Recap of Zooma's generative AI training

At Zooma, the team recently participated in a half-day training session on generative AI. The session included learning and hands-on experience. Together, we explored ways to enhance creativity, improve technical workflows, and gain a deeper understanding of AI's capabilities and limitations. As promised, here is an overview of what the Zoomers learned and accomplished together.

The primary purpose of this training was to explore how generative AI tools like ChatGPT, MidJourney, Bolt.new, and HeyGen can help us in our everyday work. From generating creative ideas to streamlining tasks, we learnt how these tools can improve our work and increase the quality of our work while ensuring we stay aware of their limitations.

AI Act

We first explored Sofie's presentation of the AI Act, a regulation on AI use and supply in the EU and EEA.

Key details include:

• Published in July 2024 and effective from August 2024, with complete application by August 2026.
• Applies to providers, importers, distributors, and deployers of AI systems and general-purpose AI models (GPAIMs).
• Stresses ethical and transparent AI practices.

• Systems considered to present unacceptable risks are prohibited.
• High-risk systems that could impact health, safety, or fundamental rights are permitted to enter the EU market, but they must meet specific requirements.
• AI systems with limited risks due to a lack of transparency must adhere to information and transparency regulations.
• Systems with minimal risks are not subject to any additional obligations.
  

Session highlights

Using Midjourney to generate images

In our first session, Alexander introduced us to generating images with MidJourney using effective prompts crafted with ChatGPT. He shared practical examples and explained how detailed prompts lead to higher-quality visuals. You can also look at his recent article on how his AI-generated images transformed his creative workflow.

At the end of the session, we teamed up to design potential images for our new office in Gothenburg.

These will be voted on, and the winning image will be printed and displayed in the office.

Key takeaways:

• The basics of using MidJourney for image creation.
• How ChatGPT simplifies the process of crafting accurate and creative prompts.
• The importance of detail in prompts to get high-quality results.

ChatGPT practical use cases

In the second session, led by  Stellan, we explored how to use ChatGPT effectively, using 13 practical use case examples. One of the use cases was using Excel to solve problems efficiently.

For example, we explored how to compare names in two columns to check for matches. If you're working in Swedish Excel, you can use the following formula in column G to determine if the first names in columns C and D are identical and display 'Ja' if they are:

=OM(C2=D2;"Ja";"")

This practical example highlighted how ChatGPT can help generate formulas, simplify repetitive tasks, and improve Excel accuracy.

Key takeaways:

• More precise prompts result in improved outputs, and you can ask ChatGPT for help crafting the best prompt. Plugins like 'Prompt Engineer' can also assist.
• Be careful of any errors. ChatGPT may confidently provide incorrect or nonsensical information, such as non-existent Excel formulas, fabricated API endpoints, or made-up legal facts. Its responses are based on its training data and online sources, so it's essential to verify critical information.
• It is not a real-time, up-to-date search engine (yet).
  

Improving technical quality with AI

In the third session, led by Yeu-Kang and Marko, we explored how AI can enhance technical quality across various applications.

They identified six key areas for AI adoption:

• Learning assistance
• Automating repetitive tasks
• Refactoring and documentation
• Code generation
• Creation of new products
• The development of tools and scripts, based on a compelling case study, demonstrated how an accessibility script optimised workflows and improved efficiency.

Key takeaways from the session included:

• Building intuition and critical thinking.
• Crafting specific and detailed prompts.
• Viewing AI as a collaborative partner.
• 'Unfriend Google' and only use it occasionally.

The session concluded with the below quote and an interactive task where we built a functional Todo app using Bolt.new. 

Adopting AI effectively requires balancing its immense potential with its limitations. Success comes from leveraging its strengths while maintaining critical oversight and a commitment to continuous learning. With AI, the saying 'work smarter, not harder' takes on a whole new dimension—enabling us to redefine what's possible.

Translating videos with HeyGen

Our final session, led by Doug, focused on HeyGen, an AI-powered video generator.

We explored its features, including:

• AI-generated avatars and voices for video production.
• Applications for customer support, sales, and internal communication.

Tips for successful translations:

• Use high-quality audio and video inputs.
• Avoid covering the speaker's mouth.
• Be transparent—don't pass AI translations off as authentic recordings.

Key takeaway:

• HeyGen is a cost-effective, scalable solution for producing multilingual video content.

Next steps and reflections

This training highlighted the importance of collaboration, creativity, and critical thinking when working with generative AI tools. As we move forward:

• We'll select the winning office visual on Friday, Fika. Stay tuned! :)
• Continue exploring these tools in our daily work.
• Always approach AI outputs with a healthy balance of enthusiasm and scrutiny.

Thank you to our trainers, organisers, and participants for making this training successful. Let's keep learning and innovating together.