Humans in the Loop: the Frontiers of Data Science

Activeloop

Humans in the Loop sheds a light on thought leaders in various verticals of data science, machine learning, and artificial intelligence. Once every two weeks, we share our interviews with leading experts from around the world, where we explore how solving a machine and deep learning problem helps achieve business goals and foster innovation. Tune in to stay in the loop with the latest groundbreaking ideas on the AI/ML landscape. You can shape this podcast by helping us pick our next guest by joining our community slack and voting! Twitter: @activeloopai Podcast & Activeloop Community Slack: slack.activeloop.ai read less
TecnologíaTecnología

Episodios

#03: Artem Rodichev: designing AI friends, optimal machine learning architecture, the future of communication
04-02-2021
#03: Artem Rodichev: designing AI friends, optimal machine learning architecture, the future of communication
In this episode, Black Mirror becomes a reality, because we’re featuring an intelligent chatbot serving as a companion, mentor and even lover. Today’s human in the loop is Artem Rodichev, Head of AI at Replika. Born in Kazakhstan, he’s worked in all major Russian tech companies, and ultimately, moved to Y-Combinator Alum startup Replika and somehow managed to get on the Forbes 30 under 30 list twice. Replika’s a company born out of its founder trying to revive her tragically deceased friend by utilizing his personal messages with the help of natural language processing and conversational AI. With Artem, we talk about the future of communication - empathetic chatbots that help people with anxiety, loneliness, depression or just have a good chat with them. We discuss the fundamental differences between talking to another human being and an empathetic chatbot, and how these can be used to make people happier. On the data science side of things, Artem walked us through Replika’s machine learning architecture - and their multimodal approach comprising vision, speech and text models. Because we were pretty constrained in terms of time, we concentrated on the text component powering their dialogue engine: retrieval, reranking, generative and other models. We’ve also touched upon GPT-3, and why it is not always the best solution to a given problem. Artem also explained how they evaluate what framework to use in which setting to provide a more immersive, and enjoyable experience to the user that makes them happier.