For the 23rd Belgium NLP Meetup, we're invited to Ghent by NLP startup Tekst. On Thursday May 2nd, the company opens its doors to seasoned NLPers and newcomers alike, for our familiar program of talks, discussions and networking opportunities. As always, there are three presentations on the agenda: Philippe Bogaerts will show us how to best implement NLP applications from a security standpoint, Vignesh Baskaran will discuss planning and reasoning by autonomous agents, and Tiebe Parmentier will explore the relevance of finetuning in today's NLP landscape dominated by large language models.
As always, doors open at 7 pm, when there will be pizza and drinks. The talks themselves start around 7:30 pm. Tekst shares its offices with In The Pocket, at Sassevaartstraat 46 in Ghent. We look forward to seeing you there!
NLP: The Ultimate Weapon in Purple Teams' Arsenal for Application Security
Philippe Bogaerts
It is fascinating to see how AI and closely related technology is changing the way we build applications. In this talk, Philippe will highlight the benefits, but also the dangers and pitfalls of how we implement this awesome technology from a hacker's point of view. Philippe strongly believes the benefits overshadow the dangers, and hopes to contribute in such a way that we bridge the gap between over-excited app development and the caution exercised by security teams.
Do not expect complex math or algorithms, just some common sense, tips and tricks from someone who has been securing applications for a long time.
Planning and Reasoning with LLMs for Autonomous Agents
Vignesh Baskaran (Hexo.ai)
LLMs can be more than just next token predictors. Their applications can span beyond just Question Answering or Chatting. LLMs can be assigned a task such as making a reservation at a restaurant. They can come up with a plan: divide the task into steps/sub-tasks, start executing each sub-task (action), monitor the output of each action and reason out in cases of success or failure and adapt the plan based on its reasoning. Such systems are called Autonomous Agents. In this talk we will focus on how such Autonomous Agents can plan and reason out using LLMs to perform tasks.
Finetuning: Between Luxury and Necessity
Tiebe Parmentier (Tekst)
With language models getting bigger and better all the time, the question arises: is finetuning even needed anymore, or are there easier and more effective ways to achieve optimal accuracy for a particular NLP task? In his presentation, Tiebe will explore the impact and usefulness of finetuning, particularly for narrow, specialized applications.
