The 38th Chaos Communication Congress (38C3) in Hamburg, Germany, was the latest installment of the annual four-day conference on technology, society and utopia organised by the Chaos Computer Club (CCC) and volunteers. From the long list of talks, I chose six I found especially relevant for readers of Smart Content Report and myself. I used AI summaries to better understand the topics and perspectives discussed in these talks beforehand.
Waiter, There’s An LLM In My Search!
In this talk Martin Hamilton discusses the growing prevalence of large language models (LLMs) and their impact on internet searches and information quality. He highlights the challenges users face in finding reliable content amid the rise of sponsored and SEO-driven material, as well as the dangers of malvertising. Martin warns about the transition from surveillance capitalism to surveillance fascism, particularly regarding sensitive information related to healthcare and gender transition support. He emphasizes the absurdity of AI-generated content, such as hallucinated images, and the difficulty in discerning real information from misleading sources.
To combat these issues, he suggests strategies for users to regain control over their search experiences, including adding parameters to Google searches and utilizing browser extensions. Martin also proposes the idea of creating a meta search engine that focuses on community interests and specific domains rather than attempting to index the entire web. He discusses the potential of using APIs from platforms like Mastodon to surface trending topics while preserving user privacy. Lastly, he invites listeners to join a community initiative aimed at improving search quality and addressing the challenges posed by generative AI content.
Language: English
Link to the talk: https://media.ccc.de/v/38c3-waiter-there-s-an-llm-in-my-search
Automation and Empathy: Can We Finally Replace All Artistic Performers with Machines?
This talk by Moritz Simon Geist explores the intersection of music, emotion, and automation. He discusses whether emotions can be effectively communicated through music without human involvement, raising questions about the nature of performance and the role of machines. Geist reflects on his journey as a robotic musician and his experiences with creating installations that react to human presence, such as his 2022 project “Don’t Look at Me,” which utilizes algorithms to adapt its music based on audience interaction.
He contrasts robotic performances with human performances, emphasizing the importance of agency and emotional connection in captivating art. Geist introduces the concept of the “uncanny valley,” which describes the discomfort people feel towards nearly human-like machines, affecting emotional bonds. He argues that while robots can replicate technical skills, they lack the interpretive skills and spontaneity that characterize human performances.
Despite advancements in automation, Geist concludes that fully replacing human performers on stage is not yet feasible due to the missing elements of agency and emotional engagement. He invites further exploration into the creative aspects of performance that may remain uniquely human.
Languages: English, Deutsch
Link to talk: https://media.ccc.de/v/38c3-automation-and-empathy-can-we-finally-replace-all-artistic-performers-with-machines
Resource Consumption of AI – Degrow or Die
Thomas Fricker, a physicist specializing in cloud security, discusses the substantial energy and resource consumption of data centers, particularly in relation to AI. He compares data centers to factories that produce computing power, emphasizing their massive energy needs, which can reach up to 300 megawatts in Berlin. The increasing demand for AI is driving an exponential growth in energy consumption, with predictions suggesting that by 2030, data centers could consume 9% of the U.S. energy supply.
Fricker highlights the environmental challenges posed by this growth, including water usage for cooling and the depletion of resources like uranium and copper. He points out that while nuclear power is often proposed as a solution, there are significant limitations and environmental concerns associated with its expansion. Additionally, he criticizes the job creation claims linked to data centers, noting that the investment per job is disproportionately high.
Fricker calls for a reevaluation of growth models, advocating for a degrowth approach to mitigate the environmental impacts. He stresses the need for public awareness and action against the unchecked expansion of data centers, urging local communities to question the resource consumption associated with such developments. He warns that without significant change, the exponential growth of data centers could lead to severe ecological and economic consequences.
Languages: English, Deutsch
Link to talk: https://media.ccc.de/v/38c3-resource-consumption-of-ai-degrow-or-die
Klimaschädlich by Design – die ökologischen Kosten des KI-Hypes
This talk by Friederike Karla Hildebrandt and Constanze Kurz discusses various aspects of climate change and technology’s role in addressing environmental issues. It highlights the significance of data and statistics in understanding global warming and the impact of large language models (LLMs) on sustainability efforts. There is a focus on the collaboration between tech companies like Google, Amazon, and Microsoft in developing solutions to mitigate climate change. Their presentation touches on the challenges of energy consumption and transparency in AI training processes. Additionally, the talk addresses the importance of community involvement and academic contributions in tackling these global problems. The speakers emphasize the need for sustainable practices within the tech industry, including recycling and responsible resource management. They also discuss the potential for universal basic income and compute as part of future solutions.
Languages: Deutsch, English
Link to talk: https://media.ccc.de/v/38c3-klimaschdlich-by-design-die-kologischen-kosten-des-ki-hypes
Let’s build dodos! How generative AI is upturning the world of synthetic biology and hopelessly overwhelming traditional governance instruments
This talk by Margret Engelhard discusses the intersection of artificial intelligence, synthetic biology, and nature conservation, focusing on the implications of rapidly advancing biotechnology. Engelhard is a risk assessor at a federal agency and highlights the increasing prevalence of genetically modified organisms (GMOs) and the challenges of assessing their risks to biodiversity. Advances in DNA sequencing and synthesis, alongside the integration of AI, are transforming the field, allowing for more complex genetic modifications. She raises concerns about the unpredictability of these engineered organisms, particularly regarding their potential impact on natural ecosystems.
Examples of innovative applications, such as gene drives that ensure the persistence of modified genes in wild populations, are presented, alongside ethical questions surrounding the de-extinction of species like the dodo. Her presentation emphasizes the need for a public discourse on the societal desirability of such biotechnological advancements, as existing regulatory frameworks struggle to keep pace with technological changes. Engelhard argues for a balanced approach that includes effective regulation, societal transparency, and ethical considerations in the development and release of genetically engineered organisms.
Languages: English, Deutsch
Self Models of Loving Grace
And finally, this talk features a discussion on the nature of consciousness, artificial intelligence (AI), and the human mind, led by a researcher with a rich academic background, Joscha Bach. He explores philosophical questions surrounding self-awareness, consciousness, and the potential for AI to replicate or understand these concepts. Key topics include the scaling and universality hypotheses in AI, which suggest that larger models may lead to greater intelligence and that different models might converge on similar structures when given the same data and tasks. Bach argues that consciousness could be a simpler construct than often perceived, emphasizing the importance of self-modeling and the relationship between the mind and the body. He also touches on the notion of “zimbos,” entities that believe they are conscious but may lack true self-awareness, and questions whether AI could fall into this category. His talk dives deeper into the implications of AI systems simulating consciousness and the ethical considerations of creating such beings. Bach advocates for a deeper understanding of consciousness as a fundamental aspect of both biological and artificial systems, suggesting that AI research may provide insights into these profound questions.
Languages: English, Deutsch, Polski
Link to talk: https://media.ccc.de/v/38c3-self-models-of-loving-grace