How Generative AI Works (a Very Rough Guide)
-
Locked
Jarsto van Santen
Specialist Test Engineer
Talk Description
Let's build a language model together!Â
Think it's impossible in the timeframe? Well, it's not going to be a general purpose Large Language Model. In fact, it's going to be a highly limited, incredibly Tiny Language Model. It's not about what the Language Model can do in this case, it's about what building it will show us.
I believe that to make truly effective use of a tool we need to have a rough understanding of how it works. But for many testers and automators LLMs seem to be a black box: we don't know how they work, so we're not sure when to use them or not to use them.
Making a TLM based on audience input is just one of the ways in which this talk aims to give a rough guide to how Generative AI works. So that we don't end up doing the equivalent of trying to push a nail into a wall with a screwdriver, or using a chainsaw instead of sandpaper.
What you’ll learn
By the end of this talk, you'll be able to:
- How to effectively use generative AI
- Why generative AI struggles with some tasks
- What we can and can't fix with better prompting
Jarsto van Santen
Specialist Test Engineer
Jarsto has been playing with code for a quarter of a century now, having started at 11 years old. Despite studying law he eventually saw sense and turned his hobby into a career. He now works as a Specialist Test Engineer at DUO (part of the Netherlands' ministry of Education).
At work he is active in the Automation Serviceteam - developing internal tools, helping and teaching colleagues, and looking critically at what should and should not be automated - as well as taking on testing-related matters that affect the entire organization, coordinating with specialists from other DevOps disciplines, and advising management as needed.
In his free time Jarsto is mildly obsessed with world history, science fiction and fantasy, and all sorts of technology he doesn't get to play around with at work (yet). Somehow he never has the time to watch as quite many movies/series, read quite as many books, or play quite as many games as he thinks he should.
Suggested Content
