This video is in German, so here is a short summary.
TLDR: The potential of Large Language Models in AI technology is discussed, along with the importance of protecting intellectual property and investing in Deep Tech for Europe’s relevance in the global market.
Read more: Video: Frank Thelen and Aleph Alpha founder Jonas Andrulis talk about the impact, opportunities and threats of AI- 🤖 Jonas, founder of AlephAlpha and former Apple AI manager, discusses the potential of Large Language Models and the difference between Apple’s AI R&D and Special Projects divisions.
1.1 Large Language Models (LLMs) are a breakthrough in AI technology that can process and condense vast amounts of text, but are limited to text and lack general intelligence.
1.2 GPT is a family of AI that includes narrow AI for specific tasks and the potential for artificial general intelligence to develop beyond human capabilities.
1.3 Jonas is working on a hot topic and introduces himself and his work.
1.4 Jonas, founder and managing director of AlephAlpha, previously worked at Apple in California in the management of AI research and special projects department.
1.5 Apple’s Special Projects division operates differently from its AI R&D division, which focuses on gaining knowledge rather than creating products, and is stronger in AI research than is publicly perceived.
1.6 Apple keeps their unreleased projects tightly locked away, but has recently started attending conferences and giving presentations, while still relying on surprise and delight for product launches, and Aleph Alpha was founded in 2019 as a European-based AI company. - 🤖 AI technology is advancing to understand complex structures beyond human limitations, but it still has limitations and is not suitable for critical and complex work such as in the medical or legal field.
2.1 AI technology is limited to human-defined signals and cannot achieve superhuman insight, but a new generation of AI is emerging that can understand complex structures beyond human limitations.
2.2 Observation can lead to learning and navigating different worlds, including computer game worlds, which opens up new possibilities similar to DeepMind’s ability to understand and create games independently.
2.3 Imitation learning and independent learning have been used in research to teach machines to play games and understand language without human annotation.
2.4 Chat GPT is not suitable for critical and complex work where correctness, complexity, and trust play a crucial role, such as in the medical or legal field. - 💡 Companies can protect their intellectual property and rely on internal knowledge with a reliable model that provides sources and evidence for every factual statement.
3.1 The model can provide reliable answers based on internal knowledge without the need for external sources, and can also provide sources and evidence for every factual statement to ensure context and responsibility.
3.2 You can run the models on-premise or in the cloud, and data can be moved into the system in a legally compliant manner without needing an expert.
3.3 Companies see their value creation at risk with their intellectual property, so they protect their internal knowledge and processes to safeguard their livelihood. - 🤖 AI systems are built from internet crawls, but companies with high-quality data are reconsidering freely accessible data for commercial use.
4.1 To improve the model’s knowledge, one can fine-tune it with additional data and allow it to access a knowledge base at runtime.
4.2 Models of skill level can use archives to search for information and references, but there is a debate about the legality of using freely accessible data sources like LinkedIn and Reddit.
4.3 AI systems are built from common crawls of the internet, but platforms and companies with high-quality data are reconsidering freely accessible data that others can benefit from commercially.
4.4 Respect data marked as non-commercial and unusable, as it is not intended for training other models. - 💰 European deep tech start-ups face funding challenges compared to other regions, with some founders having access to significantly more funding than others, making private investment necessary for growth and innovation.
5.1 The speaker shares their experience with financing deep tech start-ups in Europe, highlighting the differences in funding and growth issues compared to other regions.
5.2 Different funding environments exist, with some founders having access to significantly more funding than others, as seen with OpenAI receiving $100 million from Elon Musk, while German-European investors in deep tech financing have provided a total of 28 million, which is still three orders of magnitude less.
5.3 We set the tone in innovation with multimodality two years ago and now have explainability, making us the only ones with this feature, and the challenge is to stay competitive and capital-efficient.
5.4 Politicians should support companies in energy storage, AI, and other important technologies to maintain Europe’s independence, but current funds are insufficient, so private investment is necessary.
5.5 Investing in AI is difficult due to the long-term investment and technical evaluation required, and Microsoft appears to be the current winner in the market. - 💻 Big tech companies are making moves, but it’s uncertain if they can all stay relevant, and there’s an opportunity for European companies to build 100 billion plus companies in the new industry.
6.1 Big tech companies like Alphabet, Apple, and Amazon are making moves in the industry, but it remains to be seen if they can all stay relevant.
6.2 The possibility of a tech giant losing its dominance in the future is high, but it doesn’t necessarily mean there will only be one winner in the end.
6.3 Google and Alphabet have great potential in the field of AI, while Apple may surprise us with their personal data combined with models.
6.4 There is an opportunity for companies, including European ones, to build 100 billion plus companies in the new industry and defend themselves against changes in added value.
6.5 New technology and the current industrial revolution have enormous potential for start-ups and medium-sized companies to strengthen Europe’s value creation potential, but the challenge of social restructuring is a concern. - 💡 By 2030, humans may create transformative AI that can independently generate new knowledge, while Germany’s focus on regulation may hinder their adoption and investing in Deep Tech is crucial for Europe’s relevance in the global market.
7.1 Transformative AI, defined as AI that can do or transform a large part of knowledge work and independently generate new knowledge, is estimated to be built by humans in the year 2030 according to a study by the Open Philanthropy Foundation.
7.2 By 2030, there may be a general artificial intelligence that can develop itself further and transform knowledge work, but it is not yet an AGI scenario with singularity and completely autonomous systems.
7.3 Germany’s excessive focus on regulation instead of innovation in AI may result in the country being left behind as paying customers and users of the technology, with control and value creation going elsewhere.
7.4 Investing in Deep Tech in Europe is crucial to prevent Europe from becoming irrelevant in the global market dominated by China and the USA, but it requires a significant amount of capital. - 🚀 Politicians should support key technologies to prevent being left behind and to have a strong industry that can be financed directly by companies and entrepreneurs, as digitization with sovereign AI can open up great possibilities for growth and investment opportunities.
8.1 The speaker finds it exciting that there is another dimension and believes that politicians should support key technologies to prevent being left behind and to have a strong industry that can be financed directly by companies and entrepreneurs.
8.2 Digitization with sovereign AI can open up great possibilities for German and European companies, consultancies, and administration, leading to potential growth and investment opportunities.
8.3 Masters of Doom is a book that combines gaming, entrepreneurship, and software innovation, and the speaker also recommends cognitive scientist Joscha Bach for his unique perspective on AI.
8.4 The speaker discusses the importance of collaboration with strong partners in order to face the challenges of the current era and highlights the positive aspect of finding such partnerships.
Leave a Reply