Complex Systems and Artificial Intelligence
This is a meetup group that I run as personal hobby/passion. AI is really cool and it mixes together the perfect amount of programming and math that I look for (probablities and statistics are two studies that have always made me really happy). This is the description from meetup page itself:
If you are highly interested in artificial intelligence and discussing the complex systems behind it, then this meetup is for you! The intention of this meetup is to have in-depth discussions with researchers and developers in the field, provide technical presentations, and build a network of professionals looking to share their knowledge.Here's a link to my meetup page!
Past meetups:
AlphaGo Zero & Meltdown/Spectre
Description:
This month's discussion will explore AlphaGo Zero, the latest evolution of Google DeepMind's AlphaGo AI. AlphaGo is the first computer program to defeat a professional human Go player, and the first to defeat a Go world champion. AlphaGo Zero tackles the same problem but with a different approach; it is trained completely through its own experiences without data from any external sources (hence the "Zero"). Zero is more powerful and efficient than all previous iterations of AlphaGo, and was able to beat the previous strongest version by 100 games to 0. We will present the AlphaGo Zero white paper in an engaging technical presentation and understand how this system works.
On a separate but equally interesting note, we will discuss the speculative execution exploits Meltdown and Spectre. Meltdown and Spectre exploit a hardware flaw in branch prediction, where the CPU of a computer will make predictions on what it will compute next in order to maximize speed and efficiency. However, this gives a window of opportunity for malicious applications to bypass memory isolation to read and access the contents of the entire memory. We will discuss the implications of this exploit and how it could possibly affect the field of AI.
===== This is an academic-style journal club where one person chooses a topic or paper and presents it. Participants are expected to have read the paper or other material suggested by the discussion. =====
PRESENT A TOPIC AT FUTURE EVENTS: Generally, the discussion leader doesn’t have to be an expert on the subject, but should be interested enough in it to read the paper thoroughly so as to give a decent presentation. After informally presenting the topic, the group can then ask questions or open discussion surrounding the topic. The presentation should be informal (slides are allowed, but whiteboard/chalkboard is preferred), and this should be a discussion, not a one-way transmission of information by the presenter.
The reading material doesn’t have to be a white paper. In the case of widely-known topics a chapter of a textbook, Wikipedia article, or other material can be suggested.
Ethical Decision Making with AI
Description:
As the field of AI advances, researchers have started to play around with the capabilities of their systems to test what questions AI could answer. While the useful questions concerning sentiment and classification are already considered pretty abstract, researchers have started to ask questions that go a step beyond. Ethical decision making is something that humans themselves cannot even define effectively, yet AI systems have been made to approach this problem. At this meeting we will discuss the implications of this while closely following the MIT white paper "A Voting-Based System for Ethical Decision Making." This paper discusses the decisions a self driving car would make in extreme circumstances, where the ethical decision is decided by the results of online polls taken by random people. Along with this paper, we will discuss systems that claim to be able to identify potential terrorists or criminals, and even someone's sexual orientation, just from their face. Is this a ground breaking application of AI, or just an embellished resurgence of the archaic pseudoscience of physiognomy? This will be an open discussion, so feel free to come in and state your stance!
Apple's new AI engine
Description:
Apple recently came out with news about their revolutionary custom SoC, the A11 Bionic neural engine. The company boasts up to 600 billion operations per second for real-time processing. The engine allows for many powerful AI uses, along with specific machine learning algorithms that can enable a new wave of smart phone security. Will this new chip actually live up to these claims? Is this mostly marketing fluff or the real deal? At this point we can only speculate, however there are enough technical resources out there that can help us form a better idea, such as apple's coreml documentation. Come to our monthly discussion where we will explore this topic in depth!
Monte Carlo Sim. Value & Policy Netwk
Description:
As an extension of our market trends and industry demonstrations, we will be hosting a monthly journal club to discuss specific topics in the Complex systems and AI ecosystem. This month's discussion will explore Monte Carlo simulation with value and policy networks, as we take a look into the underlying algorithms and data structure behind the game GO (paper link below): https://storage.googleapis.com/deepmind-media/alphago/AlphaGoNaturePaper.pdf
This will be an academic-style journal club where one person chooses a topic or paper and presents it. Participants are expected to have read the paper or other material suggested by the discussion leader. PRESENT A TOPIC AT FUTURE EVENTS: Generally, the discussion leader doesn’t have to be an expert on the subject, but should be interested enough in it to read the paper thoroughly so as to give a decent presentation. After informally presenting the topic, the group can then ask questions or open discussion surrounding the topic. The presentation should be informal (slides are allowed, but whiteboard/chalkboard is preferred), and this should be a discussion, not a one-way transmission of information by the presenter.
The reading material doesn’t have to be a whitepaper. In the case of widely-known topics a chapter of a textbook, Wikipedia article, or other material can be suggested.
Discussion w/ IEEE-HKN
Description:
At this event we will be networking and discussing all news AI! Feel free to share any details about your background with AI or any projects/research you are working on. We will discuss a variety of topics relating to the field, so be prepared to give your input.