October 2018 Summaries

A group at Clarifai developed a "Spooky or Not" app during a hack day, allowing users to text images and receive feedback on whether the images are spooky, using a combination of Python, Clarifai, and Twilio services. The app leverages Clarifai’s ability to analyze images using both general and custom models to identify spooky concepts, while Twilio facilitates the sending and receiving of text messages via a Python Flask web server. The process involves setting up a Python environment, installing necessary packages, configuring both Clarifai and Twilio accounts, and using ngrok to make the local Flask server accessible online. This project illustrates the integration of machine learning and communication technologies, offering an engaging example of how these services can be combined for creative applications.

In the blog post discussing the film "Transcendence," the author critiques the portrayal of artificial intelligence in movies, emphasizing the discrepancies between cinematic depictions and current technological realities. The movie, which features a scientist's consciousness uploaded to a sentient computer, exemplifies common misconceptions about AI's capabilities, such as the existence of general AI and its potential to control humanity. The post argues that while AI in films often appears omnipotent and autonomous, real-world AI remains limited to narrow applications like computer vision and natural language processing, far from achieving a general AI that can replicate human cognitive abilities. The piece also highlights the importance of distinguishing fact from fiction to foster informed discussions about AI's ethical and societal implications, especially as the technology continues to evolve.

Facial recognition technology, a form of artificial intelligence, identifies individuals by capturing and analyzing facial images against a database. It has several use cases, including face verification, closed-set recognition, and open-set recognition, each with different requirements and challenges. The technology operates through a pipeline that involves detecting faces, aligning them, extracting features using deep neural networks, and matching these features against a database of known faces. The effectiveness of facial recognition depends on factors such as the angle of the face, known as yaw, and is influenced by the quality of the training set and computational resources. While deep learning can enhance each stage of the recognition process, the system's performance is determined by how well it manages variations in facial images and the match metric used to compare facial vectors.