Links of the week #41

AD/ADAS

• Revisiting ASAM OpenSCENARIO V2.0 (Oct. 2024) by Applied Intuition , with a focus on three categories of scenarios

AI

• Will ASICs Dethrone GPUs for LLM Inference ? (July 2024) by Nicolas C. , talking about hardware, for once !

In crypto mining, the initial reliance on GPUs and FPGAs (Field-Programmable Gate Arrays) gave way to the rise of ASICs (Application-Specific Integrated Circuits) from China, which eventually took over due to their superior efficiency and performance for specific algorithms. Could we be on the cusp of a similar revolution in AI Inference hardware?

• Using Google NotebookLM with “Understanding…” (Oct 2024), by Olivier Ezratty . Olivier's compendium about quantum technologies was part of my selection last week, but to be honest due to the more than 1500 pages it's still in my to-read list. In this article Olivier shows how he used NotebookLM with his document, and rates the answers provided by the LLM. Here's his conclusion :

At this point, NotebookLM is “emulating” me to some extent but not entirely, hopefully. It is a useful tool to consolidate information that is scattered in long documents or in several documents. It helps uncover some myths and misconceptions, beyond the famous “quantum hype” that creates a reality distorsion field around the real and potential capabilities of quantum computers. But NotebookLM won’t really reason and it can make mistakes, lacking “judgment” in some situations. You need to know the rules of the game here and always fact-check what these tools are producing.

• STATE OF AI REPORT 2024 (Oct. 2024), over 200 slides, by Air Street Capital . I selected below the one relative to the predictions

CAD / PLM

• The Future of Open Source in PLM: Can It Solve Key Problems? (Sep. 2024), by Oleg Shilovitsky , who also mentioned in his post this interesting discussion initiated by István Csanády about open-source software in CAD

In my view, the future of FOSS PLM is still uncertain. The last decade has seen increased investor interest in PLM solutions, with major exits like Arena, Aras, Upchain, and Onshape after long life cycles. While these examples demonstrate the growing importance of PLM, none have embraced the FOSS model, except of Aras Enterprise PLM.

Robotics

• Some Thoughts on Robotics Startups (Oct. 2024), by Michael Ferguson

Some of these firms claim that RaaS is the new SaaS: spoiler alert - it’s probably not! Robotics is capital intensive. Especially if you are buying all these robots, keeping them on the books, and then renting them out with a 1-2 year payback window. Even worse: robots age a lot worse than servers. (...)

Finally, while many folks will tell you that “hardware is hard” - the bigger problem is that hardware is SLOW. Supply chains have improved from the days of the pandemic, but they are still slow, inefficient and generally a bit of a hot mess. So when you do suddenly land all those orders - good luck getting the parts you need to actually fulfill the order quickly.

Python

• What is (was?) GIL in Python? (Oct. 2024), by Avi Chawla , explaining Python's global interpreter lock which restricts a process from running more than one thread at a time.