Conversations about AI: Ed 3

Hello Sisters! Hope your week is going swimmingly.

This week we have a packed newsletter designed to help you have those early conversations with colleagues about using generative AI in your organisation.

But first... In the News

 Things you may have missed and some you couldn’t have missed even if you had tried.

International Women’s Day was on Friday last week. I’m not sure what side of the fence you are on but opinions ranged from 'Hate it' to 'Love it' with lots of shades in between. Turns out that the phrase ‘Happy Womens Day! is a pet hate of many women here on LinkedIn, but perfectly acceptable to many others. This just goes to show that we may be grouped into a category entitled ‘women’ but we are not as predictable as ChatGPT may think we are. More about this in the section on the power of prompt engineering. This was the Emmeline community's IWD post

Other news published the day after last week’s newsletter extolling the virtues of Gemini Advanced's text to text chatbot capabilities, a freezing cold pale of water was poured over Gemini Advanced's text to image agent. As Sundar Pichai, CEO of Google apologised to the world for offensive images generated by GA, analysts called for his head on a platter.

Worth mentioning that although not all agents are chatbots, all chatbots are agents. If you would like to know the reason why, in plain English, you will be very welcome to join one of my collaborative workshops and all will become clear. Details to be published shortly.

Buckingham Palace published a photo of our future queen and her happy children in their garden on a sunny day. Unfortunately this photo didn't help celebrate Mothering Sunday so much as stimulate the ongoing debate about Kate's health. This is of concern to people who care about Kate as well as the photo agencies who question the sincerity of the image.

Update: Buckingham Palace remained silent last night after Getty, Associated Press, Reuters and AFP removed the image and advised clients to remove it from all their platforms.

This story has made it to our newsletter because it highlights the general public’s mistrust of all things ‘new fangled’. So, if you are struggling to come to terms with this new technology which is capable of predicting the edges of photos, glorifying shrubs with cloaks of abundant greenery and applying wonky fingers to miniature princes, then imagine how employees feel when they read that a 'robot won’t take their job but their employer remains Stoically silent on the topic. Are you ready to start talking about generative AI? We'll make some progress towards that goal today - if you are in a hurry, find your tribe here.

Today’s theme

This week’s theme is designed to help us start conversations with our colleagues to explore how we as female leaders, innovators and influencers move forward in the new world of work.

Talking with Tech – Discussing strategy

Get on top of the tech Jargon - You are equal

Whatever domain you work in whether it be HR, legal services, finance etc and whatever sector you work in, the research says most businesses will need a Generative AI strategy in place by 2026. This will involve a few stages which we can discuss in future editions of this newsletter, but first it’s important to build your confidence and to realise that you are an equal in this new world of work. You absolutely do not need to be a tech-expert, futurist or whatever, it is enough to know the basics and that involves some jargon busting. 

Of course, I turned to my chatbot buddy ChatGPT 4 and asked it to translate some well known tech references because I wanted short and sweet, plain English responses. I also got a bit bored asking straight questions so I mixed it up as you will see.

I’ve variously used a Knock Knock joke format, children’s stories and added personas for a twist. Why not try creative approaches like this yourself? As long as you don’t share private information, or ask offensive questions you will be fine.  

In no particular order here are some jargon-busting explainers:   

Me to ChatGPT: What is an LLM (Large Language Model)? Tell me in the format of a joke

ChatGPT 4: 

Knock, knock.

Who's there?

LLM.

LLM who?

"LLM-believable" how much I know, ask me anything!

Me to ChatGPT: What is an LLM Hallucination?

ChatGPT 4:  An LLM hallucination is when the magic book gets a little silly and tells you a make-believe story as if it were true, like dragons doing homework.

Me to ChatGPT: What happens if Training Data is biased?

ChatGPT 4:  If the wizard's book only learns tales from one village, it might think the whole world is just like that village, missing out on the vastness of other stories.

Me to ChatGPT: I prompted further, and this was the response – quite advanced for a 5 year old I thought:

ChatGPT 4:  ‘If the wizard's book (LLM) only learns tales from one village, it might not only miss the vastness of other stories but also start telling tales that only fit that one village's view, making others feel left out or misunderstood. As these tales spread far and wide, they could shape how people from different villages see each other, possibly leading to misunderstandings or even making some feel their stories aren't as important. When everyone's tale isn't considered, it's harder for all the villages to understand and appreciate each other's uniqueness, which is essential for a harmonious world.’

Me to ChatGPT: What is the benefit of Fine-Tuning data - respond in the style of a pessimist.

ChatGPT 4:  Sure, fine-tuning data might make responses more accurate, but let's be real, it'll never be perfect. There's always something that slips through the cracks.

Me to ChatGPT: What is the chatbot’s relationship with the LLM? – tell me in the style of a matchmaker:

ChatGPT 4:  The chatbot, a charming communicator, seeks the wisdom of the LLM, a deep thinker with vast knowledge. Together, they're a perfect match, bringing out the best in each other.

Me to ChatGPT : How does a GPT work?

ChatGPT 4:  GPT processes vast amounts of text to learn language patterns, enabling it to generate responses that can mimic human-like understanding. While it's a breakthrough in AI, enhancing communication and creativity, it doesn't truly "understand" in the human sense and can sometimes produce errors or biased content, reflecting its training data. This technology presents both exciting possibilities and challenges that need careful management.

Me to Me : What is Prompt Engineering?

I’ll answer this one … in plain English, prompt engineering is the skill of asking great questions of AI agents.

Talking with chatbots - Prompt Engineers and Prompt Userneers

Prompt Engineers usually have coding skills and develop AI behind the scenes, shaping responses with their questions. Users like you and I interact directly with AI during our working day.

Both roles are crucial; and interestingly they rely on each other for AI to learn and improve.

Engineers tweak AI based on user interactions, and users adapt their questions as AI evolves. It's a shared effort to enhance AI communication.

So, I’m inventing a new phrase just for users like you and me: ‘Prompt Userneers’.

You are a Prompt Userneer so don’t be intimidated by your co-worker in the tech department. However, it will come as no surprise that to go from amateur to effective Prompt Userneer, you need to invest time in refining the skill.

How to level up your Prompt Userneer skills

I’ve written a book to hone your Userneer skills. This is for people who like learning through stories woven to illustrate the techniques required.

If you like reading human stories and want an unfair advantage at work, it pays to buy my book ‘The ChatGPT Advantage’.

Why ‘Prompters’ are power players in AI circles

Prompt ‘Engineers’ are power players because:

beware - this is written by a chatbot which presumably has integrated bias.

They steer the training data in a few key ways to improve AI performance and relevance:

Selective Data Inclusion: They carefully choose which datasets to include in the training process, focusing on high-quality, relevant, and diverse data sources to reduce biases and enhance the model's understanding.

Data Cleaning and Pre-processing: By cleaning and pre-processing data, prompt engineers ensure that the training data is consistent, free of errors, and formatted in a way that's most beneficial for the AI to learn from.

Feedback Loops: They use feedback from AI performance to identify gaps or biases in the training data. This involves analyzing instances where the AI's responses were incorrect or not optimal and then adjusting the training data accordingly.

Augmentation Techniques: Prompt engineers may augment the training data with synthetic or semi-synthetic examples to cover rare or underrepresented scenarios, helping the AI to generalize better across different contexts.

Curating Scenario-Specific Datasets: For AI models designed for specific tasks or industries, they curate specialized datasets that reflect the unique challenges and language of those areas, improving the model's effectiveness in those contexts.

Regular Updates and Re-evaluations: Training data is regularly updated and re-evaluated to incorporate new information, trends, and feedback, ensuring the AI remains relevant and effective over time.

Through these strategies, prompt engineers shape the development and ongoing learning of AI models, guiding them and it is down to the engineers whether or not the data has the capacity to be accurate, fair, and aligned with user needs.

Prompt ‘Userneers’ are power players because:

Beware - this is written by a chatbot which presumably has integrated bias.

In contrast to prompt engineers who focus on the technical side of training AI, Userneers engage with AI systems from a user perspective. Their role involves:

Crafting Effective Prompts: They develop and refine the prompts they use to communicate with AI, aiming to get the most accurate and useful responses. This requires understanding how to phrase questions or commands in a way that the AI can best interpret.

Feedback Provision: Prompt Userneers provide valuable feedback on the AI's performance, noting where it excels and where it struggles. This user feedback is crucial for further refining AI models.

Adapting to AI Capabilities: They adapt their queries and interactions based on their understanding of the AI’s strengths and limitations, optimizing how they use the AI for various tasks.

Exploratory Interactions: By experimenting with different types of prompts and observing the outcomes, Prompt Userneers contribute to discovering new uses or uncovering limitations of the AI systems.

Sharing Best Practices: They often share insights and strategies for effective AI interaction within their communities or organizations, helping others to improve their own AI engagements.

Prompt Userneers essentially bridge the gap between the AI's technical capabilities and the practical needs or goals of end-users, effective Userneers will ensure that AI tools are used as effectively as possible in real-world scenarios.

Identify tools that appear to solve your problems

What are 'suitable' problems

Well, partnering with AI co-pilots doesn't mean giving up the work you're passionate about. Instead, think of it as an opportunity to offload the tedious, monotonous tasks that drain your energy. This way, you can focus more on the meaningful, creative aspects of your work—those areas where you truly shine and make a unique impact. Embrace this collaboration as a chance to do more of what you love, enhancing your own value and contribution.

Researching solutions

As 'Hybrid Human-AI working' is going to be the norm I advise getting to know more about it. Not ‘tech for tech’s sake gadgets’, but ‘knotty problem-solving co-pilots’.

You could work top down or bottom up. You can explore the amazing resource that is Linkedin and subscribe to newsletters like this one for the low down on Generative AI developments from a user perspective. Or, you could identify the problem and conduct targeted research. I like to adopt both methods simultaneously, but it depends how much time you have available.

Talking with L&D – To get your team GenAI ready

Starting the conversation

When talking with L&D (Learning & Development) about upskilling your teams it pays to have an understanding of the learning process per se. The first thing L&D usually focuses on, is assessing the current skills status before getting into planning stage.  Very sensible!

In the context of investing in (L&D) for working effectively with AI co-pilots, it's essential to focus on power skills that are both foundational and observable in the early stages of the learning journey. These skills not only facilitate smooth adoption and interaction with AI technologies but also can be assessed relatively early to guide further training and development efforts. Key skills to assess first include:

Fundamental Power Skills

1. Adaptability: This skill is crucial in the rapidly evolving AI landscape. It involves being open to learning new technologies, adjusting to changing work processes, and adapting strategies based on AI feedback and developments.
2. Critical Thinking: The ability to evaluate AI-generated content, make informed decisions, and apply logical reasoning to solve problems. Critical thinking helps in discerning the accuracy and relevance of AI outputs.
3. Communication: Effective communication skills are essential for interacting with AI tools, specifying requirements, and interpreting AI-generated insights. Clear communication helps in refining AI outputs and collaborating with team members on AI-related projects.
4. Ethical Understanding and Bias Awareness: Recognizing the ethical implications of AI applications and being aware of potential biases in AI models are fundamental. This skill is vital for ensuring responsible use of AI technologies.
5. Collaborative Mindset: Willingness to collaborate is key, especially as AI co-pilots often augment team capabilities. Collaborating effectively with both AI tools and human colleagues ensures a synergistic work environment.

Assessing current ‘Power Skill’ status – Team Readiness

Focusing on these skills early on can help L&D teams identify existing competencies, pinpoint gaps, and tailor training programmes that align with the needs of working alongside AI. These skills not only serve as a foundation for more advanced technical learning but also ensure that employees are prepared, responsible, and adaptable users of AI technology.

More on these skills in future newsletters –

Good News on the Assessment Front

Emmeline’s immersive ‘Power Skills Assessment’ platform is launching at the end of this month. More details next week.  I will keep all sisters informed and offer some early seats to a live demo to avoid FOMO syndrome.

Next steps, Sister

Congratulations! You are embarking on this journey as a GenAI Userneer.  I recommend cultivating conversations with colleagues and using your Power Skills to make those conversations count.

And before I go…

Embedded gifts:

I’ve been using Miro boards this week and notice that it the Miro team has added a GenAI chatbot for use when inspiration fails us. This triggered a thought… it’s worth checking out our usual online tools to discover newly embedded GenAI agents that can help you nail that task faster, better, more efficiently. 

New tools:

And I recommend taking a free spin on SparkItUp.ai –  this tool helped me write a training programme in half the time and with better results than I would have got.. (wait for it… ) using  ChatGPT 4! Which is pretty awesome.  SparkItUp.ai elegantly helped me stay focused  – it’s so easy to get distracted by shiny things.

Tip… click each icon at least once to make sure you don’t miss the point.

HELP!

If you have enjoyed this edition, please share it, like it or leave a comment. It energises me.

Until next time sisters, peace and prosperity!

Carolyn S.