The second session of AICVS was conducted on 4th August, 2018 by Kirit Thadaka who is a research engineer at Zero labs where he leads the Zero fabric project. The session was attended by 170 students of Cummins. The title of his session was, "WHAT'S THE DEAL WITH AI?" Kirit started the session by discussing the various subfields of Artificial Intelligence. Artificial Intelligence can be mainly divided into three groups, Symbolic learning, Statistical Learning and Neural Networks (Deep Learning). Symbolic learning is based on high-level "symbolic" , ie, human representations of logic and search. It can be subgrouped into the fields of Robotics and Computer Vision. This differs from Statistical Learning which relies of data to learn and forms a predictive function based on the given dataset. Statistical Learning is applied in fields like NLP (Natural Language Processing) and Speech Recognition. Neural Networks are modelled on the human brain and nervous system. Neural Networks can be of different types like CNN (Convolutional Neural Networks), RNN (Recurrent Neural Network) and LSTM (Long Short Term Memory). The next topic discussed was the history of AI ranging from brief introductions to in depth discussions of the developments in the field of AI over the years. The evolution of this field right from Turing Machines to Watson winning a game of Jeopardy to AlphaGo Zero were discussed and left everyone in awe of the advances made up to this point but also left them aware of the advances yet to be made. Kirit then discussed various Learning Algorithms like Supervised Learning, Unsupervised Learning and Reinforcement Learning. Learning Algorithms are methods that are used to process data to extract patterns appropriate for application in a new situation. Supervised Learning is one in which the dataset is labeled and consists of a set of training examples while Unsupervised learning is used to draw inferences from a dataset consisting of input data without labeled responses. Reinforcement Learning deals with how the software agent takes actions in an environment so as to maximize some notion of cumulative reward. This was followed by a interactive discussion with the students about whether they think machines are going to take over the world and how soon where some students presented their view points. Kirit then discussed what he felt was lacking in the Machine Learning and Artificial Intelligence Industry which were mainly, biases in data which could lead to human bias being replicated in AI models, the Brute Force approach to solving most AI problems today where the model is trained rigorously on a dataset and an excessive amount of training that is needed for a model to perform a minor task. We then discussed Creativity in Artificial Intelligence and examined a Deep dream Image. We then tried to figure out which audio clip was Wavenet and which one was human. It was incredibly difficult to distinguish between the two and most of us got it wrong! We then listened to some music created by an AI model. He then talked about the different levels of AI mainly, Artificial Surface Intelligence, Artificial General Intelligence (AGI) and Artificial Superintelligence. Artificial Surface Intelligence which is intelligence which permits machines to perform tasks in specialized domains like chess. AGI is the intelligence of a machine that could successfully perform any intellectual task that a human being can. Artificial superintelligence is a term referring to the time when the capability of computers will surpass humans. It goes a step beyond, and posits a world in which a computer’s cognitive ability is superior to humans. We then watched a video about Atlas. Atlas is the latest in a line of advanced humanoid robots we are developed by Boston Dynamics. Atlas' control system coordinates motions of the arms, torso and legs to achieve whole-body mobile manipulation, greatly expanding its reach and workspace. Kirit then gave the students a demo on Jupyter Notebooks on a text generation model he had created. The model analysed Game Of Thrones text and generated text of its own. The model can be found on Kirit's GitHub Page. He explained the model and how the previously discussed concepts and paradigms were being used. We thank Kirit for taking the time out to come and hold a session on ML on behalf of the AICVS Club. We hope that all the students who attended the session learned a lot about AI. We hope to collaborate with Kirit in the future and wish him all the best for his future endeavours! Written by Michelle Davies.
2 Comments
|
Archives
February 2020
Categories |