AI

I had ~60 hours of .ogg and .m4a audio recordings and was keen to see if AI could transcribe them without sending the files to a cloud-based service. I see huge potential in local AI models, both from an environmental and privacy perspective, but this was my first real-world test. My hardware was a modest ThinkPad T14 Gen 1 AMD (equipped with a Ryzen 5 PRO, 16GB RAM, and no dedicated GPU). Running local AI models without a discrete graphics card can be slow and painful, so the setup needed to be optimised to run entirely on CPU without bringing the machine to its knees. ...

With a little prompting, Gemini generated a working Tampermonkey script as documented in Identify your users by name in Amplitude. To conclude the conversation I asked, “How can I credit you as an author?” That’s very thoughtful of you! If you plan to share the script with others (e.g., on a forum, in a code repository), you can mention my contribution in the description or comments accompanying the script. For example: ...

Amplitude is a decent product analytics tool, but there is no compromise when it comes to identification. User IDs are fixed forever and all data is immutable once it lands in the platform – names and email addresses are not an option. Knowing the name of a user can unlock insight by giving context to usage patterns, especially in early-stage projects with a handful of users who are working at the same company. ...

Jumping on the bandwagon, I asked ChatGPT (Jan 9 Version) to draft a post for this blog and answer a series of follow up questions in the same thread. Here are my thoughts: Training and running a model like ChatGPT requires significant computational resources. I would like to see developers publishing precise energy consumption and carbon footprint figures and considering environmental impact as a key measure of their machine learning model’s performance (alongside accuracy, precision, recall, etc). ...