As of a month ago, I’ve been considering doing a Masters in Artificial Intelligence, because I just loved being a student so much the first time. The thought of, once again, not having any money, studying for exams, and living in student accomodation (FYI, worse than a van) was something I just couldn’t resist.
Not quite. I want to do a postgraduate degree for 2 reasons:
- Artificial Intelligence will soon be prevalent in nearly every area of our lives and so being equipped with fundamental understanding in these technologies is super important to me. Oh, and I like it.
- Having a Master’s degree boosts credibility (help me get a job).
The downsides:
- Costs £20K
- Not being paid at the same time
- Move away from my lovely girlfriend (it’s only a year but still)
So, while I am filling out applications. I’m also looking at an alternative option.
Creating an MSc for myself.
How? Using AI, books, and projects.
I’ve found the units from all the best Master’s courses in the UK that interest me the most:
MSc Advanced Computing at Oxford
- Bayesian Statistical Probabilistic Programming (Oxford)
- Computational game Theory (Oxford)
- Concurrent Algoritms and Data Structures
- Graph Repsentation Modelling
- Machine Learning
- Quantum Processes and Computation
- Quantum Information
- Foundations of self programming agents
Didn’t apply because 1) Don’t meet the academic requirement 2) I could only provide 1 academic reference and they wanted at least 2
MSc Compututing (AI ML stream) at Imperial:
- Mathematics for Machine Learning
- Probabilistic Inference
- Reinforcement Learning
- Knowledge Representation
- Statistical Information Theory
- Software Engineering for Machine Learning Systems
- Distributed Algorithms
- Computational Finance
- Cryptography Engineering
- Principles of Distributed Ledgers
- Quantum Computing
Applied – got rejected (didn’t meet the entry requirement)
MSc Aritificial Intelligence at Edinburgh
- Accelerated Natural Language Processing
- Methods for Causal Inference
- Applied Machine Learning
- Computational Cognitive Neuroscience
- Algorithmic Game Theory and its applications
- Reinforcement Learning
- Computer Programming for Speech and Language Processing
- Blockchains and Distributed Ledgers
- Introduction to Quantum Programming and Semantics
Applied – still waiting. This course looks incredibly interesting and I spoke to a couple of people who had done it and it seems to attract an interesting and entrepreneurial stream of people so fingers crossed for this one.
I also applied for UCL Statistical Computing and Machine Learning – this course is a bit more specialised and wouldn’t allow me to study modules in wider areas of Computing but seems like a very rigourous course. Still waiting.
Now I didn’t want to completely outsource my education and rely completely on me getting into these courses to study topics that I want to learn about. So I’ve created my own curriculum.
ChatGPT aggregated the above list to come up with a list of 7 topics that cover most of my interests along with some elective units. I will do a project in each of them – it’s the best way for me to learn and develop practical skills.
This is ChatGPT’s course:
MSc in Artificial Intelligence and Computer Science
Core Units
- Foundations of Machine Learning:
- Resources: Pattern Recognition and Machine Learning by Christopher Bishop; Deep Learning by Ian Goodfellow et al.
- Project: Image classification using CIFAR-10 with a deep neural network.
- Mathematics for Machine Learning and Probabilistic Inference
- Resources: Mathematics for Machine Learning by Marc Deisenroth et al.; Bayesian Reasoning and Machine Learning by David Barber
- Project: Bayesian classifier for stock price trend prediction.
- Reinforcement Learning
- Resources: Reinforcement Learning: An Introduction by Sutton and Barto; OpenAI Spinning Up
- Project: Lunar Lander
- Quantum Computing and Quantum Information
- Resources: Quantum Computation and Quantum Information by Nielsen and Chuang; IBM Quantum Experience
- Project: Implement the Deutsch-Jozsa algorithm on a quantum simulator.
- Statistical Information Theory
- Resources: Elements of Information Theory by Cover and Thomas; MIT OpenCourseWare lectures
- Project: Analyze information gain in a dataset using information theory.
- Knowledge Representation and Computational Cognitive Neuroscience
- Resources: Artificial Intelligence: A Modern Approach by Russell and Norvig; Computational Neuroscience by Dayan and Abbott
- Project: Develop an ontology or knowledge representation system for medical diagnostics.
- Concurrent Algorithms, Distributed Systems, and Blockchain Technology
- Resources: Distributed Systems: Principles and Paradigms by Tanenbaum and Van Steen; Mastering Blockchain by Imran Bashir
- Project: Create a simple blockchain with transaction verification and consensus.
Elective Units
- Natural Language Processing
- Resources: Speech and Language Processing by Jurafsky and Martin; Hugging Face NLP Course
- Project: Build an LLM twin to create content.
- Algorithmic Game Theory
- Resources: Algorithmic Game Theory by Nisan et al.
- Project: Simulation of a multi-agent auction, analyzing outcomes using game theory.
- Software Engineering for AI and ML Systems
- Resources: Designing Machine Learning Systems by Chip Huyen; Coursera MLOps Course
- Project: Deploy a machine learning model using Docker and Kubernetes.
Capstone Project Options
- Healthcare Predictive Modeling: Bayesian predictive model for health outcomes.
- Quantum Cryptography: Develop a quantum key distribution protocol.
- Multi-Agent Simulation: Build a market simulation with game-theoretic strategies.
- NLP for Social Good: Create a legal or medical document summarizer.
Based on all of the above. I’ve come up with 4 projects that I can complete that combine multple modules and should give me a solid grounding in Artificial Intelligence, and Dsitributed Computing (blockchain).
MSc Artificial Intelligence By Me
- Build and deploy LLM twin to write content – This will teach me software enginering for machine learning systems and integrate aspects from NLP.
- Build a blockhain for facilitating trade and investment in decentralised communities – this will teach me about ditributed algorithms and blockchain technologies
- Build a reinforcemnt learning agent to trade cryptocurrency futures – this will likely also involve a lot of data analysis and I can explore statistical information theory.
- Build a Bayesian Classifier to predict what weight you should train at next based on past performance
These are small-scale projects with the most practical applications I can think of. Units that weren’t inlcuded were too theoretical to do a feasible project on that would create value.
I’ll start with the first one – building and deploying an LLM twin, we’ll see how this goes. I’ll also be writing posts explaining interesting concepts I come across.