Contact

Knowledge Transfer in the Field of Artificial Intelligence

KNOWLEDGE TRANSFER | KEYNOTE SPEECHES | TRAININGS & WORKSHOPS

AI in Practice

Opportunities for You as a Person & Your Business

Artificial intelligence offers numerous possibilities for companies. This presentation shows you the risks and challenges of working with AI technologies as well as the opportunities and practical applications.

What topics will we cover today?

Terms

AI, LLM, Prompt

Risks

DeepFakes & Digital Violence

Challenges

Regulation & Data Protection

Opportunities

Practical Applications

Tools

AI Systems in Use

BOBBY and the AI BOX

LENA - Learning Engineered Neural Assistant

An advanced AI system specifically developed for learning processes that functions as an intelligent assistant.

Daniela Purps - founder AI Inside

Rob Galler - Motivation - Ideology - Kinside

Change: Learning & Knowledge

Limited KNOWLEDGE

LEARNING BY HEART

In the era of encyclopedias and reference books, storing knowledge was the priority – learning meant retaining facts permanently in one's mind.

1985: Not always available

Unlimited KNOWLEDGE
FINDING PURPOSEFULLY

With the internet, the focus shifted: no longer knowing everything, but knowing where and how to find it – research became the core competency.


1997: Always available

Define RESULTS
SHAPE THEM PURPOSEFULLY

With AI, a new phase begins: learning now means generating precise results through good prompting – the human transforms from a knowledge consumer into a knowledge creator.

2021: Part of society

25 Years of Change in Fast Forward


Technological Evolution

The development of knowledge sources has changed dramatically — from physical encyclopedias to the internet, all the way to AI systems that are becoming part of our daily lives.

New Forms of Interaction

Modern technologies such as smart glasses and augmented reality are fundamentally changing how we interact with information and integrate it into our everyday lives.

Future Perspectives

AI will increasingly be embedded in our environment, continuously supplying us with relevant information without us having to actively search for it.

Terms & What is AI Actually?

AI

Artificial machines that solve tasks that would otherwise require human intelligence – e.g. understanding language, recognizing images, learning from data, and supporting decisions

LLM

Large Language Model: trained on vast amounts of text, generates and understands language – answers questions, writes/summarizes/translates, and is the foundation of many chatbots.

PROMPT

An instruction or request given to the AI

Generative AI

AI that generates content such as text, images, video, audio, or software code in response to a prompt.

3 Types of AI

ANALYTICAL AI

Analysis & Planning

  • Analyze data, recognize patterns, gain decision-relevant insights
  • Process large datasets to predict outcomes or classify information
  • No content generation — pure analysis and recognition

GENERATIVE AI

Creativity & Sparring

  • Generate new content (text, image, audio, video) from learned patterns
  • Output-oriented: respond to prompts with original, synthesized results
  • In contrast to analytical AI: creation, not classification

RPA*

Execute & Automate

  • Software robots that automate repetitive, rule-based digital tasks
  • Mimic human system interactions via predefined workflows
  • No AI or machine learning — pure rule-based execution.

*Robotic Process Automation

AI in business requires the combination of analytical & generative AI.

Well-Known AI Variants

Assistant

  • Conversational AI systems that respond to natural language inputs (text/voice)
  • Task-oriented: answering questions, executing defined tasks on user request
  • Operate within fixed parameters — reactive, not autonomous

Tools: ChatGPT, Gemini, Claude

Avatar

  • Visual or voice-based digital personas that represent AI systems
  • Human-like, engaging interaction interfaces
  • Use cases: customer service, education, virtual environments


Tools: Heygen, Synthesia

Agent

  • Autonomous AI systems that independently execute complex tasks
  • Goal-oriented: decision-making, tool/API interaction without constant human input
  • Chain actions across workflows to achieve defined outcomes

Tools: Manus

ROBOTS - The New Companions?

ROBOTS - the new companions?

South Africa is a leading robotics market in Africa, but not a top-ranked country globally.
The Robotics Market in Africa is witnessing steady growth, driven by factors such as increasing industrialization, growing demand for automation in manufacturing processes, and government initiatives to promote the use of robotics. However, the market growth rate is being impacted by challenges such as high initial costs and lack of skilled labor. (statista)

An LLM does not calculate truth, but probabilities

How the Transformer Architecture Works

Is it even important to know?

Complexity & the near Run in the AI Hype

AI for THAT

MAD

LLM Board

Let's start simply as a plain explainer….

Let's focus on what MATTERS

Instead of getting lost in complexity, we focus on practical applications of AI.

GPT, LLM & CO – explained super simply

At their core, large language models are systems that predict, based on probabilities, which text should follow a given input.

The PROMPT: The art of instruction

A "prompt" is an instruction or cue given to the AI to perform a specific task.

The key to AI: Defining results

What matters is knowing in advance what outcome you want and how to instruct the AI accordingly.

What is Artificial Intelligence?

Generative Pre-Trained Transformer Explained Using ChatGPT as an Example

Well-Known LLM Models and Chatbots




What does Prompt mean?

PROMPT - Request / Instruction

A "prompt" is, in the context of Artificial Intelligence, an instruction given to the AI to perform a specific task. THE ART OF IT…. is knowing in advance what result you want to achieve.

Prompt Architecture



Risks… DeepFakes & Digital Violence

What are Deepfakes?

The term is a combination of the words "deep," referring to "deep learning," and "fake," meaning a forgery or fabrication.

Technological Foundation

Deepfakes are based on complex neural networks that analyze large amounts of data to generate deceptively realistic forgeries.

Societal Risks

The technology enables the creation of fake videos and audio content that are barely distinguishable from real ones and can be used to spread disinformation.

Example Prompt: "Generate an image of the pirate Captain Jack Sparrow" in Midjourney

AI Video & Image Generation – 2022-2026

2022

First AI image generators with recognizable artifacts

2023

Improved quality and level of detail

2024

Integration of video generation and animation

2025/2026

Photorealistic results – indistinguishable from real ones

Which Image is "REAL"?

MIDJOURNEY & FLUX 02 2025 - The distinction between AI-generated and real images is becoming increasingly difficult, posing new challenges for authenticity verification.
(#2 is real and not AI generated)

What is real – What is AI generated?

Danger

Challenges

4 things to always keep in mind!

In addition to critical, moral, and logical thinking skills that one should always be aware of...there are

What intention is behind a message?

classic intentions in fakes: fear, time pressure, authority, money, data

Who sent
this message?

Sender quality: For business decisions, only verifiable senders and known channels count

What is the source
of the message?

Sender is not the source; without a source it is a claim, not information

Can the message be verified online?

always check or clarify internally

Artificial Intelligence: Challenges

The EU AI Act is part of the EU's efforts to regulate the use of artificial intelligence (AI) in the Union and to ensure that AI systems are safe, transparent, and ethically sound.

Unacceptable Risk

Prohibited AI systems

High Risk

Strict regulation

Limited Risk

Transparency requirements

Minimal Risk

Self-regulation

South Africa is not yet governed by a dedicated AI Act equivalent to the EU AI Act. Instead, AI-related compliance currently relies on existing legislation — especially POPIA — while the government is developing a National AI Policy that is expected to evolve into a broader regulatory framework later in 2026.

Even without a final AI Act, the direction is clear: organisations should prepare for risk-based oversight, stronger accountability, sector-specific guidance, and closer scrutiny of high-impact AI use cases.

Greatest Challenge: MINDSET

Since February 2, 2025, companies are obligated

According to Article 4 of the EU AI Act, companies must ensure that their employees have sufficient AI competencies.

Cultural change is necessary

The greatest challenge lies not in the technology itself, but in the willingness to adopt new ways of thinking and question traditional ways of working.

Continuous training

Companies must invest in the AI education of their employees in order to remain competitive and meet the requirements of EU regulation.

South Africa’s AI Shift

Pragmatic adoption, rising expectations, and growing pressure to reskill

Cyclical Economic Development Regularities

Innovation

New technologies emerge

Growth

Spread and economic upturn

Regulation

Legal frameworks emerge

Adaptation

Economy adapts to new conditions

The introduction of new technologies such as AI historically follows a cyclical pattern, in which phases of innovation are followed by growth, regulation, and economic adaptation.

Source: Russian economist Nikolai D. Kondratiev

Development Speed
of Digital Core Technologies

Chat GPT – The 100 Million User Threshold

*in years

Weekly ChatGPT Users

Sources: OpenAI (2022–2026), Stanford AI Index 2026, TechCrunch, Backlinko

The Hidden Water & Energy Hunger of AI

The Training Problem: Power Consumption

GPT-3 Training
34 days of training | 175 billion parameters

* Annual consumption of ~ 120 German households

1,287

MWH*

GPT-4 Training

* Annual consumption of a small town

~60,000

MWH*

1 million Megawatt hours = 1,000 Gigawatt hours = 1 Terawatt hour

Source: Mosharaf Chowdhury, University of Michigan

The Global Dimensions

Tech Giants Dominate Energy Consumption

1 TWh equals the annual electricity consumption of approximately 450,000–500,000 South African households.

Sources: Google Sustainability Report 2025; Microsoft Sustainability Report 2024; IEA – Energy and AI Report 2025; Amazon/AWS Environmental Report 2024; Meta Sustainability Report 2024

The Water Crisis - The Invisible Thirst

AI data centers consume 2x more water than conventional data centers

GPT-3 Training
34 days of training | 175 billion parameters

*~ 3.5 Olympic swimming pools

5.4 M

Liters*

Google USA 2021

*the annual consumption of a city the size of Gqeberha (Port Elizabeth)

12.7 B

Liters*

Source: Li, P. et al. (2023): "Making AI Less 'Thirsty': Uncovering and Addressing the Secret Water Footprint of AI Models" Universities UC Riverside & University of Texas Arlington

What does this mean in concrete terms?

1 ChatGPT Request: 2-4 Wh

= 1 LED bulb for 1 hour

1 Image Generation: 3.6 Wh

= 100% smartphone charge

10-sec. Video: 1 kWh

= 10 hours of TV watching

1 ChatGPT Conversation (20-50 questions):

10-30 Image Generations:

0.5-sec. Video

= 500ml of water


Sources: Epoch AI (Feb. 2025), Stanford University & AXA (2024): "Energy Scaling Laws for Diffusion Models", MIT Technology Review (2025): "We did the math on AI's energy footprint", Li, P. et al. (2023): "Making AI Less Thirsty", UC Riverside & UT Arlington

Opportunities: Conserving Resources with AI

AI for Less Chemicals in Food, Soil & Water

Live Examples (Impulse Talk Only)

Fridge and Your Blood Values

AI can analyze health data and provide personalized nutrition recommendations tailored to the contents of your fridge.

Professional Prompting

In the education sector, AI can develop innovative learning activities and create personalized learning materials — e.g. with Gamma — including text, images, and layout.

Employee Onboarding

AI assistants can support the onboarding process for new employees and provide personalized training materials.

Automation

From a WordPress blog post, ChatGPT automatically writes platform-specific promotional content for Facebook, Instagram, X, LinkedIn, and as an email to colleagues — all according to predefined guidelines.

Transcription with Plaud

What if someone could objectively and with multi-layered complexity record minutes and transcribe them for further processing.


Notebook LM

Analyze and summarize URLs or uploaded documents. Create podcasts, mind maps, flashcards, and summaries to facilitate learning and information processing.

Suno: Create Music

From a prompt with song lyrics and a music genre, Suno creates individual custom music.

Travel Planning with Manus

From a prompt with travel dates, location, budget, and preferences, the agent builds a website with restaurant suggestions, cultural events during that time, sights, and more.

Build a Website with Lovable

From a prompt with your requirements, facts, and context, Lovable builds a fully functional website with an imprint, privacy policy, cookie banner, contact form, and all images and texts in your corporate design.

The AInside team

AI doesn't start with data, but with

trust, attitude and curiosity.

With us, the HUMAN is at the center.

Daniela Purps - founder of AI Inside

Neo Potele


Saskia Pircer


Rob Galler - founder of Kinside Germany






© Kinside - Rob Galler, 2026 – all rights reserved.


Note: Content on this platform is created using AI systems and is editorially reviewed before publication.