spray bottle level, spray area. Assessing the spray surface area images can be correlated with the quantity of spray used to corroborate the efficiency of spraying

The IRS campaign requires extensive training since the SOP needs to be followed strictly for good coverage. Some of the parameters to be observed are distance from the wall, angle of application, speed of application, spray pattern, and overlapping. Hence assessing the images/videos of the spray application can help to predict the coverage and efficiency of the campaign.
Identifying the type of vector (mosquito) by image scanning and corresponding insecticide can be used for LLIN (Long Lasting Insecticide Nets) and IRS. To minimise the development of insecticide resistance in mosquito populations, it is advisable to use different classes of insecticides for IRS and LLINs. If LLINs contain a specific class of insecticide, it may be beneficial to use a different class for IRS in the same area. Hence during the IRS campaign, the LLIN available can be assessed to check the compatibility or vice versa.

Use case 2: Taking measurement of work completed for public works
Context: In the realm of civil engineering, where an average of 150 to 200 projects are executed by an Urban Local Body (ULB) within a single financial year, each project entails weekly measurements, verification, and approval by authorities. These projects typically have an average completion period of 4 to 5 weeks. In addition, site measurements are essential for the accurate estimation of work. However, the process of preparing such estimates is time-consuming. AI offers the potential to significantly streamline and expedite these tasks, making them more efficient for civil engineers.

Implementation Scope: Government has used these techniques in the past (road, borewells) using images being captured via drones but this was limited to pilots. MuktaSoft is currently being implemented in 2 ULBs (Odisha), which will soon be scaled to 20 ULBs; we can explore this further with the state government.

Use case 3: Fraud detection | Revenue maximisation assessment and identify properties in an area and identify gaps in property assessment

Context: There are multiple leakages in property taxes as citizens do not report
Type of establishment: commercial setup in residential area
Extent / area of establishment: multi floor area is reported as single to reduce property tax
Missing properties

Data sources: Existing DIGIT property/water tax/trade licence data, GIS open buildings data

Proposed solution: Using GIS and building / settlement data and datasets captured in DIGIT, image scanning can be used for revenue optimization during property tax collection. Assessing the building location, parameters such as type of construction, number of rooms, elevation, number of floors can help in estimating the revenue leakages. AI can leverage data analytics to prompt the gaps in settlements using different data sets.

Implementation Scope: Can be taken up in Punjab where fiscal debt is one of the highest and would help maximise revenue for the government.

Use case 4: Recording attendance

Context: Can we look at innovative ways to record attendance for

1. Teachers at school
2. Health workers during health campaigns
3. Contractual workers for public works
4. Attendance of MGNREGA related work beneficiaries on worksite

It can involve scanning a QR code, taking their picture and mapping it with geo coordinates, etc and generating a daily report.
Concerns around data privacy, consent and security need to be thought through for this use case. The implementation of this use case is required but also might be adopted with resistance. Hence needs government policy / support for implementation.

Theme 3: Participatory Budgeting | Social audit via citizens’ voice
Use Case 1:Participatory Budgeting
Context: Traditional budgeting processes often suffer from a lack of transparency and citizen involvement, leading to a disconnect between government priorities and the genuine needs of the people. This gap can result in inefficiencies, misallocation of resources, and diminished trust in the government.
Governments also want to understand if the money they have spent on something, are the citizens benefiting from it? The audit is generally done by government employees, it would be interesting to see if citizens can be a part of this feedback.

Implementation scope: DIGIT's AI models streamline data collection and citizens inputs, analysis, and reporting, accelerating the budgeting process and eliminating workflow inefficiencies. By analysing historical budget data and socioeconomic indicators, decision-makers gain insights into community priorities and the potential impact of budget allocations. Leveraging DIGIT's data, AI can enhance the efficiency, accuracy, and effectiveness of participatory budgeting.

For Example, analyse patterns using feedback from citizens via phone calls, messages from the citizens to listen to how budgets need to be allocated to different heads. Basically understand what is important for citizens and allocate budgets accordingly.
Reference : https://decidim.org/

On completion of a public works initiative, governments can connect with citizens and understand if they are satisfied with the work that was undertaken.

Theme 4: Using AI in public works / Public Finance Management space
Context: MUKTASoft is an application that has been developed using the works platform in Odisha. It enables ULBs to manage the contracting of works to community based organisations, execution and releasing the payments. It has three types of bills to be paid,
Wage bill to make payments to wage seekers, Purchase bill to make payment to suppliers, and Supervision bill to make the payment to community based organisations.

Capturing Attendance: To generate a wage bill attendance of wage seekers is a must which is currently captured using a mobile application by tapping into full day, half day present, or absence. We are looking for a more simple and authentic way of capturing the attendance.

Planning of Work: The estimated amount of the work is derived from an estimate which is prepared into the system, but there is no way to plan a work with the timelines. We are looking for an auto planning algorithm which can use the measurement data for planning. Measurement Book having measurements is offline as of now.

mGramseva
Context: mGramSeva is an application that has been developed as part of the iFIX exemplar in Punjab. It enables rural local bodies to manage revenue and expenditures related to water supply projects using a mobile app. The app will be used by members of the water management committee and collection agents. The collection agent will be able to add consumers (or households), generate bill and receipt acknowledgements.

Scope for Analysis on data :
Fiscal sustainability of GPWSC: To assess the fiscal sustainability of the "GPWSC,". Fiscal sustainability typically refers to the ability of a GPWSC to maintain its financial health and meet its long-term financial obligations. Based on the data available w.r.t revenue and expenses, the model should be able to assess if the GPWSCs are fiscally sustainable.

Demand generation: Based on the operational data, the model should be able to predict and alert if the demand is needed to be modified appropriately. This can also be based on the population in the village and other projects that are planned. Data for this may have to be scouted from other sources/systems.

Bill collection (on date and delayed)
Revenue behaviour (anomalies) and Expense behaviour (anomalies)
Risk of collections, Fraud and compliance
volumetric delivery of the water for villages
Combining demographic data, past payment data and PGR data to predict the propensity of a citizen to pay so revenue collection efforts can be effectively redirected

Theme 5: Data prediction, personal assistant for employees and intuitive query builder

Use case 1: Dashboard + intuitive query builder
Context: Program supervisors and administrators (in health campaigns specifically but also applicable to other use cases) use data on DSS (dashboards) to make decisions. Sometimes, the data can be too much and it gets difficult to establish correlations, patterns and anomalies. Can we use AI to

• Provide a daily summary of how the campaign performed on the day.
• Identify anomalies that need further investigation (real time information that is actionable - say, a team that needs attention; or a set of data that doesn't appear to be recorded from the expected location)
• Give a prediction on, say, excess and shortage of stocks and the suggested movement of stocks between districts.
• Allow the ability of administrators to 'ask questions' on the performance of campaigns.
  And all this needs to be presented in plain language.

Use case 2: Create, Search and query large databases in minimal time
Context: Road Cuts are done for various utilities, O&M and New Installations. A lot of independent organisations are involved in handling the utilities and require road cuts for their implementation (New and Repairs & Maintenance). ULBs receive requests from the Organisations for the Road Cut Permissions and provide permissions in a decentralised way based on their preference and priority. Lack of information on the age of the Road, climatic conditions, traffic conditions, curation of the road cut and limited idea on the time it takes to restore.

Proposed solution: We can build a Dataset of All the Roads and update regularly on the Repairs and Maintenance of the Roads. This will enable to take a Decision based on the Facts and field Conditions.
Role of AI: Search and query comprehensive Data of all the Road Cuts across the City which are active and probable measures to be taken can be derived by the AI tool.

Implementation Scope: Road Cut is one of the most irritants to the Citizen and has a huge impact politically. If the AI tool provides comprehensive data and suggests the corrective/preventive measures the problems can be brought down significantly.
We do not have a buy in from any government yet but this use case is scalable across departments for different data sets which are used across departments (not only road cuts)

Use case 3: A digital personal assistant for government employees: Query reports, timesheets etc. prior to meetings
Instead of spending time collating data around delayed projects, pending payment etc, an aI chatbot can help Mechanical Engineer or a Junior Engineer with talking points with a contractor. It can be to pull out contractor details and provide information on delayed projects, pending payments, quality of work etc. so they can have a constructive discussion and highlight relevant issues.

Use case 4: Better planning based on previous initiatives
Context: Health campaigns go through a macro and micro planning process prior to the actual campaign to estimate the following

1. Target population (who?)
2. Resources required (what do we need?)
3. Where are they located? (geo spatial distribution)

Macro planning is based on previous campaign datasets, world population data and geo spatial settlement data. It is done one to two years before the actual campaign starts.

Micro Planning involves understanding who are the target citizens, where are they located, what resources will I need, how do I plan the transport/logistics?

Both of these exercises are done manually and also leverage tech in parts. For example, few campaigns use geo enabled microplanning to automate the process and make it more efficient. However this still involves pulling different datasets together and assigning teams to designated areas in the same or different systems. AI can streamline and simplify this arduous, tedious and manual-intensive task.

Theme 6: Revenue/ Expenditure Analysis to improve Financial Management of state/ Union government or department.

Use Case1:
Context: Any government be it state or union collects revenues from various sources like taxes, fees etc, and also raises capital receipts through debts. On the other hand there are payments and outflows in the nature of recurring expenses like salaries and others like payments for projects, loans etc. which are committed but become due on occurrence / completion of certain milestones. The cash availability with the finance department may not match the revenue and expenditure side needs at all times.

Problem Statement: While the overall budget in terms of total receipts and expenditures are planned at the beginning of the year, these may not take care of the cash-flow needs of the treasury at all times. Receipts or debt inflows may be expected later but the payments might become outstanding earlier, leading to the treasury or finance department constantly prioritising the payments in an ad-hoc manner and holding reserves of cash to meet an unexpected liability. This results in delay in delivery of essential services, vendor payments etc., creating mistrust in the government financial ecosystem.

Based on historical trends of revenue, receipts and expenditures, systematic alerts can be generated in advance as to when the cash balance is expected to go below matching expenditure requirements, so that an action can be taken in advance to either defer payments or raise cash. AI based models can help specify a range in terms of days for which delays are expected or payments are expected to be held up. This will result in greater trust and transparency in the government financial ecosystem.

[meghana-peopleplusai]

Find below a list of AI use cases within the agriculture sector (primary production) :

• Monitoring Soil Health: AI systems can conduct chemical soil analysis and estimate missing soil nutrients. Agrocares (Dutch Company) Nutrient Scanner, analyses data from soil samples, provides estimates of missing nutrients and gives customized recommendations, allowing farmers to adjust their fertilizer application and irrigation practices.

• Crop Monitoring and Advisory : AI-powered sensors and image recognition tools, allows farmers to detect and track crop changes. This would enable them to predict when crops will reach maturity, leading to cost savings and reduces wastage. Cropin (Indian company) is an example of an AI-powered agriculture analytics firm that utilises AI to enable crop monitoring and advisory.

• Smart Irrigation: AI-powered irrigation systems offer a number of benefits. It allows us to monitor and adjust the amount of water being applied to crops based on moisture levels of the soil, soil characteristics and weather conditions, increasing water efficiency. It also identifies maintenance problems with irrigation systems, such as clogged nozzles or broken pipes, saving water, time and money | Fasal is an example of a company that utilises AI technology to check water availability in the soil and ensures irrigation requirement is met -based on crop, its stage, and soil characteristics

• Insect and Plant Disease Detection: AI-powered systems can detect insect and plant diseases. It can also recommend effective treatment of the same. DeHaat is a Gurgaon-based start-up that provides farmers with access to more than 3,200 agricultural inputs and AI-based customised crop advisory on pest and disease management, delivered via mobile app and call centres. Taranis is another example of a company that uses AI, computer vision and machine learning to analyse high-resolution images of crops, identify signs of stress or disease and suggest solutions. This reduces the need for broad spectrum insecticides and decreases pesticide resistance.

• Intelligent spraying and weeding: Weeding and pest control can be automated with AI technology. Using computer vision, weeding robotics is remarkably precise, resulting in a 90% reduction in pesticide usage. Based on data analytics, these tools can calculate how much pesticide is needed for each field. Blue river technology's ‘See and Spray’ machine differentiates crops and weeds, applying herbicide only where needed.Bosch’s BoniRob is an example of an agricultural robot which finds and removes weeds by driving a bolt into the earth.

• Chatbot for farmers: Chatbots can act as an interface between farmers, customers and distributors. Farmers can use them to answer questions about products, services offered, order supplies, and check inventory levels. Microsoft's FarmVibes.Bot provides farmers with personalized advice and recommendations based on data.

• Farm data collection and Analysis: : Microsoft’s FarmVibes.AI extracts intelligence from farm data and remote sensing sources. This can help farmers and agribusinesses make informed decisions.

• Farm Geotagging: Geo-tagging can help users find a wide variety of location-specific information and AI can be used to achieve this.