Über uns
There are many Artificial Intelligence meetup groups in Munich. This group is specifically focused on Artificial General Intelligence (AGI), which is also known as Strong AI or human-level AI. This subfield is about creating an agent which can learn to accomplish any task that human beings or other animals are able to perform.
Systems like MuZero or GPT-4 show sparks of AGI, but are not AGI themselves. MuZero can only learn to play different games, but cannot create new music, for instance. GPT-4 is good at a lot of different tasks, but it also sucks in mathematics. An AGI will be able to learn everything that a human can potentially learn. This includes playing games, making music, solving mathematical problems, driving, etc.
The only example of an AGI system in this world is the human brain. Therefore, most people interested in AGI start by studying the human brain and try to replicate some of its working principle in machines. Though not widely known, this is the research approach motivating the main figures in Deep Learning e.g. Frank Rosenblatt, Yoshua Bengio, Geoffrey Hinton, Yann LeCun, or Demis Hassabis. This is the research approach that eventually led to the architectures behind MuZero, GPT-4 and other awe-inspiring neural network based systems. But the journey is not over yet and many mysteries remain.
This research effort mainly happens in academia and in big tech (FAANG). But I believe that there are many smart people outside the elite circle of academia and big tech who want to research, experiment and contribute to this exciting quest. LAION, an organization that created the dataset behind Stable Diffusion, is actually run by a high school teacher in Hamburg and goes to show that such contributions are possible.
I privately research AGI in my spare time. However, I miss having a peer group that makes research in academia and big tech so fulfilling. You can bounce your ideas off each other and learn together. So I thought it will probably be a lot of fun if we could bring together a peer group through meetup.com. That was the reason for starting this group.
The purpose of this group is to bring together people from all walks of life who are studying AGI, in order to create a collaboration network within the city of Munich and its surroundings.
Here are some frequently asked questions about this meetup group.
Q: What is the meetup language?
A: The meetup language is English.
Q: How are you different from other AI groups?
A: AGI is primarily a research field and can be characterized as "hard tech". Therefore, we are not very interested in the economic or social side of AI e.g. how to create business value with narrow AI applications, economic ramifications of automation, and safety concerns. These topics are of course not excluded, but these are also not our main focus. The main focus is the research and development of AGI.
Q: Do I need to have any prior knowledge?
A: Nope. However, it will help to have some prior experience in Machine Learning or Cognitive Sciences (e.g. computational neuroscience) or Robotics. If you don't have it, you can still come, and hopefully, the events will inspire you to learn more.
Q: Do you recommend any specific books or resources?
A: To get a general overview, we recommend Architects of Intelligence by Martin Ford. This is just a series of interviews with the people at the forefront of AI. It is a pretty light reading and can be finished in a couple of weeks. To learn about Deep Learning, we recommend Deep Learning: A Visual Approach by Andrew Glassner for those who want a light but serious introduction without the use of mathematics or programs. We recommend Deep Learning by Francois Chollet for those who want a code-heavy, maths-light introduction. For those who don't mind heavy maths, we recommend Deep Learning by Goodfellow, Bengio, and Courville (freely available). For computational cognitive neuroscience, we recommend Computational Cognitive Neuroscience by O'Reilly, Munakata et. al (freely available), and the Neuromatch video series. For Robotics, it's probably best to get a ROS-supported robot (or even a Raspberry Pi or Arduino-based robot) and play with it.
