ABOUT LANAILABS
LanaiLabs is an Artificial Intelligence Lab focused on advanced uses of AI in the enterprise. We are dedicated to making AI safe for everyday life. Our team is composed of experts in AI, software engineering, and data science.
Who We Are
We assembled a talented and unique team with over 100 years combined in data science, technology startups, and data science. The hype of generative AI and GPT is relatively new, but members of our team have been researching and developing AI for decades. One of the areas of interest has been how to best identify language that has been generated by a machine versus a human. Much of our interest in this area started back when team members were working on the large language models (LLM) for IBM Watson. Shortly after IBM Watson won Jeopardy! in 2011, it was clear that LLMs were going to quickly achieve even greater heights in both natural language processing (NLP) and natural language understanding (NPU). This motivated members of our team to start researching and building methodologies to better distinguish machine-generated content from human.​
In this photo provided by Jeopardy Productions, Inc., Ken Jennings, left, and Brad Rutter, right, pose after the episode of "Jeopardy!" that aired Wednesday, Feb. 16, 2011, when Watson, the IBM-created megabrain, beat the veteran champs with a total of $77, 147 over two exhibition matches.(AP Photo/Jeopardy Productions, Inc.)
Why We Are Different Than Others?
Our ongoing research and understanding of how advanced LLMs and machine learning NLP/NLU have evolved led us to the use of common principles and methods used in physics for the last couple of centuries, such as linear regression and concepts around energy and variety (many of us on the team are actual physicists by trade). As humans, we don’t choose our words based on probability (like LLMs), but rather we select words in a much more complex way that is rooted in many facets of what makes us human. For instance, we often choose words based on prior life experiences or emotions. In cognitive science and psychology, this is referred to ad hoc category construction, a way that we categorize objects, concepts, or stimuli in a fluid manner, versus relying on pre-defined concepts or probability of what word should come next, something that a machine simply cannot do now and may never be able to do.
