It’s completely natural for humans to attribute human characteristics to generative AI systems that seem to behave like us. And since language is one of the most profoundly human characteristics on earth, it explains why we tend to compare LLMs like GPT or PaLM (Bard) to humans, and even attribute emotions to them.
But in the same way that we don’t expect an airplane to flap its wings like a bird, or a submarine to swim by using wave-like movements like a fish, it’s important to remember that LLMs do not use language the way humans do. They are an immensely useful invention to generate language but are blind to experiences that aren’t encoded in language and have no true emotional understanding.
Listen to this fascinating discussion between Vasagi Kothandapani, Bart Maczynski and Marina Pantcheva in our latest Globally Speaking Podcast episode and learn all about the fundamental differences between language learning in LLMs and humans and how we can encode ethics in AI.
Bart Mączyński is a VP of Machine Learning at Language Weaver. Since 2000, when he joined Trados, Bart has held consulting roles helping enterprise and government customers with translation management platforms, terminology systems, and machine translation. His current focus is the practical application of linguistic AI in emerging use cases.
Born into a multicultural family, Marina Pantcheva paired her multilingual background with an eccentric childhood dream that led her to mastering 11 languages…and still counting. After working as translator, educator, consultant, and founding her own language school, she turned to exploring the elementary particles of language in her PhD research on Nanosyntax. In 2014, Marina exchanged the academic life for the high-paced world of localization, and currently leads a multifaceted team which develops solutions for crowd localization. In her spare time, Marina reads about science and engages in research inspired by the vast amount of data she encounters in her daily work.
Vasagi is a Senior Vice President of Strategic accounts responsible for multiple global accounts and also head of RWS’s TrainAI data services practice, responsible for delivering complex, cutting-edge AI training data solutions to global clients operating across a broad range of industries. She has 27 years’ of industry experience and held various leadership positions in Business Delivery, Technology, Sales, Product Management, and Client Relationship roles in both product development and services organizations globally.
Vasagi has spent most part of her career in Banking and Financial services, Technology and Hospitality practices managing multiple Banks and Fintech accounts and led several technology and Digital transformation initiatives. Vasagi holds a Master’s degree in Information Technology and a Post Graduate Certificate in Artificial Intelligence along with several industry certifications in Data Science, Architecture, Cybersecurity and Business Strategy.