Understanding our planet

Large language models (LLMs), like BERT, GPT-3, BART, and T5, seem to understand and process human languages in many ways. They are trained on tons of information to help us with tasks such as understanding and translating texts. But we do not know if they understand texts that mix different languages, like when someone switches between English and Dutch in a conversation. That's called code-switching (CS). We’re investigating whether LLMs need to recognize CS to translate these mixed-up texts accurately. So, imagine trying to teach a robot to understand both English and Dutch when people use them together.

Nowadays, we can hardly avoid interacting with technologies that speak or otherwise mimic how humans communicate with one another. Devices ranging from phones to supermarket checkout machines often address us without warning and create the impression that they understand natural languages. In this talk, I argue that such technologies should & quot;shut the hell up" as recommended by a meme frequently shared on social media. I contend that unwelcome, artificial social interactions initiated by technologies which can & quot;talk" may distress or even harm us, thus, should not occur in the first place.

What would you say if I told you zombies were real? Not the dead come back to life mind you, but the kind infected by parasites, like in the film 28 Days Later or the television series the Last of Us! Over the last decade, scientist have discovered many different parasites capable of manipulating the behavior of their hosts, including worms, bacteria, viruses, protists, and even fungi. Join me over a pint to discuss the fascinating biology behind how some of these parasites, the zombie fungi, hijack the minds of insects to better spread their infectious spores to new, unsuspecting hosts.

Please use this link on the speaker bio.
For reference, please use your name and surname in the subject.
With this payment method no tickets will be send to you. For confirmation please keep the payment proof.