Jonas Mayer
TNG Technology Consulting GmbH, Senior Consultant
TNG Technology Consulting GmbH, Senior Consultant
Munich, Germany
Actions
Jonas Mayer is a Senior Consultant at TNG Technology Consulting. As Head of Innovation Hacking, his main focus lies on creating innovative showcases and prototypes in soft- and hardware. Since 2018 he's been working on numerous projects ranging from market-leading Realtime Deepfakes, an LLM Shitposting AI, all the way to autonomous drone racing. As a keynote speaker, he has been talking about the Innovation Hacking projects at over a hundred conferences all across IT and Tech. Prior to joining TNG, Jonas studied Informatics: Games Engineering at TU Munich. More information can be found at innovation-hacking.com.
Jonas Mayer ist Senior Consultant bei TNG Technology Consulting. Als Kopf des Innovation Hacking Teams liegt sein Fokus auf der Entwicklung von Prototypen und Showcases, die innovative Technologien in Soft- und Hardware erlebbar machen. So arbeitete er seit 2018 an verschiedensten Projekten, wie zum Beispiel Echtzeit-Deepfakes, einer LLM Shitposting-KI oder autonom fliegenden Minidrohnen.
Vor seiner Zeit bei TNG studierte Jonas Informatik: Games Engineering an der TU München. Mehr Informationen findet man auf innovation-hacking.com.
Links
Area of Expertise
Topics
AI Lofi HipHop - How Not To Generate AI Music
Wie gut kann KI-generierte Musik schon sein? Um diese Frage zu beantworten, haben wir einen 24/7-Live-Stream erstellt, der bis in alle Ewigkeit KI-generierten Lofi-HipHop spielt.
Musik ist im ständigen Wandel - von Live-Musik über Schallplatten bis hin zu Lofi-HipHop-Live-Streams auf YouTube. Auf dem Höhepunkt des KI-Hypes startete das TNG Innovation Hacking Team den Versuch, den nächsten logischen Schritt zu gehen: einen endlosen Strom von kontinuierlich frischer Musik, komponiert und produziert von Algorithmen und KI. Heraus kam das "AI lofi radio", ein schwer zu ertragender 24/7 Live-Stream, gefüttert von einer Software, die verschiedenste KI-Modelle mit echtzeitfähiger Produktionssoftware kombiniert.
Dieser Vortrag gibt einen Einblick in die KI-gestützte Musikproduktion aus einer naiven IT-Perspektive. Zunächst werfen wir einen Blick unter die Haube, wie Computer Musik verstehen und verarbeiten. Anschließend zeigen wir, wie verschiedene KI-Modelle genutzen werden können, um Musik zu komponieren oder direkt Audio zu generieren. Nach der Theorie, werden wir auf verschiedene Überlegungen zur Umsetzung eines KI-Musik-Generators eingehen und unsere größten Lernerfahrungen teilen. Wir zeigen, wie man entspannte Drum-Grooves und eine dazu passenden Melodie erzeugt und diese in programmatisch synthetisiert. In gewohnter Innovation-Hacking-Tradition, runden wir den Vortrag mit einem Feuerwerk von Live-Demos ab.
AI Slides - Generating Slide Decks with Artifical Intelligence
In the age of GPT-4 and Stable Diffusion, complex slide decks can be generated automatically at the click of a button. Even better! They can also be delivered by AI itself!
Has this ever happened to you? There's a corporate presentation tomorrow, but the slides are nowhere to be found! Well, no worries! TNG proudly presents AI Slides: This AI system generates a complete slide deck, you only need a title. Too tired for the delivery? No Problem: The AI even takes over the presentation itself! A personalized Deepfake Avatar armed with the knowledge of the entire internet guides the audience through the slides.
This talk will give you a technical overview of AI Slides and the underlying artificial intelligence. Using live demos, our still-human-for-now speakers demonstrate how GPT-4 can generate headlines and content in multiple languages and how Dall-E 3 or Stable Diffusion fill the slides with appealing images. Afterwards, the speakers will show you how to give the Avatar a voice using text-to-speech engines like Amazon Polly or Resemble.ai and finally a face using Living Portrait technology.
Feel free to bring an idea for your own AI-generated presentation to the talk. Some pre-generated talks can already be found at ai-slides.com.
The AI Elections: How Technology Could Shape Public Sentiment
In the age of Deepfakes and Large Language Models (LLMs), the threat of misusing AI for shaping and manipulating public sentiment has reached unprecedented levels. In this talk, we shed light on the capabilities of Artificial Intelligence (AI), and its potential implications for our societies.
As pivotal elections in both the UK and USA loom in 2024, it is important to understand the evolving landscape of political influence, particularly in the digital realm. Remembering the impact of Cambridge Analytica’s manipulation during the Brexit era serves as a reminder of how information can be distorted to sway public opinion.
In this keynote, we show the capabilities of modern AI and how they can be misused for shaping public opinion, using examples like Realtime Deepfakes and LLM-powered chatbots. Through a series of live demonstrations and showcases, we unveil the inner workings and technical foundation of these neural networks, highlight their capabilities and their weaknesses, and give an outlook on how we as a society can resist manipulation.
In a world where technology shapes narratives, understanding its potential and pitfalls is crucial. This talk aims to spark discourse and promote a future where technology serves societal betterment, not manipulation.
The Shitposting AI - An Ironic Solution to Fruitless Online Debate
Using modern AI approaches such as OpenAI's GPT we created fully autonomous robot heads that engage in heated social media discussions, completely taking the human out of the loop. The TNG Innovation Hacking Team created a prototype of an end-to-end natural language understanding system, employing techniques such LLM Text Generation, Text-To-Speech (TTS) and Robotics.
Social media comments and threads have become the predominant medium for public discussion. However, discussions on X, Reddit and Facebook are notorious for their poor debate culture and missing conclusiveness. The obvious solution to this tremendous waste of time is automation of such fruitless discussions using an LLM powered bot.
In this talk, we will give an introduction to different facettes of Natural Language Processing while showcasing how we used them for building the Shitposting AI. Kicking things off, we take under the hood of LLMs: what makes them so powerful, how to collect your own shitposting dataset and how to constantly fine-tune your LLM on the newest data. We also take a look at the current state of the art of Text-To-Speech and how we built the robot heads. As custom for the TNG Innovation Hacking Team, we showcase these concepts with a plethora of live demos ending with the grand finale: A live, AI-powered robot discussion about the most pressing issues of the modern day.
Pushing Deepfakes to the Limit - Fake video calls with AI
Today's real-time Deepfake technology makes it possible to create indistinguishable doppelgängers of a person and let them participate in video calls. Since 2019, the TNG Innovation Hacking Team has intensively researched and continuously developed the AI around real-time Deepfakes. The final result and the individual steps towards photorealism will be presented in this talk.
Since its first appearance in 2017, Deepfakes have evolved enormously from an AI gimmick to a powerful tool. Meanwhile different media outlets such as "Leschs Kosmos", Galileo and other television formats have been using TNG Deepfakes.
In this talk we will show the different evolutionary steps of the Deepfake technology, starting with the first Deepfakes and ending with real-time Deepfakes of the entire head in high resolution. Several live demos will shed light on individual components of the software. In particular, we focus on various new technologies to improve Deepfake generation, such as Tensorflow 2 and MediaPipe, and the differences in comparison to our previous implementations.
Jonas Mayer
TNG Technology Consulting GmbH, Senior Consultant
Munich, Germany
Links
Actions
Please note that Sessionize is not responsible for the accuracy or validity of the data provided by speakers. If you suspect this profile to be fake or spam, please let us know.
Jump to top