Dreams on demand? ft. Building your prompting skills

✨Welcome to untechnical✨
This is tech content made for the untechnical.

For those who are new here, this weekly newsletter is an easy and approachable way for you to incrementally build your knowledge and skills in all things new tech, AI and automation, while also staying up-to-date with a curated summary of the week’s AI/tech news.

So let’s get into it!

Tech Tip of the Week 🍎 

Prompting Basics (Part 2)

Prompting is the most important skill to learn when adopting generative AI.

Last week, we looked at the fundamentals of prompting generative AI tools, like ChatGPT. If you missed it, I’d recommend taking a moment to revisit what we covered as up-leveling your prompting skills is the most effective way to get better results out of ChatGPT or similar AI tools.

This week, we’re going to build on some of the foundational concepts we covered last week by talking about the format of your prompts.

I know what you’re thinking, surely generative AI is much smarter than worrying about the order we say things in or whether we use headings but you’d be surprised what a difference these little things make.

1. Headings

Use headings (and subheadings) as a way to clearly structure your prompt.

Bonus: if you want to take this up a level, there’s a basic text structure referred to as ‘Markdown’ that helps add formatting to text (also used in HTML code). It’s not as scary as it sounds, I promise. Check out how to use Markdown here.

2. Dot points / Bullet Points

Add dot points to your prompt to help break up text so it’s easier for the AI model to process (and focus) on individual points.

3. Threads

Start a new thread for each new topic or task.

If you are using tools like ChatGPT, when you interact with the tool, you automatically start a thread (or a conversation). The more you chat within the thread, the more context and understanding the model has about you and what you are trying to achieve.

But when you change topic in the same thread, ChatGPT can get ‘confused’ as it may try to reconcile the context of the ‘old’ conversation with the new topic and subject matter. This can reduce the relevance and quality of its outputs quite drastically. So as a general rule, make sure you’re starting new threads whenever you want to change the topic of conversation!

If you’re using ChatGPT or another generative AI tool this week, why not try and implement any or all of these tips? They may be simple, but they work!

Weekly News Digest 🗞

There’s no shortage of news about tech, AI and automation but it’s hard to know what you should tune into. Here, I summarise noteworthy news from the week and why you should care.

1: Anything seems possible with AI - even our dreams aren’t off-limits with Prophetic’s new dream device!

What happened? 

• Tech startup Prophetic has introduced Morpheus-1, an AI model trained on brain data and a wearable headband device to induce lucid dreams on demand. (Yes, you read that right!)

Tell me more:

• Aptly named after the Greek god of dreams, Morpheus-1 uses a halo-like headband device that detects brain waves using ultrasound beams that stimulate specific parts of the brain associated with lucid/conscious dreaming.

• Self-described as the world’s first ‘multimodal generative ultrasonic transformer’, the device uses a combination of machine learning, EEG data, and ultrasound waves to detect REM sleep and induce and stabilise lucid dreams.

Why is this important/interesting? 

• Will this unlock the mysteries of human consciousness? Or is it too early to tell as the startup hasn’t provided tangible proof (yet) that their dream machine works?

• Regardless, this invention is giving 🎬 Inception (the 2010 thriller movie) where dreams could be constructed, manipulated and shared.

2. Pika researchers discover method to create even better generative AI images from text-based prompts.

What happened? 

• Researchers at the idea-to-video startup Pika.Art have developed a new approach to improving images generated from complex text prompts.

Tell me more:

• Simplifying the research paper, the researchers discovered that better image results were yielded by breaking down text prompts into simpler sub-prompts and creating regions within the image to help focus the AI model.

• The end product image is made up of a combination of the smaller ‘regions’ created by the specific sub-prompts.

• This approach has been coined RPG (Recaption, Plan, Generate).

  • Recaption: Transform text prompts into highly descriptive ones.

  • Plan: Partition the image space into complementary subregions and assign different sub-prompts to each subregion.

  • Generate: Diffuse the overlapping regions to create a cohesive image.

Why is this important/interesting? 

• While text-to-image models (like Midjourney and DALL-E) have come on leaps and bounds over the past year, they have struggled with processing complex text prompts.

• This research goes to show that breaking a big task down into smaller, more manageable tasks is not only a winning tactic for humans but generative AI, too!

3. Starting a large-scale AI project? The US Government wants to know

What happened? 

• The Biden administration is readying to use the Defense Production Act to compel companies like OpenAI, Google, Amazon and other up-and-coming AI tech companies to inform the US government when they train an AI model using a significant amount of computing power.

Tell me more:

• These new rules could come into effect as soon as next week and will likely require key information about highly sensitive AI projects inside Big Tech companies to be disclosed to the US government.

• For example, the US government may be the first to know when work or testing happens for OpenAI’s next innovation, GPT-5.

Why is this important/interesting? 

• The rise of OpenAI’s ChatGPT caught the world (and many governments) off guard. As a government, it’s difficult to effectively govern and make laws about technology they don’t know about, yet.

• These changes seem to be a mechanism for governments to keep in the know about what new tech is on the horizon, so they’re not caught unawares, again!

Tech Tool Spotlight 💡

Here, I deep-dive into a selected tech tool to show you what’s possible and ways you can apply the tool.

Pika

Have you ever had an idea in your head about a great video, but not had the technical skills to make it come to life?! Me too…

But with AI, and tools like Pika we’re getting closer to being able to have an idea and turn it into a fully-fledged video with no video editing required.

Pika is positioning itself as an idea-to-video platform, which means you can turn a text-based idea into a moving image within seconds.

Watch what’s possible with Pika in < 1 minute…

Before you go, if you think anyone else would benefit from reading untechnical newsletters, please share far and wide!

Feedback 🎯

This newsletter is a two-way street. Learning more about what you’d like to see more/less of and what content is going to support you best is incredibly important.

💬 Do you have any feedback that may help improve future editions of untechnical? Any questions? Or content requests?

Get in touch by replying to this email.

That’s all for this week!

✨ Thanks for tuning in!