Data Science
Demystifying buzz about AI
Human dream of creating intelligent machines that can mimic human brain is not new. The term artificial intelligence was first coined in 1956 & first neural network (ANN) was invented in 1958. Over last six decades AI has gone through natural evolution with multiple hype cycles. In this blog we try to demystify current buzz about AI and ways in which AI can potentially transform human lives!
Indicators of AI technologies on accelerated growth path!
AI technologies required to store, process and share large amount of data. Impact of AI technologies have increased many fold in recent times due to a)
reduced cost per byte of storage, b) reduced cost per byte of network bandwidth and c) reduced cost per instruction execution by processor. This has been possible due to evolution in telecom technologies (e.g. 5G), storage technologies (e.g. SSD) and AI chipsets (e.g. NVIDIA GPUs).
According to McKinsey report ‘Technology Trends for 2022’, the potential annual impact from AI is $10 to $15 trillion. Following are few indicators of AI technologies on accelerated growth path.
- Improved AI Models Training Speed – 100% improvement in training speed for AI models in year 2021 from year 2018
- Rapid Innovation in AI technologies – 30X increase in AI patents filed in year 2021 from year 2018
- Substantial Public and Private Investment -$100b private investment in AI companies in year 2021
- Market Capitalization of AI Chipset companies – One of leading AI chipset company NVIDIA’s market capitalization touched $400b recently, doubling in last couple of years.
What are the most noteworthy AI technologies?
Wide range of technologies and use cases are clubbed together in umbrella of AI technologies, below is a brief summary on some of most noteworthy AI technologies.
- Computer Vision – Computer vision is a branch of AI/ML that involves machines to process visual data such as images/ videos extracting complex information and produce varied interpretations. This help with wide ranging use cases that impact many industries e.g. autonomous driving cars, facial recognition systems for national security, unmanned quality control for industrial automation, video analytics solution like intruder detection etc.
- Natural Language Processing – Human civilization across world developed spoken and written language over centuries. Natural Language processing is a branch of AI/ML that involves machines to process written and spoken natural language data & produce varied interpretations. This help with various use cases like virtual voice assistant (chatbox) or virtual voice assistant (Alexa/Siri) using automated voice recognition and response systems (Conversational AI) among many others.
- Deep Reinforcement Learning – These technologies a combination of deep learning and reinforcement learning and involves machine/agent learns to make decision based on either live or historical data in different environments using algorithms inspired by brain neural networks. This helps to build machines/robots that can learn & perform human tasks (using robotic-arm motion control). Such machines/robots improve process efficiencies like automating, manufacturing line or can be deployed in environments risky for humans e.g. Fire/flood & help save human lives.
- Applied AI: Machine Learning, Deep Learning – Machine learning (ML) is a branch of AI that uses statistical techniques to enable machines to "learn" from data, a process known as “training” a “model”. Deep Learning (DL) is an area within ML that mimic layers of neurons in human brain to learn complex patterns in data. The “deep” refers to the large number of layers of neurons in contemporary models that help achieve better performance gains. ML/DL algorithms are being applied across industries in different use cases to both increase revenues and decrease cost. Some example use cases are Fraud Detection in Financial industry, product recommendations in Retail Industry, predictive maintenance in Manufacturing Industry, path/schedule optimizations in logistics industry.
New Frontiers for AI technologies
While evolution is taking place a fast pace in all technologies listed in the previous sections. AI world has open up to new frontiers being addressed with high energies by stakeholders in both academics and industry. We expect fast paced movement resulting in creation of new AI technologies for these frontiers.
- Going beyond Correlation & finding causes (Causal AI) – Machines learning and deep learning algorithms learn to find correlation between input and output variables in historical data. Models learnt finding these correlations then apply the same correlation for predicting outcomes for new set of input data. This approach has shortcoming that target outcome would be limited to parameters for which training data is available. Causal inference methods enable modelling of counterfactual or “what-if” scenarios for which training data isn’t (or can’t be) available, but these scenarios that are essential for scientific experimentation or business decision-making. Some use cases of causal AI include dynamic pricing for taxi service, discount for customer acquisition/retention etc.
- Interpreting results of ML algorithms (Explainable AI) – To drive confidence in business to trust outcome of AI algorithms, the results need to be interpretable. For example when a bank rejects a loan to customer due to low credit rating, bank employee should be able to explain to customer the reasons for his/her low credit rating. This is needed both for transparency & sometimes mandated by law of land. Explainable AI (XAI) is a branch of AI that deals with interpreting results of AI. Different algorithms and practices help AI engineers build solutions to help explain /interpret results.
- Building safeguards from AI deployments (Responsible AI) - Experts and common public alike suspects potential catastrophic risks with wide spread adoption of AI technologies. Responsible AI intends to bring required policy safeguard to ensure AI technologies abide laws, incorporate ethics and implement technical and social robustness to mitigate potential harm. Responsible AI tenets include human agency oversight; societal and environmental well-being; technical robustness and safety; privacy and data governance; transparency; accountability; and diversity, non-discrimination and fairness
We at IKTARA have built competencies to help enterprises accelerate their AI journey. We provide consulting, design, development, deployment & operation services to clients worldwide. Please get in touch at info@iktara.ai for any queries or suggestions, we would be happy to engage with you.