This funding of Black Semiconductor caught my attention because of Aachen (one of my favorite cities) and because it's a major recent development not directly focused on Generative AI.
Black Semiconductor is focusing on leveraging graphene for chip-to-chip connectivity, moving away from traditional electronic connections towards photon-based interconnects.
Black Semiconductor, however, is pushing towards commercial scalability and integration into existing semiconductor manufacturing processes. I can imagine its impact on complex data processing tasks like streaming where latency will be reduced through innovations on a foundational hardware level.
I built aiplanet.com where numerous beginners accessed free AI learning resources provided by a diverse group of contributors (primarily experienced AI practitioners). Building on the common advice, here are some insights to consider, given your existing context:
- AI/ML is diverse, with data scientists specializing in different areas. I know AI experts who have still not delved into LLMs; they have their specific focus areas. AI/ML skills encompass a wide range of topics, and data scientists often have specific focus areas. Continuous exploration and reading are crucial. Resources like paperswithcode.com are valuable for discovering new research areas and domains.
- While time-consuming, Kaggle offers exposure to robust modeling and validation skills. These skills are critical, though they are only a fraction of what's needed for real-world projects. It's beneficial to expand beyond these skills. This being said, it does give bragging rights. I've seen company founders, like those at H20.ai, often highlight their Kaggle Grandmasters.
- My current role is at Pathway.com. Over 80% hold of my colleagues PhDs, and our CTO has co-authored with folks like Geoff Hinton and Yoshua Bengio (I find that cool actually :)). But this environment may reflect my bias towards academic research. This being I said, I believe that strong foundational understanding is essential and also valued, especially when tackling complex challenges.
- Active participation in forums and communities related to the frameworks you use is highly recommended, like TensorFlow User Groups. At Pathway.com, we welcome those interested in stream data processing to our community. Engaging in these forums offers the chance to receive support from the original creators and leading community members. Other notable communities include DataTalks.Club and MLOps.Community.
It's interesting how it gets Jupyter working from streaming workloads. Jupyter notebooks are the de-facto choice for most data scientists around me. Haven't really tried it out myself but I imagine this could be a value addition for streaming ML use cases for sure.
Black Semiconductor is focusing on leveraging graphene for chip-to-chip connectivity, moving away from traditional electronic connections towards photon-based interconnects.
There was a similar news around graphene but for graphene-based chips by folks at Georgia Tech. https://spectrum.ieee.org/graphene-semiconductor#:~:text=Res....
Black Semiconductor, however, is pushing towards commercial scalability and integration into existing semiconductor manufacturing processes. I can imagine its impact on complex data processing tasks like streaming where latency will be reduced through innovations on a foundational hardware level.