Ashis Panigrahi Innovation, Society, Technology

Everyone will be a prompt engineer!

Generative AI is dominating conversations due to its transformative impact across industries. From generating creative content to aiding in problem-solving, these systems are revolutionizing how we approach tasks. With their ability to innovate, automate, and personalize, Generative AI technologies are at the forefront of shaping the future of technology and human interaction. Ensuring technology advancements benefit all, not just the privileged few, is paramount. While major tech players and governments strive for inclusive and equitable technology use, individuals must also acquire basic skills to prevent widening skill gaps. Accessible technology empowers everyone to participate fully in the digital age, fostering a more equitable and prosperous society.

Crafting well-designed prompts empowers users to steer the AI towards desired outcomes, maximizing creativity and relevance

Introduction

Rule based systems vs Trained systems

Trained systems learn from the data, they adapt and improve over time. Whereas, programmed systems follow predefined instructions without any learning capabilities. Today’s intelligent systems (using AI technology) are trained, not programmed. Instead of programming or writing specific rules to solve a problem, AI systems are fed with examples of ‘what it will encounter’ in the real world. AI systems then detect the patterns and produce their own ‘rules’ based on the examples. While trained systems offer flexibility and adaptation, programmed ones excel in precise execution.

So, Artificial Intelligence (AI) is the science and engineering of making intelligent machines, that can perform tasks which typically requires human intelligence (John McCarthy, 2007). At its simplest form, Artificial Intelligence is a field which combines Computer Science with robust data set.

AI systems typically perform functions like ‘classifying data’ (e.g. assigning labels to images), ‘grouping data’ (e.g. identifying customer segments with similar purchasing behavior), or ‘choosing actions’ (e.g. steering an autonomous vehicle). 

What is Generative AI?

Generative AI (GenAI) is the latest development in AI field, where the primary function of the AI system is to generate content, content that is similar to/ indistinguishable from human-created content.

Generative AI is crucial because it empowers us to create new content, ranging from art to music to text, with minimal human intervention. Mastering generative AI allows for innovative problem-solving, creative expression, and automation across various fields, making it a valuable skill in today’s rapidly evolving technological landscape.

Who are the consumers?

Consumers of generative AI systems can be diverse, ranging from individual creators and artists to businesses and organizations across industries such as entertainment, marketing, design, education, healthcare, and more. These systems cater to anyone seeking automated content generation, personalized recommendations, enhanced creativity, or improved efficiency in their work processes. Additionally, consumers could include developers and researchers interested in advancing the capabilities of generative AI itself.

Why to embrace generative AI?

Understanding generative AI skills can be incredibly valuable for several reasons:

  • Creative Expression: Generative AI empowers individuals to express their creativity in new and exciting ways, whether it’s through generating art, music, literature, or even entirely new concepts. This skill allows for innovative exploration and pushes the boundaries of traditional creative processes.
  • Problem Solving: Generative AI techniques can be applied to various problem-solving scenarios, such as generating synthetic data for training machine learning models, designing optimized structures in engineering, or simulating complex systems for research purposes. By mastering generative AI, individuals gain a powerful toolset for addressing real-world challenges.
  • Career Opportunities: With the increasing integration of AI technologies across industries, proficiency in generative AI can open up numerous career opportunities. From roles in software development and data science to creative fields like design and entertainment, individuals with generative AI skills are in high demand and can command competitive salaries in the job market.

Can Gen AI systems make mistakes? 

A big yes. While it strives for accuracy and coherence in it’s response; there might be occasions where AI systems can misinterpret the context and provide information that’s outdated or inaccurate.

Large Language Models (LLM)

Large language models are advanced AI systems trained on vast amounts of text data, capable of understanding and generating human-like language. They exhibit impressive language comprehension, enabling them to perform tasks such as text generation, summarization, translation, and question answering. 

Generative AI systems are based on large language models which forces users to understand how they are trained. Training a LLM involves feeding it vast amounts of text data to learn patterns and language nuances. The process begins with pre-training on a diverse dataset to impart general knowledge and language understanding to the model. Fine-tuning is the next step which involves training the LLM on a more specific dataset for a particular task or domain. During training, the model adjusts its internal parameters through iterations to minimize prediction errors and improve performance. Finally, validation datasets are used to monitor the model’s performance and prevent overfitting during training. 

Despite their capabilities, challenges such as ethical concerns, biases, and energy consumption highlight the need for responsible development and deployment of these models.

Key advantages of Gen AI systems

  • Creative Output: Gen AI can generate novel and creative content, including images, music, and text, which can be valuable for artistic endeavors, content creation, and design projects.
  • Automation of Tasks: Gen AI can automate repetitive tasks, such as data entry, image generation, and text summarization, freeing up human resources for more complex and creative endeavors.
  • Personalization: Gen AI can analyze large datasets and user preferences to generate personalized recommendations and experiences, enhancing customer engagement and satisfaction in various industries such as e-commerce, entertainment, and marketing.
  • Problem-solving: Gen AI can assist in problem-solving by generating solutions, optimizing processes, and identifying patterns in data, enabling organizations to make data-driven decisions and innovate more effectively.
  • Efficiency and Scalability: Gen AI can improve efficiency and scalability by automating workflows, optimizing resource allocation, and accelerating innovation cycles, leading to cost savings and competitive advantages for businesses and organizations.

Known dis-advantages of Gen AI systems.

While Generative AI (Gen AI) offers numerous benefits, it also presents some disadvantages:

  • Ethical Concerns: Gen AI raises ethical concerns regarding the authenticity and ownership of generated content. There are debates surrounding the potential misuse of AI-generated content for spreading misinformation, creating fake news, and manipulating public opinion.
  • Bias and Discrimination: Gen AI models trained on biased datasets may inadvertently perpetuate biases and discrimination in generated content. This can lead to unfair outcomes, reinforce stereotypes, and exacerbate societal inequalities.
  • Security Risks: Gen AI poses security risks, including the potential for malicious actors to exploit vulnerabilities in AI systems to generate fake identities, bypass authentication mechanisms, or launch cyberattacks such as phishing scams and social engineering attacks.
  • Loss of Jobs: The automation capabilities of Gen AI may lead to job displacement and unemployment in certain industries, as AI systems can perform tasks more efficiently and cost-effectively than humans. This can have socio-economic implications and require workforce retraining and reskilling initiatives.
  • Hallucination/ Dependence on Data: Gen AI models require large volumes of high-quality data for training, which can be expensive, time-consuming, and resource-intensive. Additionally, reliance on data may raise privacy concerns and require compliance with data protection regulations. Moreover, if the data is biased or incomplete, it can lead to inaccurate or unreliable AI-generated outputs.

As per Mckinsey report, economic potential of Generative AI is $4.1 Trillion each year, that is 4.4% of the total economic output. Generative AI has the potential to change the anatomy of work, augmenting the capabilities of individual workers by automating some of their individual activities. Generative AI will have a significant impact across all industry sectors. Banking, high tech, and life sciences are among the industries that could see the biggest impact as a percentage of their revenues from generative AI.

How to unlock the potential of Gen AI?

To unlock the potential of Generative AI as an individual, one should learn and experiment by diving into learning Generative AI techniques/ tools/ frameworks. Experiment with small projects to understand its capabilities and limitations.

Continuous Learning is needed to keep up with the latest research, advancements, and best practices in Generative AI. Individual may want to share findings, insights, and projects output with the community in order to foster collaboration and exchange ideas, expertise for growth and learning.

The key is to also stay informed about ethical considerations surrounding generative AI, such as bias mitigation and privacy concerns; and ensure responsible usage and deployment of AI-generated content.

Be a prompt engineer

Prompt engineering is a strategic and creative endeavor that shapes the interactions between humans and artificial intelligence systems. This is the process of crafting effective and specific prompts to guide artificial intelligence systems. It involves designing questions or statements that elicit desired responses from AI models. It is crucial in influencing the output and behavior of AI, ensuring accuracy and relevance in the generated content. By carefully constructing prompts, engineers can control the direction and focus of AI models, leading to more tailored and useful responses. 

The effectiveness of prompt engineering can greatly impact the overall user experience and usefulness of AI-powered tools and services.

Conclusion

In summary, prompt engineering serves as a strategic tool for harnessing the power of Generative AI, by guiding the model’s output towards desired outcomes. Crafting well-designed prompts can steer the generative AI model to produce outputs that align with specific goals or criteria, ensuring precision and relevance in the generated content. Prompt engineering can help mitigate biases in generative AI outputs by providing prompts that encourage fair and unbiased content generation, promoting diversity and inclusivity. And finally, tailoring prompts to specific domains or tasks enables generative AI models to produce content that is relevant and useful within those contexts, enhancing their practical applicability.

References:

McCarthy, John (2007), WHAT IS ARTIFICIAL INTELLIGENCE? Stanford University, https://www-formal.stanford.edu/jmc/whatisai.pdf

Mckinsey Report (2023), The economic potential of generative AI: The next productivity frontier. https://www.mckinsey.com/capabilities/mckinsey-digital/our-insights/the-economic-potential-of-generative-ai-the-next-productivity-frontier.

Ashis Panigrahi Innovation, Technology

Blockchain and IoT technology for Train Signalling System


Railway systems/networks are divided into sections (also known as blocks) to avoid collision between trains; because more than one train is NOT permitted to run on the same section of the track at the same time. The recent train accident in India along with the casualty numbers has warranted to revisit the train signalling system and how technology can come for rescue. The probe report flags multiple protocol breaches including signalling system as reason for the accident (1).

Combination of Blockchain, IoT and AI technology will play a big role in addressing this real-life challenge.

Signalling system plays a vital role in a Railways network to monitors the trains movement and operates the trains in a safe manner. Finding issues with signal failure  before time will not save lives and national infrastructure. Key objectives of a Rail signalling system are:

  • Maintain a safe distance between trains
  • Control the trains movement at junctions
  • Prioritise track allotments based on business rules (Fast train vs slower trains vs goods trains)

Internet of Things (IoT) based event logger are being used in many countries to monitor and log the data related to train movements, but many a times this logger is not integrated or overlooked by personnels during the decision making which are primarily manual in nature. Also the usage of IoT systems are limited to predict failure of equipments to alert signal engineer for maintenance rather than using in real-time decision makings.

Problem statements:

In the above diagram, a passenger train has got priority over a goods train. When both the trains approaches a station which is equipped with a “passing loop” track, Peer to Peer (P2P) negotiation takes place. In this case, passenger trains negotiate with Station master to get access to the main line, and good train negotiate with station master to divert into passing loop track. Just imagine when trains are also passing in opposite track at the same time, the station master will get engaged in similar P2P negotiations. Such a P2P coordination can become with more number of trains participating in the network and also if train lines are crossing each other (diamond crossing). The real treat to safety is, the outcome of such a bilateral coordination (P2P) is provisional until all the coordinations are completed (between station master and participating trains).

Concurrent one-to-one coordinations are potentially complex and inefficient or even dangerous if done manually or by unsophisticated approach. All the coordinations between peers must be able to execute the same safeguarding logic that take into consideration of all relevant actors within the given railway network.

Secondly, in some cases it is seen that, once a section of track is allowed to one train post P2P negation, the track fails to switch as intended and hence creating accident situation.

Solution Components:

Blockchain: Blockchain technology provides both centralized and decentralized aspects of a solution.

It will hold the network-wide data at central level which is unique truth and can be shared across all participating nodes (information like train routes, timetables, passenger reservations, infrastructure reservations etc.). It will hold network-wide rules to safeguard trains in terms of smart contracts (stored on the ledger). These are software defined rules and will get executed when certain conditions are met.

The governance will be decentralized, business network administration is based on consensus and on transparency. Transactions are verified (endorsed) and ordered (serialized) by peers/orderers distributed over the participating network. When participating trains in a section or within a defined proximity met certain criteria consensus being taken in defined channel and corresponding smart contracts are verified and ordered for execution. This way, any bottleneck from central broker that is responsible for information distribution is avoided.

Internet of Thing: IoT devices are used to provide real-time situational data from the ground.

Once consensus is arrived in Signalling system for a train to get access to a particular section, the Switch operators makes sure the required switches happens before a green signal provided to the train driver.


So the IoT system will be a participant of the Blockchain system to provide a finality to the transactions before it get executed. IoT devices can be used to monitor the track changes and pass/fail flag will come back to Blockchain system as input before it commits the transaction. Computer vision/ camera solutions can be also used alternatively.

The status of the signal (green or red) on a particular track can be easily  broadcasted by a wireless transmitter to any near-by approaching train. This come handy especially early morning hrs with less visibility where train driver need not have to look for the signal outside, but will get the status inside the driver cabin itself.

IoT devices using RFID and other technology can be used to build Anti Collision Systems for railway which can sense static or slow moving objects on the track. Indian railway has been using Train Collision Avoidance System in certain sections of the network successfully (2).

Conclusion

The Blockchain-based signalling System will be a software defined safety mechanisms that can be considered to automate train traffic management. Exponential technologies such as Edge Computing, 5G, IoT and AI can be used to bring right capability for  implementation of such a system.

Blockchain and IoT solutions together will provide the real time data to all the stakeholders including Station Masters, Section Controllers, Signal Engineers and others. By sharing real time data on train movement with data integrity and transparency in place, this solution will enhance decision making capability. All the participating trains can communicate with each other, and decisions will be executed through smart contract that can avoid train delays which typically happens due to information gap and lack of trust with the existing systems.

Data will be stored in a distributed ledger using cryptography which will ensure data integrity and tamper proof. Applications and user interfaces can be build on Blockchain system to authenticate participants (driver, dispatcher, station master) based on face recognition or similar technology to avoid security risks.

References:

  1. Indian Express (2023, 5 July). Probe report flags multiple protocol breaches for Bahanaga Bazar crash. https://www.newindianexpress.com/states/odisha/2023/jul/05/probe-report-flags-multiple-protocol-breaches-for-bahanaga-bazar-crash-2591586.html
  2. India Times (2022, 3 April). Explained: How Railways’ Indigenous Anti-Collision System Kavach Can Prevent Train Accidents. https://www.indiatimes.com/explainers/news/indian-railways-indigenous-anti-collision-system-kavach-565956.html

 

Ashis Panigrahi Society, Technology

Smartcity 3.0

Smart city is not a new concept, it exists for more than a decade now. Rather, I would say the definition of Smart city has changed over the time. During early 2000, definition of Smart City was more towards, connecting all disparate systems like land management system, traffic, telecommunication, electricity network, water supply, sewerage system; to operate the city in a seamless fashion. Focus was more on responding to crisis; either caused by nature or man-made, orchestrating various sub systems of a city to work in tandem and act in a wholistic manner. I call this as smart city 1.0. In the second wave, it was more towards deploying various types of technology methods, sensors devices to collect real-time data and use the Insights gained from those data to better manage the assets and resources. In this Smart City 2.0, focus was more on making city services more efficient by making sure bottlenecks in road infrastructure, energy distributions, sewerage network are identified and addressed before even they arise. Now we are in 3rd wave of Smart city which I termed as Smart City 3.0, where the focus is to address the human aspects, the community aspects, to drive economic growth and at the same time to improve the quality of life and improve collaboration among citizens. Extreme urbanization has got many challenges for the cities, either people have migrated from rural areas to the city centers or people live in city’s outskirts and drive to their work place as housing near city centers are not affordable. This housing crisis is due to Euclidean zoning where city divides land into residential, commercial, and industrial areas that resulted into road congestions as people travel long distance to work on daily basis.

Following are key characteristics of the smart city 3.0 where technology as enabler, is going to play major role in transforming it:

  • Making city more liveable. More than 50% of city’s landscape is occupied by roads and parking lots (Gardner, 2011), cars on the road are oversized, no place for pedestrians, and not enough playgrounds for kids. Challenge is to plan/design our cities for people and not for machines.
  • Affordable housing. Housing in cities became expensive for young professionals and lower income group citizens. Ideas and plan for affordable housing for all sections of people to be promoted.
  • Reduced traffic congestion. City to offer workplaces closer to affordable housing and provide access to all the services like schools, restaurants, shopping and other within walking distance where they live. This will not only cut down traffic congestions but will promote collaborations among citizens.
  • Ride sharing. Electric vehicles will solve the air pollution problem, but wouldn’t address the road congestion challenges that cities are facing today. Culture of ride sharing needs to be nurtured in the city. Research has shown that the risk of collision between pedestrians and motorised vehicles reduces when more number of people starts riding cycles (Elvik, 2009).
  • Innovation Districts: Establishing innovation districts in city center with closer proximity to university will attract young talents into the city, will promote new and diverse ideas.

References:

  1. Gardner, Charlie: 2011. Old Urbanist: https://oldurbanist.blogspot.com/2011/12/we-are-25-looking-at-street-area.html
  2. Elvik, Rune: 2016. Accident Analysis and Preventions:  https://www.sciencedirect.com/science/article/abs/pii/S0001457509000876?via%3Dihub
Ashis Panigrahi Technology

Application of Digital Twin Technology

Digital Twin technology is much more than representation of 3 dimensional view. This can produce value when augmented with real time data collected from IoT devices and analysed with Machine Learning/ AI techniques. As the data is expanding exponentially with billion devices getting connected where sensors are providing reliable data on real time basis; Digital Twin technology are coming handy in analysing and assessing products, services and processes for efficiency, better maintenance, and bringing many other benefits. Since the data continues to flow, simulations made by the digital twins can learn, improve, get continuously updated on near real time basis and models can be deployed in the physical world very quickly.

This is a promising technology with a time to acceptance of 5 to 10 years as featured in Gartner’s hype cycle for emerging technology trends for 2017 (Gartner, 2017). Market forecast is to have one-half of companies using Digital Twins to improve their products, services, and to solve real life challenges like traffic management, climate changes among others in next five years or so.

What is a Digital Twin?

Digital twin is a digital representation of a physical entity or system. The object can be an actual physical assets like building or can be a product like vehicle or a city by itself. These virtual replica gives control over the product or process from the design phase to the deployment phase, hence we will see a growing demand of digital twin technology globally. Digital Twin is getting widely adopted due to multiple benefits it offers, such as real-time monitoring, reduce product defects, shorten time to the market, extend the life of assets and reduction in production cost.

Why Digital Twin technology is important?

Digital twins are powerful masterminds to drive innovation and performance. Digital twin technology helps companies improve the customer experience by better understanding customer needs, develop enhancements to existing products, operations, and services, and can even help drive the innovation of new business.

Digital twins are virtual replicas of physical devices that data scientists and IT pros can use to run simulations before actual devices are built and deployed. They are also changing how technologies such as IoT, AI and analytics are optimized. Digital twins can marry human and AI to produce something far greater and simulate complex systemswhich wasn’t possible otherwise. Also combined with AR, VR and related technologies provide a framework to overlay intelligent decision making into day-to-day operations. Digital twins offer a real-time look at what’s happening with physical assets, which can radically alleviate maintenance burdens.

Where Digital Twin technology can be used?

Electronics & Manufacturing Industry are expected to b ethe largest users of this technology. But increasingly Digital Twin technology is used in government, Automotive,  Energy and Utilities, Retail, Healthcare and many other industries. The applications range from Product Design & Development, Inventory Management, Manufacturing Process & Planning, city Planning, Traffic managements and others.

Retail Industry: During the pandemic retail industry was one of the most affected business next to travel and hospitality. Consumer experience is critical in the retail industry and Digital twin implementation played a crucial role in augmenting customer experience by creating virtual twins for customers and modelling fashions for them on it. Other benefits of Digital Twins are in the area of floor space management, security features implementation and energy management among others.

Smart Cities: Creating a replica of the physical world Digital Twin is rapidly becoming indispensable technology to visualize the city  in real-time; layered with buildings data, urban infrastructure, utilities network, traffic data and others helped city authority to take real time decisions while planning new and making adjustments to existing conditions.

Tourism: Many nations have started using Digital Twin as a tool to promote their cities to attract tourists by creating high-quality 3D models supplemented with Augmented Reality and Virtual Reality applications. This virtual tourism also got a boost during COVID-19 pandemic with limited travel options for tourists.

Industry 4.0: Digital twin technology applied to in-use products provide comprehensive insights into usage patterns, workload capacity; a holistic view of the health and performance of equipment enable companies to carry out Maintenance and replenishment of spare parts to minimize time-to-service and avoid costly asset failures.

Manufacturing: Usage pattern and use feedback are incorporated into product designing where Digital twin has a significant influence simulating the next generation of products and aligning them to manufacturing pipeline.

Healthcare: Digital twin technology can be used for pre-operative planning by simulating the medical conditions, taking all medical risks into account and plan for better outcome. In addition to simulating process flows to identify inefficiencies and bottlenecks.

Current challenges for adoption

Three Dimensional drawings are the foundation of Digital twin. Most of the industries are still working on two Dimensional drawings currently and adoption of Digitization is the biggest bottleneck. As number of digital Twins increases; version control and storing the up-to date replica of physical assets in digital format can be being considered as a challenge to address. Finally, as massive amounts of data being collected and utilized in Digital Twin initiatives, it has potential risk for Data security and privacy which needs to be addressed carefully and responsibly.

References: – Gartner’s Top 10 Strategic Technology Trends (2017). Link: https://www.gartner.com/smarterwithgartner/gartners-top-10-technology-trends-2017/

Ashis Panigrahi Economics, Society, Technology

Contact Tracing & Worker Safety: two pillars of safe re-opening of economy

As world is desperately looking for a comeback on it’s economic activities, we all need to play our part by following safe distance measures and quarantine rules imposed by the respective governments. IBM has developed offerings for Worker Insight Solutions which receives inputs from optical and thermal imaging cameras, Bluetooth beacons and mobile phones to provide analytical insights to workers and supervisors for a wide range of critical use cases targeting for COVID-19 situation.

For details please visit LinkedIn page.

 

Ashis Panigrahi Technology

Contactless Transactions

Summary

Now, considering the spreading pattern of COVID-19 where ATM stations including physical cash (paper/plastic notes and coins) will carry the infections up to an alarming number of hrs or even days will expedite each countries journey towards contactless payments. At the same time, we need to plan ahead and try to leverage the existing ATM network by repurposing them to provide new services to mankind.

Looking forward to hear thoughts from all of you. thanks.

For details please visit LinkedIn page 

Ashis Panigrahi Technology

Where to start with Blockchain Journey

There are blockchain platforms available as SaaS options where you can quickly establish and grow your network without worrying much about the underlying infrastructure. But, you need to pay attention into long term plan for Blockchain platform, because once Cryptographic hashes are generated for transactions and blocks, the same can’t be recreated if you change network platform. Hence much attention to be given on deployment aspect of the network at the beginning considering the future plan. It is recommended to perform proof-of-concepts on the same platform where you plan to productionize the network.

Getting it right in terms of use case selection, network participants, agreement between parties, network protocol selection and hosting option means that success is more likely to follow….

For details please visit LinkedIn page

 

Ashis Panigrahi Technology

Centralization vs Decentralization: A Healthcare use case

In a permission Blockchain network, the patient will have control over own medical data and decision on sharing which part of the data with which parties in the network. In addition to data security, data privacy, data immutability, Blockchain solution will also enable ‘right to be forgotten’ where individual’s personal data will be removed from the network when she/he exits the network. This can also incentivise individuals when they are willing to share their medical data (post data anonymization) with research organisations or Pharma companies. The biggest advantage it will provide is the data portability, where individual will have access to their data all the time (changing jobs, changing countries, changing insurance etc..). Decentralised Ledger Technology along with support of decentralised Application (dApps) architecture will encourage individual developers and start-ups to come forward and create patient centric and doctor centric utilities applications for efficiency improvement and better user experiences which will incubate innovation and new revenue models for individuals. #blockchain #dApp #healthcare

For details please visit LinkedIn page  

Ashis Panigrahi Technology

Distributed vs Decentralized Transactions

One good thing in Distributed Apps (dApps) is that, it disperse the economy by incentivising the participants. For example, knowledge management apps like Lunyr, where each participating node is incentivised to contribute articles, and peer-review articles. Using trust engine nature of  blockchain protocol, contents can be kept up-to date access can be controlled if required. 

For details please visit LinkedIn page  

Ashis Panigrahi Technology

Blockchain as Simplified

Blockchain is digitized, is de- centralized, transactions are tamper-proof (meaning if you want to update a record, it is a new transaction and has to be agreed by all parties involved), it enforces the transparencies. Participants must be identified/identifiable, Networks to support high transaction throughput performance and low latency of transaction confirmation. The participants in a network may be competitors in the same industry and may not 100% trust each other; but the defined governance model such as a legal framework to handle disputes can bring required level of trust between participants, Privacy and confidentiality of transactions is the key. If we consider the same supply chain use case we discussed earlier, some suppliers might be given preferred rates as a means of either solidifying a relationship, or promoting additional sales. If every participant can see every contract and transaction, it becomes impossible to maintain such business relationships in a completely transparent network – everyone will want the preferred rates!

For details please visit LinkedIn page