I was attending a very valuable meetup about AI and Data Science. Valuable because speakers were well experienced in Machine Learning, working with data streams and Natural Language Processing.
Paweł Noga – How to join AI?
Paweł started with examples of Machine Learning solutions and one of the most interesting was GPT-2. This model takes text as an input and generates more paragraphs about the given topic. It adapts to the style and content of the input text and its release was suspended until November because it was considered as a dangerous tool to create fake-news.
Paweł was talking about hardware and software stacks of Machine Learning. It was the introduction to describe the skills needed to start working with AI. Python is usually a basic requirement. To learn more please check Data School recommended by Paweł.
Michał Staśkiewicz – If or when will AI replace lawyers?
Michał was talking mainly about working with text and documents. How to automatically find key pieces of information in dozens of financial reports? Or how to increase sales based on brief notes of marketers? The NLP algorithms can find the patterns if you have enough data. How much is enough? Hundreds of thousands will do the job.
Michał said that for machine learning training he is using GPUs and workstations in the office because it’s quite an affordable solution. On the other hand, Paweł said, that he’s using cloud solutions for his work (like AWS, Azure) because of flexibility in computing power he can get. I guess both have some pros and cons depending on your needs.
Wojtek Ptak – Machine Learning on Data Stream
Wojtek was talking about a transformation from monolith to event-driven architecture. Kafka, Spark, Hive or Flink are some of the tools he was using to manage tons of emails that FreshMail was sending. These tools helped also to distinguish between wanted messages and spam.
What’s important: It was not only the change in technology. It was also about the human side of transformation.
Rewire your Culture, Technology Success will Follow
Transparency, taking a risk and training at every level of the company are required to introduce changes in a team. It help’s to choose the right technology. More about changes in their team.
I remember that Wojtek recommended Martin Fowler’s presentation about event-driven architectures. I think this is it:
I’m going to watch it and maybe write about it.