The Week Ahead

Jon Reades

A few ‘fun’ facts about your professor…

Jon:

• Has a first degree in literature.
• Has never studied computer science and has only ever taken night classes in programming.
• Worked for a dot.com start-up in New York before transferring to London.
• Has taught undergraduate geography at King’s, from which this module evolved, and has taught this module for five years.
• When not working on teaching, being head of department, and squeezing in some reseach, Jon can be found looking after his two young daughters, mountain biking, or hiking in Scotland.

Useful Information

All of the content will remain available to you after our workshop:

1. The main micro-site: jreades.github.io/jaipur/ – talks, applied sessions, and some suggested readings.
2. A quick introduction to Python: jreades.github.io/code-camp/ – 12 self-paced sessions focussing on the basics of Python.

There are plenty of other resources online (Khan Academy, MIT Open Courseware, etc, etc) if you want to go further. And, of course, our Urban Spatial Science MSc and Connected Environments MSc.

What Are We Trying to Do?

This workshop hopes to achieve four things:

1. To expose you to the power of spatial analysis in Python.
2. To expose you to a set of tools that support reproducible analysis.
3. To (hopefully) convince you that going further with this is worth your time.
4. To signpost resources that will support further learning.

This knowledge is intended to be transferrable.

Principles

• Software should be free (as far as practicable).
• Software should be open (as far as practicable).
• Software should run on all platforms (you get the idea).
• Software should reflect what you will encounter in the ‘real world’.

Overall Structure

• Day 1: Getting Started: introducing open source, code and data.
• Day 2: Preparing: setting up, managing, and linking data.
• Day 3: Reproducing: making outputs and analysis reproducible.
• Day 4: Reinforcing: making maps and visualisations.

The Challenges

• To learn a bit of programming and to connect it to the bigger picture.
• To be ok with learning to walk before you run.
• To learn not to rely (too much) on ChatGPT.
• To communicate your thoughts through code and text.

This is a new one for us too. We don’t want to pretend that ChatGPT doesn’t exist. It’s how you will do your work. Unquestionably. But it is also a trap. This year we’re hoping to show you that.

The Rewards

• Skills that are highly transferrable and highly sought-after professionally.
• Problem-solving and practical skills that are valued by the private and public sectors.
• A whole new way of seeing the world and interacting with it.
• Lots of support along the way… if you remember to ask for it!

Getting Help

Hint: There’s lots of help ‘out there’.

Study Aids

When you need an answer right now:

• Google
• Stack Overflow
• Slack

When you want to learn more:

• Medium
• Pocket

Google will become more useful as you learn more and this is definitely one class in which “I Googled it” is a good answer.

As of early September 2020, Stack Overflow contains over 1.5 million Python questions alone! Chances are someone else has had your question before.

Where’s ChatGPT in this? We’ll get to that!

Before You Ask for Help

From the Computer Science Wiki:

• Draw a picture of the problem
• Explain the problem to a toy or inanimate object (really!)
• Forget about a computer; how would you solve this with a pencil and paper?
• Talk it through out loud
• Explain the problem to a friend

To which we would add:

• Use print(variable) statements liberally in your code!

We’ll cover this last bit as we get more used to coding!

When to Ask for Help

• When you get warning messages from your computer’s Operating System.
• When you cannot get the coding environment to run at all.
• When even simple commands return line after line of error code.
• When you have no clue what is going on or why.
• When you have been wrestling with a coding question for more than 20 minutes (but see: How to Ask for Help!)

In order to learn you do need to struggle, but only up to a point! So we don’t think that giving you the answer to a coding question as soon as you get stuck is a good way for you to learn. At the same time, I remain sad to this day that one of the most insightful students I’ve ever taught in a lecture context dropped out of our module because they were having trouble with their computer and thought it was their fault nothing was working right. By we had realised what was going on it was too late: they were so far behind that they didn’t feel able to catch up. We’d rather that you asked and we said “Close, but try it again” than you didn’t ask and checked out thinking that you couldn’t ‘do’ programming.

How to Ask for Help

In addition to what we have provided, we like the “How to ask programming questions” page provided by ProPublica:

1. Do some research first.
2. Be specific.
3. Repeat.
4. Document and share.

If you find yourself wanting to ask a question on Stack Exchange then they also have a guide, and there are plenty of checklists.

There’s also useful ideas on how to get help that covers things like ‘how to get a reply from your Prof’ and ‘where to look for help’.

A Final Note: LLMs

We can’t pretend they don’t exist.

Why Use Them?

Many programmers use LLMs as part of their coding for three reasons:

1. They can help to spot bugs, redundancy, and other issues that impact the performance of large applications (i.e. feedback).
2. They can provide information about different libraries and strategies the developer can use, as well as completing code begun by the developer (i.e. guidance or training).
3. They can help to ‘translate’ code and design patterns between languages (i.e. re-use).

This is very much a ‘brave new world’ and we are all trying to figure it out on the fly.

How to Use THem

LLMs like ChatGPT can help you to learn to be a better coder by providing guidance and feedback:

• Use AI like a patient, but scatty tutor.
• Ask it to explain. Then ask it again.
• Always test its suggestions against reality and logic.

LLMs as co-authors

Using ChatGPT as your co-pilot is not the same as using ChatGPT as your co-author.

How (Not) to Use Them

Here’s what we’ve noticed so far:

• Don’t delegate learning to the LLM, it doesn’t work.
• Don’t let it write paragraphs for you because LLMs like flowery language (e.g. ‘intricate methodologies’, ‘exhaustive reviews’, …).
• Don’t expect LLMs to come up with an argument for you because they strugle with structure/coherence.
• Don’t expect LLMs to get the details right (they don’t ‘like’ referencing!).

Read the Fine Print

Own Your Mistakes

So… Study like you’re learning a new language. Recognise you are all starting at different levels. Talk to other students. Ask for help when you need it!¹

So… we do hope to convince you that:

• Anyone—and this includes you—can code.
• Learning to code does not require mathematical ability.
• Learning to code does not require linguistic ability.
• Learning to code does require practice. And more practice. And more again.

1. Don’t take my word for it, Prat et al. (2020) in Nature link language learning to programming language learning!

One More Thing…

• You will get things wrong.
• We will get things wrong.
• We will assume that you are trying your best.
• Please assume the same about us!
• It’s going to be messy, but we’re really excited about it!

References

Chiang, T. 2023. “ChatGPT Is a Blurry JPEG of the Web.” The New Yorker. https://www.newyorker.com/tech/annals-of-technology/chatgpt-is-a-blurry-jpeg-of-the-web.

Heaven, W. D. 2022. “Why Meta’s Latest Large Language Model Survived Only Three Days Online.” MIT Technology Review. https://www.technologyreview.com/2022/11/18/1063487/meta-large-language-model-ai-only-survived-three-days-gpt-3-science/.

Prat, Chantel S, Tara M Madhyastha, Malayka J Mottarella, and Chu-Hsuan Kuo. 2020. “Relating Natural Language Aptitude to Individual Differences in Learning Programming Languages.” Scientific Reports 10 (1). Nature Publishing Group UK London:3817. https://www.nature.com/articles/s41598-020-60661-8.

Wolfe, C. R. 2023b. “Program-Aided Language Models.” https://cameronrwolfe.substack.com/p/chain-of-thought-prompting-for-llms.

———. 2023a. “Program-Aided Language Models.” https://medium.com/data-science/program-aided-language-models-93d226c7d9a0.