Proxify Developers’ Post

Having your own company is filled with ups and downs. But the positives beat the negatives. Check out our latest episode of Builders with Sohrab Hosseini of orq.ai out now on Spotify, Apple Podcasts, Google Podcasts and YouTube. 𝐋𝐢𝐧𝐤𝐬 𝐭𝐨 𝐭𝐡𝐞 𝐞𝐩𝐢𝐬𝐨𝐝𝐞 𝐢𝐧 𝐭𝐡𝐞 𝐜𝐨𝐦𝐦𝐞𝐧𝐭𝐬 𝐛𝐞𝐥𝐨𝐰. #podcast #coding #softwaredevelopment #developers #programming

Thanks to Proxify and Proxify Developers for the great chat we had on entrepreneurship and the 99 challenges on the road for orq.ai. Links to the podcast in the comment!

How has technology shaped education over the last 50 years—and where might it take us next? In the latest episode of Education Matters, I sit down with Hall Davidson, a true pioneer in edtech and a fantastic storyteller. We discuss: ➖ Practical advice for edtech companies navigating schools. ➖ Key takeaways for district leaders shaping their tech strategies. ➖ Hall’s reflections on AI and its place in the classroom. ➖ The importance of the humanities in learning. Whether you're an educator, policymaker, or edtech entrepreneur, this episode offers unique perspectives and actionable ideas. Listen now and join the conversation: 🔗 Spotify: https://lnkd.in/geVG_DRY 🔗 Apple Podcasts: https://lnkd.in/dcdWeJBz #EducationMatters #EdTech #AI #K12 #EducationLeadership Discovery Education

I had a blast talking with Alex Gray on his podcast. Here are 5 key ideas from our conversation ⤵️ 1) AI-generated content should be used as a supplement to teacher expertise rather than a replacement for teachers' knowledge and guidance. 2) We must be mindful not to let big tech make us small humans! 3) AI presents an opportunity to reduce the burnout rate among educators, who are increasingly burdened with administrative tasks. 4) It is crucial to train students to analyse AI-generated content critically. 5) AI-generated content should be used as a supplement to teacher expertise rather than a replacement for teachers' knowledge and guidance.

From Big Tech to EdTech: Jeremy Gu's Mission to Transform AI Education In this episode of the Sync'd Up podcast, we sit down with Jeremy Gu. Jeremy is a former Director of Data Science who made a bold transition to founder of AI Strategy and Education. He shares his journey from working at Amazon and Uber to launching his own company, the challenges of early-stage entrepreneurship, and his vision for transforming education with AI. Tune in to hear insights on leadership, adapting from corporate life, and the future of personalized learning. Link to the full episode 👇

𝐇𝐚𝐯𝐞 𝐲𝐨𝐮 𝐞𝐯𝐞𝐫 𝐢𝐦𝐚𝐠𝐢𝐧𝐞𝐝 𝐠𝐨𝐢𝐧𝐠 𝐛𝐚𝐜𝐤 𝐭𝐨 𝐬𝐨𝐟𝐭𝐰𝐚𝐫𝐞 𝐝𝐞𝐯𝐞𝐥𝐨𝐩𝐦𝐞𝐧𝐭 𝐰𝐢𝐭𝐡𝐨𝐮𝐭 𝐭𝐡𝐞 𝐬𝐮𝐩𝐩𝐨𝐫𝐭 𝐨𝐟 𝐚𝐧𝐲 𝐝𝐞𝐯𝐞𝐥𝐨𝐩𝐦𝐞𝐧𝐭 𝐞𝐧𝐯𝐢𝐫𝐨𝐧𝐦𝐞𝐧𝐭𝐬 (𝐈𝐃𝐄𝐬)? While that may be clearly a thing of the past, there is another evolution currently happening in the software development process: #𝑨𝑰-𝒑𝒐𝒘𝒆𝒓𝒆𝒅 𝒄𝒐𝒅𝒆 𝒂𝒔𝒔𝒊𝒔𝒕𝒂𝒏𝒕𝒔 Tools like #GitHub #Copilot, Amazon #CodeWhisperer, #Tabnine, and SAP's own #Joule have started becoming an integral part of the developer's toolkit. They help developers go from idea to software much faster and more efficiently. One thought that's been on my mind lately is: what does it mean to go back to unassisted coding? Here is a small podcast episode on this topic. Hope you enjoy it! Let me know in the comments what you think. Will we ever see that happen? 𝐏.𝐒: If you're curious to listen to/watch other episodes of my podcast, head over to any platform of your choice (YouTube, Spotify, Apple Podcasts) and check for "𝐊𝐢𝐬𝐡𝐨𝐫𝐞'𝐬 𝐏𝐨𝐝𝐜𝐚𝐬𝐭". #AI #codeassistants #developers

Very much enjoying playing with the #AI features of the Google #NotebookLM, especially the option to take some text and generate a two person "podcast" discussing the topic! I recently shared a LinkedIn article (https://lnkd.in/eak32Cvz) about my talk at this year's Dublin Tech Summit. I was able to give this link to NotebookLM and it generated the attached dialog. LinkedIn doesn't seem to like just plain audio files so I used Clipchamp to bring in images of the slides that I used in the presentation. What do you think? Useful learning tool?

The future of learning is not the written word. That's according to Victor Riparbelli, co-founder of one of the UK's only unicorns, Synthesia . Mankind started to communicate with cave drawings after all ...except those tablets with the 10 commandments of course. It's all about images, and moving ones (that was harder in caves and much easier with laptops et al) So as I build out the Podcast Guest Blueprint course I'm following their lead. Instead of lots of dull writing, and after hours of trying to make power point sound interesting, I'm hiring AI course trainers. If you'd like to see more of these two, and the course that they are helping me to create on being the golden podcast guest, just DM me. If you want to try Synthesia out for yourself...use this link to make a short video. https://lnkd.in/eJ5NxRVh #AI #podcast #courses

🧐 Machine Learning Case Study: Apple building transcripts for Podcasts Apple recently updated the Podcasts app with transcriptions. You can now listen to podcast episodes alongside readable transcriptions. Of course, at Apple's scale, there was no way to do this by hand. ML to the rescue! Aside from the transcripts being high quality, they released a blog post detailing how they got there. Instead of using the common academic benchmark WER (word error rate) they used HEWER (human evaluation word error rate). The former takes into account every error (including “ah”, “umm”, "yeah" vs "yes" etc) whereas the latter focuses on errors that make the transcript unreadable or have a different meaning. Often transcripts will have a large WER but a much lower HEWER (the script is still readable). The takeaway? This is a perfect example of where default academic benchmarks (while useful) are not always the most ideal for your production use case. When a new model comes out with state-of-the-art results on academic benchmarks it can be tempting to adopt it immediately. However, the correct approach is: Develop a test set for your use case -> experiment, experiment, experiment (with existing and new models)! For example, working on Nutrify, I've found that a fine-tuned 2-year old pretrained computer vision model outperforms almost all recent models on our custom dataset when accounting for performance/latency tradeoffs. However, I know this may change at any given moment. Hence, experiment, experiment, experiment! --- Links in top comment. #ai #machinelearning #apple #podcasts

Here's a reference to a machine learning case study based on Apple's development of their transaction service for Apple Podcasts. I'm sharing it because the summation point reiterates what I wrote about in the AI Strategy post last week. Deliberate experimentation is key to getting AI right. You can read my post here - https://lnkd.in/eTRgVFun