 Our next presenter is Brendan Bolt. Their title is Understanding AI Invented Language. Some of you may remember back in 2017, when some Facebook researchers had to shut down these AI chatbots because they invented their own language that humans couldn't understand. Yes, these news stories were an exaggeration, but they got some of it correct. There really is a field which looks at how AI invents language. It's called emergent language. It's also true that these emergent languages can be difficult to understand, but in reality the AI invented something far less sophisticated than English. Nevertheless, in the last six years, we now have AI like chatGPT which show an impressive command of language. For this reason, my thesis will develop computational methods to allow humans to intuitively understand emergent languages. There are two sides to understanding emergent language. One human-centric, the other AI-centric. The human side of understanding language is what we call linguistics, yet emergent languages can look nothing like human languages. Thus, my thesis will adapt the methods for analyzing human languages to the unique challenges of analyzing emergent languages. These analyses cover everything from low-level traits like grammar and vocabulary to high-level structures like conversation and social relationships. Studying emergent language from this perspective will help address unanswered questions in linguistics such as how humans invented language. The AI side of understanding language is called natural language processing, or NLP for short. NLP works by learning statistical patterns in language data from billions of words. It's only at this scale that AI can begin to account for the immense complexity within a human language. Yet, collecting this much data can be problematic as it can come from surveillance of social media, underpaid crowd workers, or biased content off the internet. For this reason, my thesis will also use NLP to determine if emergent language displays the same large-scale statistical properties as human language. If human and emergent languages are similar in this regard, then down the road, NLP can rely more on emergent language data and less on these troubling sources of human language data. The computational methods I am developing to understand emergent language will allow humans to understand our own language better, as well as support language-capable AI of the future. This way, the next time that AI-inventing language makes the news, it can be a celebration of new insights into human and artificial intelligence instead of stoking fears of existential threats.