Adam Broda

🎈Attention!! 🎈 Calling all SWE Technical Leaders with “at least 3-4 years” leading/mentoring teams and handling end to end large scale projects, located in the Bay Area, CA or NY Metropolitan locations, supporting Product/Backend Systems or Machine Learning/Data Science domains, and have “NOT” interviewed with Meta in the last 12 months! If you meet these qualifications, please send me a message to connect and chat about the exciting opportunities at Meta! Last week, we released the new version of Meta AI, our assistant that you can ask any question across our apps and glasses. Meta AI has been upgraded with a new state-of-the-art Llama 3 AI model. We believe Meta AI is now the most intelligent AI assistant that you can freely use, and marks a significant milestone in our AI investments. By open-sourcing this technology, we aim to encourage collaboration and knowledge-sharing among developers worldwide, ultimately leading to even more advanced tools and applications. In addition to this, we have made progress towards our vision for the Metaverse, opening Meta Horizon OS - our operating system powering Quest - so others can design more headsets in our ecosystem. This will allow new headsets to be created in the coming years, focused on productivity, entertainment, exercise, gaming, and more. We also announced new product features and styles for Ray-Ban Meta smart glasses, including AI integration. These are just a few examples of the groundbreaking work that we are doing at Meta, and we believe that there is no better time than now to join! Learn more about our Meta AI assistant built with Llama 3 here:

2

🌟 New from #NVIDIAResearch: Weight-Decomposed Low-Rank Adaptation (DoRA). A groundbreaking advancement in fine-tuning technology that's set to revolutionize how we optimize pretrained models without increasing inference costs. 👀 https://dy.si/BniXqX2

1

Did you know that of 30,000 products launched every year, 95% of the products fail? Validating your idea is required to ensure you have a successful product. How do you validate your product. Here are some ideas 💡 User Interviews or Surveys 💡 Minimum Viable Product (MVP) an opportunity to learn 💡 Wizard of OZ 💡 Build a Prototype 💡 A/B Testing 💡 Quantitative Research and many many more... If you rush in and launch a product without validation you stand a high chance going too far and then realizing no one wants it. Join us in our next Scrum Alliance A-CSPO (Advanced Scrum Product Owner) on July 22 - 25 2024 to learn how to discover, plan and launch your product successfully. Schedule does not work? We have one planned every month. Link in the comments below.

1

1 Comment

Today’s Large Language Models (LLMs) are getting a lot of negative attention for the compute power it requires to run them, but the environmental and financial price tag of AI doesn’t have to be as head-spinning as the headlines might suggest. It just requires an understanding of when scale is necessary for high quality AI outputs, and when it isn't. Read more: #salesforce #AI

6

This is why WE need YOU! Different builds Different The Rotational Engineering Program was designed to provide access & and equitable experience for candidates that historically have not been viewed as “traditional software engineers”. ❗️APPLY HERE :https://lnkd.in/gDTQjcz4

5

Absolutely fascinating to see this play out. The PR value and general buzz that Starbucks have achieved from what was 'internal comms' shows how important it can be strategically, when done right. The intercept in the venn diagram between external and internal comms is clearly strategic in nature. Well worth a read below ⬇️ #Communications #StrategicComms #SeriousIntentComms

16

3 Comments

This is an extremely thoughtful and well observed piece Tim Lee has written (thanks John for posting). Well worth reading. My take: Almost all credible autonomy technology companies are using neural networks in (increasing) parts of their architecture and development processes. They have an unmatched ability to learn patterns, recognise complex scenes and generalise at scale. But it is also true that they alone may not (likely will not) get you all the way. Many of the reasons for that are called out neatly in this article. But there are also some omissions. There are some useful hints in the recently passed UK AV Bill that refers to “in-use monitoring” as well as “explanability and transparency”. In essence; in the event of fault, the Authorised Self Driving Entity (ASDE) must be able to discover and explain the fault and then rectify it. This is inherently difficult for neural nets but standard for safety critical system. AI’s routinely make mistakes or experience phantom recall. That might be okay in training data. But it’s also possible these mistakes first appear at run-time when it’s too late and there’s no way to discover it and test for it ahead of time. Though we each have our unique magic star dust, I would wager Oxa’s architecture is more closely aligned with Waymo’s, at least philosophically and from the bedrock of safety and validation. The key word here is “validation”. They, like us, probably use a combination of technologies and techniques that match the right need. It’s likely some of those techniques utilise as many complex specialist algorithms as they do combinations of compound AI’s. Its understandable to assume that geographic restrictions at first blush might seem clunky and slow to scale. The honest answer is the early versions are. However, our bet (and I’m certain Waymo’s bet) is that the speed in which we are innovating to validate new “autonomy ready” geographies will outpace the speed in which black-box neural-net only approaches can be ready for driverless operation (if ever). This is the fastest route to generalised, validated, safe autonomy for everyone. Or what we like to call “Universal Autonomy”.

35

2 Comments