Learn to use AI

AI will be an essential skill for 2024 - here's how to learn it

Hi - 2023 is almost over. It’s the year that AI became mainstream and 2024 will likely see its reach extend further. Learning how to use it properly is going to be an essential skill, so in this newsletter I have collected some great resources to help you do that.

Some resources for learning how to prompt with ChatGPT and other LLMs

According to Google Cloud director Tarek Khalil, AI models are "just scratching the surface" and will soon skyrocket. Speaking recently to Business Insider, he is convinced that "The next few years are going to be the era of AI". In this newsletter edition, I explain how to maximise the value of AI's latest innovations by employing the latest prompt engineering methodologies.

Prompting LLMs

Prompting an AI model like ChatGPT or other large language models (LLMs) involves providing a specific text input or query to generate a response from the model. The prompts should be well crafted for best results, this is known as "prompt engineering", and there are many alternative and often nuanced approaches to the discipline. Making full use of the immense power of LLMs involves getting to grips with the finer points of prompt engineering, and for those new to the subject, the learning curve can be relatively steep. However, many free resources are available to help you with the journey. Here, I briefly review some of the more accessible resources.

Figuring it out for yourself

In "Working with AI: Two paths to prompting", Ethan Mollick argues that the need for esoteric prompting is disappearing as LLMs' abilities to figure out their users' needs are improving rapidly and that "AI will teach you how to use AI". He advocates a conversational approach as long as you also provide contextual information. He develops this into structured prompting with some valuable examples. It is an excellent place to start your journey. The article also references "Prompts for Education: Enhancing Productivity & Learning, " providing useful definitions.

Delving deeper into prompt engineering

For those who wish to get more involved with prompt engineering and indeed might view it as a good career move, an excellent place to start is this YouTube video course "Prompt Engineering Tutorial – Master ChatGPT and LLM Responses". It begins with the basics of AI and moves through the power of prompt engineering, including the nuances of language and language models, using ChatGPT 4 and APIs.

ChatGPT 4 is also the subject of this video, "Learn Prompt Engineering for ChatGPT", which looks at the nuts and bolts of the system and focuses on tokens, which are the fundamental units of text in ChatGPT and other natural language processing models.

A Beginner's Guide to ChatGPT Prompt Engineering from DataCamp provides a brief overview. For those of you who want to immerse yourself in the subject truly, a series of articles is available in Prompt Engineering Guide that explores many aspects of prompt engineering at considerable depth, including basic prompting, prompt formatting, prompt design, different prompting techniques including zero-shot prompting, few-shot prompting, chain-of-thought prompting, generated knowledge prompting, and much more.

Vanderbilt University offers a formal course on Prompt Engineering for ChatGPT, including a free trial and a monthly fee if you wish to continue with the course. However, you can progress at your own pace and receive a certificate on completion.

I also encourage you to look at Prompt Engineering - Mastering Prompt Crafting with Advanced Techniques, which covers many areas already referenced but with a different approach and several examples to work through.

Speed reading

That is a lot of material to wade through, but imagine how much progress you would make if you could take it all in at a rate approaching 25,000 words a minute – the current world record. Why not watch my latest speed reading video, which links to the research paper "So Much to Read, So Little Time?" These resources don't provide a magic speed-reading bullet and highlight the relationship between reading speed and comprehension. Sometimes, you need to slow down to understand.

Finally, you may know about Copilot (formerly Bing Chat), which brings AI powered web searching, but have you tried PerplexityAI? In its own words it’s ‘a chatbot-style search engine that uses AI technology to gather information from multiple sources on the web and provide responses in natural language’. I’ve found it to be very useful and I think you might too. Give it a go!

That’s all for now,

Giles