In our latest AI Compass workshop, AI expert Frank Tentler introduced us to helpful use cases of custom GPTs and explained how you can create a custom GPT yourself. An almost fully booked online session that was informative and entertaining and showed how AI learning can be fun. For all those who were unable to attend, here is a brief review of what it was all about and what we learnt.

Are you still prompting or are you already personalising?

OpenAI's GPT Assistant is available to anyone who is willing to set up an OpenAI account. This artificial intelligence (AI) understands and reproduces human language. Personalized GPT models offer customized solutions that adapt to your needs and provide more personalized answers. These models are easy to use and efficient for daily tasks such as creating presentations or automatically analyzing data. They are flexibly customizable, which makes them particularly useful for creating customized learning content, special content bots for social media or helping to solve problems. They also offer a wide range of possibilities in marketing, as they can deliver customized solutions that meet your specific requirements. Some examples of applications of custom GPTs are

1. Specialized content bots to create similar blog posts: A Custom GPT can be used to generate blog posts in a specific style or on a specific topic, based on existing content or specifications.
2. Email outreach bot: A Custom GPT can be used to generate automated responses to emails tailored to the user's individual writing style and requirements.
3. Automated analysis of data: By using a Custom GPT, complex data analysis can be automated to identify trends, recognize patterns and gain insights from the data.
4. Tailored learning content: A Custom GPT can be used to create personalized learning content that is tailored to the learner's individual learning style, knowledge, and interests.
5. Dedicated content bot for social media: A Custom GPT can be used to generate content for social media platforms that considers the style and preferences of the target audience.
6. Problem-solving help: By using a Custom GPT, problem-solving can be automated by the bot providing the user with relevant advice, guidance or suggested solutions.
7. Creating presentations and concepts: a Custom GPT can be used to create presentations, concepts, or reports that meet the user's requirements and address their specific needs.

Overall, custom GPTs are an easy-to-use and effective solution for various use cases. But how do these GPTs actually work and how can they be personalised? This is the question we explored in the last AI workshop with Frank Tentler , after we had already spoken to him about the impact of AI on the future of work.

A review of the AI Compass workshop with Frank Tentler

Frank presented us with impressive content in the online session and showed us in a practical way how custom GPTs work in the real world and what diverse application possibilities they offer. From demonstrating how to create and use GPT instances, to explaining the differences between AI assistants and GPT instances, to tips on how to use these tools effectively and introducing the master GPT and its functionalities, everything was covered. For those who couldn't attend, here's a quick look back at the highlights of the session and some links so you can try out custom GPTs for yourself.

Highlights of the session

The Master GPT

This work "MasterFPT Tentler" by Frollein Flow is licensed under All rights reserved

The Master GPT by Frank Tentler is a custom GPT that you can access here: Master GPT: GPT development made easy.

The "Master GPT" by Frank Tentler is an AI tool that makes it easier for you to develop and use GPT instances. With more than 100 conversations, it offers ready-made prompts and functions for image generation, browsing and data analysis. With this GPT you can learn about

The basics of GPT development:

• Introduction to the creation and use of GPT instances
• Important information about GPT instances and their possible applications

Practical applications and real-world applications:

• Use of DALL-E for image generation
• Data analysis and searches

As a conversation starter there is:

• "Let's develop a GPT!"
• "What do I need to know about GPT instances?"
• "FAQ 'GPT instances': the 15 most important questions"
• "Give me examples of useful GPTs"

These conversation starters serve to arouse interest in addressing exciting and relevant topics that directly introduce the subject of GPT instances. They are designed to promote a more profound understanding of how GPT instances work, their benefits and their potential uses.

For example, the conversation starter "Let's develop a GPT!" encourages active engagement with the development process. Just click on it and try it out. The conversation starter generated the following response for us:

This work "Gesprächsaufhänger" by Frollein Flow is licensed under All rights reserved

A conversation starter like the one in the example "Let's develop a GPT!" actively involves users in the development process, making it easier for them to develop and use GPT instances.

The Master GPT is therefore helpful for developers, companies and interested parties who want to exploit and customise the possibilities of ChatGPT. It is a practical approach that motivates people to actively engage with the topic and helps them to realise their own GPT model more quickly.

On the way to a customised assistant: We are developing "Wolfie"

Of course, we also "lent a hand" in the session and built a custom GPT called "Wolfie" with Frank.

We switched to editor mode and gained our first practical experience with the inputs in the two-table interface. Among other things, we learned that hints or so-called prompts are an important component when using custom GPTs because they give the model context, control the quality of the generated texts, personalise the interaction and improve model performance. They help to achieve more precise and relevant results. This is how our first custom GPT was created during the live session: Wolfie 😄

This work "Wolfie" by Frollein Flow is licensed under All rights reserved

From inspiration to action: enthusiastic participants hit the ground running

An inspiring introduction often leads to immediate action.

• Hilge Kohler showed us how easy it is to access AI by trying it out in practice. She has developed a storyteller that can tell stories on its own. 
• And at FROLLEINFLOW we have not been idle either and have developed the Custom GPT “Future Skills Lara”. Lara identifies key competences for the future and is a practical tool for developing future skills. We present this first version in our EPALE blog. We are already working on the next version in the background, as we believe that this process is not only fun, but also extremely helpful for those interested in future skills! 
• Frank also presents other possible applications.

What we took away from the session

What GPT instances are

GPT instances are specialised, focused AI models that can be trained for specific tasks. They differ from general AI assistants in that they can be created and used individually within a Chat GPT account. To create a GPT instance, precise instructions are given as to which tasks the instance should perform and which data it may use. For example, you can upload files or PDFs that the GPT instance should use as a reference. GPT instances are well suited for specific use cases within a team or company.

The process of creating and using GPT instances involves the following steps:

1. Task definition: defining what specific task the GPT instance should perform, e.g. text editing, proofreading or data analysis.
2. Instruction and training: providing detailed instructions on how the GPT instance should work, including the desired behaviours and criteria. 
3. Data integration: Upload relevant data (e.g. text files, PDFs) to be accessed by the GPT instance to fulfil its tasks.
4. Usage: The GPT instance can then be used within a team account or for individual purposes, drawing on the specific data and instructions.

Advantages and areas of use:

• Specialisation: custom GPTs can be tailored to specific requirements, e.g. for proofreading work or creative projects.
• Efficiency: they provide tailored solutions for specific tasks and can handle multiple data sources and instructions.
• Flexibility: custom GPTs can be continuously adapted and improved to meet changing requirements.

Practical examples:

• Proofreading: a GPT instance can be trained to check texts for spelling and grammatical errors and suggest stylistic improvements.
• Creative projects: A GPT instance such as the Vincent Van Gogh GPT can analyse artistic works and simulate emotional interactions.
• Enterprise applications: GPT instances can be used in enterprises for reporting, analysis and decision support.

Creation and management:

• Create: Users can create and customise their own GPT instances via the GPT Store.
• Manage: GPT instances can be shared and utilised within a team account.

Technical details:

• Interaction interface: Users can enter instructions and train the GPT instance via a user-friendly interface.
• Data limits: Specific data limits (e.g. number and size of uploaded files) can be set to maximise efficiency.

In summary, custom GPTs offer a flexible and specialised way to tailor AI technology to specific requirements and use it efficiently.

Deciphering differences between AI assistants and GPT instances

An AI assistant and a GPT instance are technically very similar, but different in their application. GPT instances are created and used within a chat GPT account, while AI assistants can often be integrated into other systems via APIs. To remind ourselves of this, let's briefly summarise what Frank said about this:

An AI assistant and a GPT instance are technically very similar. Both are focussed colleagues and assistants that can be trained on a specific or multiple task(s). Both can work with private data, but AI assistants can often handle more data, while GPT instances are trained with a limited amount of information that is sufficient for most purposes. GPTs allow the uploading of up to 25 text or PDF files with a limit of 20 megabytes per file, which is equivalent to half a library of materials. For everyday tasks such as writing texts, strategies and concepts, GPTs are often the simpler solution. AI assistants, which are supplied via interface solutions, can incur small charges for each query. For internal teamwork in a chat GPT account, a GPT instance makes more sense, while for external website integration, an AI assistant can be more cost-effective and efficient. Both options offer different advantages depending on the area of application.

ChatGPT naturally also has an opinion and generated the following response:

GPTs and AI assistants specialise in helping with specific tasks. While GPTs are particularly good at writing texts, developing strategies and concepts, AI assistants offer an interface solution for processing large amounts of data.

GPT instances (custom GPTs):

• Function: Custom GPTs are specialised versions of the generative language model that are created and used within a Chat GPT account.
• Application: They are tailored for specific tasks within Chat GPT, such as writing texts or proofreading.
• Data processing: Work with uploaded data (e.g. PDFs, texts) and can access the user's specific instructions.

AI assistants:

• Function: AI assistants are specialised programs that can be integrated into other systems via APIs.
• Application: They offer customised solutions for specific tasks, such as text processing, customer service or data analysis.
• Data processing: Often work with large amounts of data and can solve complex tasks through API calls.

AI assistants are therefore broadly applicable, API-based solutions for various tasks, while GPT instances are specialised models embedded in chat GPT for specific requirements.

Your experiences and questions are welcome!

Have you ever worked with GPTs or are you curious about how they can make your day-to-day work easier?

We look forward to exchanging ideas and are eager to hear your experiences and questions. Let's discover the exciting world of artificial intelligence together and utilise the opportunities it offers!

Who else has tried something after the session and would like to share it publicly? Then please let us know.