Claire Lebarz’s Post

AI - not tomorrow, but supporting some new OS data products today What's the magic behind combining AI with location data? Discover how we're harnessing AI and automation for OS location data: https://ow.ly/CpPw50SI5fw

Really impressed with the work that went into this ML, AI and Data landscape by FirstMark. We see hundreds of "out of the box" integrations on this list, proving that #cloudnative is the foundation for #AI. From #databases to applications and AI Frameworks, MinIO works with all of them. https://lnkd.in/dPQA9YHs

Interested in a sneak peak into the AI future? Then take a moment to read this today:

And that was a wrap! As #apisummit24 completes with over 10,000 attendees from the API community - making it the largest API conference by far - with dozens of awesome product and partner launches to show what we can do with *the API platform for the AI era*... one fundamental question remains open: Are you on the right side of AI? One thing we know for sure... APIs are the common language of LLMs as we build apps and transform our business with {Gen} AI. In this post, I wrote a few thoughts from my opening keynote on the API market evolution that's ahead of us. Lmk what you think!

Gemini 1.5 Flash shines for its large context window, speed, and affordability. It boasts a 60x larger context window than GPT-3.5 Turbo, runs 40% faster for 10k character inputs, and costs up to 4x less with context caching. Learn more on Vertex AI. ⇣

The 2024 MAD (Machine Learning, AI & Data) landscape report is an good read for anyone interested in the current state and potential direction of AI. This report covers a range of topics, including the current state of AI companies, the modern AI stack, and a discussion on the potential AI bubble. Check out the report here: https://lnkd.in/dHiBBNdQ.

This was such an interesting read. The considerations regarding the data processing cost (one of the main considerations for Generative AI models) and data transmission costs (one of the main considerations for traditional software workloads) are relevant for several stakeholders, including developers of AI models, CTO/CIOs and big cloud companies.

Me: $ ai 4 prompt -c 'And is there an option to disable all the API discovery of kubectl, so to minimize roundtrips on slow networks? E.g. just trust the cache or builtin schema info and execute a get pods in one round trip' GPT-4:  Yes, you can minimize round trips by disabling API discovery using the `--cached-discovery` flag with `kubectl` command. This flag causes kubectl to use cached discovery information from `$HOME/.kube/cache/discovery` and ignore the real-time response from the API server for API discovery. It can reduce latency in networks with slow response times. Here's an example: ```bash kubectl --cached-discovery get pods ``` Please note, that the cache must be populated beforehand otherwise kubectl will not be able to perform commands properly, as it doesn't know the available resources and their details. Also not all kubectl commands are compatible with this flag, depending on the Kubernetes version you're using. Also remember, that if kubectl's cached information is out of sync with the actual state in the cluster, you might face errors or inconsistencies. Cache can be refreshed using `kubectl api-resources`. ---------------------------------- Wow, that's cool! Let's go! Reality: of course there is no such option...

Right after validating your AI use cases (outlined in my last post 👉 https://lnkd.in/dW_bpEEU), the next step will be selecting the best tools & techniques for the job. The problem with that? The speed of progress right now. OpenAI barely has a moat against competitors anymore, and every inch of advantage seems temporary, lasting for a few months at best. Just a few weeks ago, GPT-4 was surpassed by Claude 3 Opus on the Chatbot Arena Leaderboard, then came back on top (last refresh: 04.13. https://lnkd.in/dpqRHzCC). Meanwhile, new open source options continue to be released out of the blue, with Databricks and Cohere making lasting impressions. Continuous updates and adoption will remain mandatory for the foreseeable future, but there is a loophole to avoid vendor lock-in AND having to start over from scratch every time: A modular AI architecture allows for swapping out proprietary models for more powerful ones as they emerge, while keeping your workflows intact. You can always check which LLM is trending and considered the “best” on various benchmarks, thanks to the open-source community. Our data scientists at Hiflylabs have compiled a snapshot of our current favorites, pictured below. 🔽 But keep in mind that choosing the CURRENT best model is not the ultimate goal—rather, it’s designing a modular platform which is flexible enough to future-proof your operations, suffering no disruption from whatever fancy new toy emerges next!