11th December 2025

Explanation - It was established in 1957 as Non-Governmental Organisation (NGO). It was officially changed its status from a Non-governmental Organisation (NGO) to an Intergovernmental Organisation (IGO) based on a Convention ratified by 34 States in 2024. Its mandate is to harmonise global maritime navigation systems, promote maritime safety initiatives, and collaborate with member states, international organisations, and industry stakeholders to address emerging challenges in maritime safety and environmental protection. Members — It comprises of 200 members, 80 of which are national authorities and 60 are commercial firms. India is a member of this organisation since 1957. Headquarters — Saint Germain-en-Laye, France. 

Explanation - Q-day is the future moment when a powerful quantum computer can break today's commonly used encryption systems. This doesn't mean data will be exposed instantly — but anything stolen and stored today could be decoded later once such a machine exists. This threat is called "harvest now, decrypt later." How Are Governments Preparing? Countries are already working on protections. The U.S. National Institute of Standards and Technology (NIST) has approved new post-quantum cryptography (PQC) methods designed to stay secure even against quantum computers: CRYSTALS-Kyber → for encryption, Dilithium → for digital signatures. These rely on tough mathematical problems that quantum computers are not expected to crack (not easy for them to solve). Experts believe breaking RSA-2048 — a widely used encryption standard — will require millions of stable (logical) qubits. RSA encryption works by multiplying two huge prime numbers. Multiplying them is easy. But figuring out the original primes from the final product is extremely hard — so hard that even supercomputers would need billions of years. At current progress, this may take 5 to 8 years, so Q-day is still a future risk, not an immediate one. 

Explanation - Machine learning — A subset of AI where data is fed into a program so it can identify features in that data. Deep learning — Trains itself to recognise patterns without human intervention (not requiring manual labelling). Neural networks — Constructed of connected network nodes composed of several layers that pass information between each other. Transformer models — Learn context using a technique called self-attention to detect how elements in a sequence are related. 

Explanation - LLMs are the "foundational powerhouses" behind tools like ChatGPT, Claude, Gemini, etc. LLMs are now becoming "multimodal (working with media types beyond text)", which means they are not limited to text only. After initial training, LLMs are further trained via "tuning" (fine-tuning or prompt-tuning) for particular tasks. 

Explanation - LLMs are built on a type of neural network called a transformer model. Deep learning involves the probabilistic analysis of unstructured data, which is a method used by LLMs. Many LLMs are trained on internet data, programmers may use a "more curated data set" for better quality learning, implying that exclusive reliance on uncurated data is not a universal rule.