Talking AI is a collection of blogs and interviews where we discuss Artificial Intelligence (AI) with our talents to gain some insight into their world and further understand how they work with AI. This week we’re talking with Lucas, our Machine Learning Engineer, as he tells us more about himself and shares his experience with AI.
First off, can you tell us a bit about yourself and your role at Talent-e?
Hello, my name is Lucas Giordano and I’m currently a machine learning engineer at Talent-e. I’ve studied AI at the École polytechnique fédérale de Lausanne with a focus on natural language processing and the ethical and fair usage of AI. At Talent-e, I develop content generation systems based on artificial intelligence and machine learning principles. I’m very passionate about my studies and enjoy cooking, photography, and team sports!
What can you tell us about the AI loop that you’ve been building with Talent-e?
We started working on the system with my colleague Yassine which basically came down to two main phases. Optimal data collection and gathering through cloud technology and prompt engineering powered by AI pioneers such as OpenAI. What I find often interesting is reflecting on the ways that we can optimize the system’s interface for our copywriters so they can work in the most cost-efficient manner.
As a machine learning engineer, what’s it like to work with copywriters on a daily basis?
As ML engineers we lack the domain expertise of whom we are building the system. We’re constantly in discussion with copywriters and obtain valuable feedback from them on how to develop the system in the way that suits them best. I’ve found that it’s always handy to have a chat to align perspectives and discuss what practical options are on the table that we can implement.
What do you think generative AI will look like in 2/3 years and what can we expect for Talent-e?
Generative AI has been decent enough for the past five years but obviously it’s not perfect. Sometimes AI generates its own “hallucinations”. Text is the easiest to generate while video and speech are much more difficult to produce and require remarkable effort. With AI in general, it’s necessary to have a discussion around its ethical and unbiased usage. How do we reach that stage? How do grow to a level where AI can produce without human supervision? As for Talent-e, the text modules will keep improving. Video and speech are potential avenues we can explore while the immediate next steps would be improving post allocation and machine learning outcomes.