Module S – Interactive Quiz – 20 Multiple-Choice Questions
Click an answer to check. Your score is tracked below.
0 / 20 answered
Q1Understand
What does the “Convergence Thesis” claim about AI and blockchain?
Explanation
The thesis holds that AI and blockchain solve different problems and reinforce each other: AI adds intelligence, blockchain adds verifiability and decentralized coordination.
Q2Understand
What coordination problem does decentralized compute aim to solve?
Explanation
Decentralized compute networks (Akash, Render, io.net) aggregate idle GPU supply globally to reduce dependence on AWS, Azure, and GCP.
Q3Understand
What is Zero-Knowledge Machine Learning (zkML)?
Explanation
zkML uses zero-knowledge proofs to prove model execution integrity – enabling verifiable inference while preserving privacy of inputs and proprietary model parameters.
Q4Understand
What distinguishes an on-chain AI agent from a standard smart contract?
Explanation
Smart contracts are deterministic and rule-based; AI agents add planning, language understanding, and adaptive decision-making on top of blockchain execution.
Q5Apply
AWS charges $32/hr per GPU; Akash charges $5/hr. A training run requires 200 GPU-hours. What is the cost saving using Akash?
A tokenized inference service charges 0.5 tokens per query; the token price is $2.00. What is the total cost for 1,000 queries?
Explanation
1,000 × 0.5 × \2.00 = $1,000$ total query cost.
Q8Apply
A project description reads: “Our platform uses AI to score DeFi loan applications and records each decision on Ethereum with the model hash, input hash, and zkML proof.” Which dimension of the Convergence Thesis does this BEST exemplify?
Explanation
Recording the model hash, input hash, and zkML proof on-chain makes each AI decision auditable and tamper-evident – the core of verifiable inference.
Q9Apply
A content creator registers a cryptographic hash of their artwork on Ethereum. A rival platform later displays identical content without any on-chain attestation. What does this indicate?
Explanation
The on-chain timestamp precedes the rival’s use; the creator can demonstrate prior existence of the content via the immutable hash record.
Q10Apply
An AI oracle feeds a Bitcoin price of $50,000 to a DeFi lending protocol. The true market price is $40,000. What is the most likely consequence for the protocol?
Explanation
An inflated price oracle allows borrowers to extract more value than their collateral is worth; if prices correct the protocol accumulates bad debt, as seen in the Mango Markets exploit.
Q11Apply
A decentralized compute network has 100 providers; 80 run NVIDIA A100s and 20 run older NVIDIA V100s. One provider uses custom TPUs. Which risk does this illustrate?
Explanation
Concentration in a single hardware vendor creates correlated failure risk; a firmware bug or supply-chain attack can cascade across the majority of providers.
Q12Apply
A DAO proposes delegating treasury management to an autonomous AI agent. Which evaluation question is MOST critical before deployment?
Explanation
Human override capability and clearly defined decision boundaries are the primary safety requirement for any autonomous agent managing real assets.
Q13Analyze
Why is AI hallucination MORE dangerous when an AI model is connected to a smart contract than on a centralized chatbot platform?
Explanation
Centralized platforms can roll back incorrect outputs; blockchain transactions are immutable, so a hallucinated decision executed on-chain may permanently alter state or move funds.
Q14Analyze
Compare centralized AI (e.g., OpenAI API) and decentralized AI (e.g., Bittensor). What is the KEY structural difference regarding censorship resistance?
Explanation
A single company controls access to a centralized API and can block users or alter outputs; decentralized networks distribute control so no entity can unilaterally censor.
Q15Analyze
Fetch.ai, SingularityNET, and Ocean Protocol merged into the ASI Alliance. What strategic problem does this merger address?
Explanation
Fragmented decentralized AI projects struggle to match the network effects of OpenAI or Google; pooling tokens, users, and developer ecosystems creates a more credible alternative.
Q16Analyze
GPU manufacturing is dominated by NVIDIA (>80% data-centre share). How does this concentration affect decentralized compute networks?
Explanation
When one supplier dominates, geopolitical restrictions (e.g., US export controls on A100s to China) or supply shocks propagate across the entire decentralized ecosystem simultaneously.
Q17Analyze
A zkML proof is generated for a credit-scoring model that returned “reject” for a loan applicant. Why is this useful to the applicant even without revealing the model?
Explanation
zkML separates execution integrity from secrecy: the applicant (or regulator) can verify which model ran and that it ran correctly, without the bank exposing its scoring algorithm.
Q18Analyze
Blockchain content provenance works well for newly created content. Why does it NOT solve the problem of existing deepfakes already circulating online?
Explanation
Provenance is prospective: registering future content helps, but the vast corpus of already-circulating manipulated media has no prior on-chain record to invalidate.
Q19Evaluate
A project claims “AI-powered blockchain analytics with our governance token.” Apply the 5-question evaluation framework. What is the MOST appropriate assessment?
Explanation
The 5 questions are: Does AI add value here? Does blockchain add value? What does the token incentivise? Is decentralisation necessary? What is the real cryptoeconomic model? None are answered by the claim.
Q20Evaluate
A government proposes requiring all AI models used in financial services to publish zkML proofs on a public blockchain. Evaluate this proposal on privacy, cost, and feasibility.
Explanation
zkML offers meaningful transparency benefits but faces three constraints: metadata leakage from on-chain records, high proof-generation cost for large models, and lack of standardised tooling across neural architectures – making broad mandates premature.