Roman Kirillov

🚀 💡Gemma 2, the next generation of open AI models, is now available to researchers and developers. It offers exceptional performance and efficiency with sizes of 9 billion and 27 billion parameters. Gemma 2 supports various AI tasks and integrates seamlessly with popular frameworks like TensorFlow, PyTorch, and Hugging Face. The models are optimized for various hardware, including NVIDIA GPUs and Google Cloud TPUs, ensuring versatile applications across different projects. This release is designed to drive innovation and accessibility in AI development, providing powerful tools for creating advanced applications. https://ow.ly/fGHp50Sx3UU #Gemma2 #GoogleCloudTPUs #MachineLearning #TensorFlow #PyTorch #HuggingFace

13

1 Comment

Spotify announced generative AI powered playlists in April in beta, but only recently seems to have started mass scale rollouts. I haven't had time to try them out till earlier this week. And I was able to prompt it with "create a playlist with soulful blues and jazz which match singers like Norah Jones and Melody Gardot with the depth of Adele's voice. I'd like a deep bass rhythm as well". And it produced excellent results. I *love* genAI usecases that reduce manual toil and this is one of them that hits the mark. Now obviously, as I listen to the music, the engineer in me leads me to always dig into the how - and while I wanted to deep dive on their GenAI approach, I landed on a more fundamental post on annotation scaling that I resonated with and is actually the underpinning to great outcomes. Better annotations, and being able to scale annotations beyond pure manual tagging leads to better fine tuning which in my experience is often a critical step over and above prompt engineering when you want to actually launch into production. https://lnkd.in/epDFCxCb #genai #music #experiencesthatreallyimprove

69

1 Comment

https://lnkd.in/eCkDwU5n An inspiring and important talk about the metaphors we use to classify AI. Whether we choose to think about AI as a new order of life, or as just liquified intelligence to extend our own abilities, will change everything about the next century.

4

Which of the two would you think is the dominant data type in cloud workloads - integers / strings? Do analytical databases serve reads or writes more? Are database workloads open or closed loop? We hope to motivate new investigations into performance and features with this new paper about how customers use Amazon Redshift.

5

Switch to Claude 3.5 Sonnet on AWS Bedrock for the most versatile, innovative, cost effective, and performant LLM model out there. With this model beating almost all benchmarks, and users world over going crazy over its mind boggling coding and visual abilities, it is a no-brainer at this point. "Claude 3.5 Sonnet, the latest and most capable model from artificial intelligence (AI) safety and research company Anthropic, is generally available today in Amazon Bedrock as part of Anthropic’s Claude family of AI models. Amazon Bedrock offers the broadest selection of high-performing foundation models (FMs) from leading AI companies, along with the capabilities and enterprise security customers need to quickly build and deploy generative AI applications. Data from Anthropic shows that Claude 3.5 Sonnet sets a new industry standard for intelligence. According to Anthropic, the new model outperforms other models available today,including OpenAI’s GPT-4o and Google’s Gemini 1.5 Pro, as well as Anthropic’s own previously most capable model, Claude 3 Opus, in areas including expert knowledge, coding, and complex reasoning. Claude 3.5 Sonnet offers better intelligence and speed than Opus at one-fifth of the price. View Claude 3.5 Sonnet benchmarks."

7

1 Comment

Great product ideas can come from anywhere. Just look at Stanley Kubrick's movie 2001: A Space Odyssey. Kubrick wanted to make sure that HAL knew that it was going to be disconnected. One problem though, Kubrick didn't know how HAL could overhear the private conversation between crew members. The solution is something that we, as Agile leaders, can learn from and apply in our companies. Check out this YouTube Short for the lesson. https://lnkd.in/gyNyUKVk

14

2 Comments

Here is a quick summary of some interesting KubeCon NA announcements by Paul Burt. Here are a few highlights. - Envoy AI Gateway was announced as well as some of the reasoning behind why specialised LLM gateways are a good idea - Prometheus 3.0 is out! - Cloud native reference architectures are starting to appear. Adobe and Allianz Direct are the first two to be posted. - The CNCF wants you to help fight against patent trolls. By identifying prior art, you can help protect open source from scummy actors. - In similar news, the Open Platform for Enterprise AI (OPEA) is being donated to the CNCF by Intel. Read the breakdown below. https://lnkd.in/egZJ_gdE

28

I've been an Apple developer for the last fifteen years or so. When I say 'Apple developer', that means I've signed up for the company's developer program, downloaded all the development tools, signed up and abandoned multiple developer training programs, and have yet to write my first application. One benefit to the annual program fee is access to the developer beta software loads for iOS, OSX, and the many other flavors of beta software for the AppleTV, iPad, etc.. The key benefit, it appears, is getting to admonish myself for putting unstable code on my primary communications devices and bemoan the impact on battery life they inevitably entail. The beta iOS that dropped today, 18.1, is the first to incorporate the not-yet-quite-walking-upright Apple Intelligence features for owners of iPhones of model 15 vintage or newer. This is the first taste of mass adoption of AI from a UX perspective other than the unsexy 'plumbing' AI/ML work being done behind the wall in search, recommendation engines, and the like, and I am anxious to see how society responds to this new tool. I haven't checked Reddit yet to see if someone has created a r/appleintelligencefails subreddit, but it's inevitable that people will ask Siri to do wildly complex tasks that it will fail spectacularly at. (We can have a separate conversation on why the default assistant voice is female, and the subtle conditioning that is inculcating on a generation of users). Gen AI is in the 'stall' phase of early enterprise adopters, in that the hype is beginning to sputter in the face of enterprise users saying that they don't know how to apply it to their own business needs, and the uncertain return on investment of trials and projects. (Edward Norton voice here) "I am Christian's complete lack of surprise" as this is the same cycle that repeats itself with the newest must-have technology, from IoT to Cloud to pick-your-favorite-cybersecurity-flavor-of-the-week. A very small subset of companies have a use case in mind to apply the new tools to, and the rest wander about waiting for clarity, or (usually under pressure from an executive that wants to show relevance in the market) start an ultimately-doomed pilot project without clear goals or outcomes. The current wave of artificial intelligence hype is well founded, it will be transformative, just as previous waves of hype around AI were equally well founded in how they have transformed so many aspects of our daily lives. The predictions about major changes in the short term will come back to burn the predictors when the large-scale change doesn't happen overnight, but the long-term changes will be unprecedented across all industries. When in doubt, remember the (misattributed to Bill Gates as 'Gates' Law, but more likely from Stanford computer scientist Roy Amara in the 1960s) old adage: "we overestimate the impact of technology in the short-term and underestimate the effect in the long run".

63

7 Comments

From responding to reasoning: how the AI landscape is evolving, with OpenAI o1 on the forefront! Great and very complete article from Sequoia! Really interesting!!

7

I feel enormously privileged to be able to work alongside such talented thinkers as Demis Hassabis, John M. Jumper, and Geoffrey E. Hinton. Huge congratulations on their respective Nobel prizes. This demonstrates two things to me: the incredible talent pool at #Google and that #AI has moved beyond the hype stage and is here to create real-world impact. Their contributions lay the foundations and are a testament to the transformative power of AI in shaping our future. #google #deepmind #ai #alphafold

39

It's official! The London Observability Engineering Meetup is back for an exciting talk about scaling and managing SLOs at a global scale! 🌏 Join us on Thursday, July 11th, where Alex, Senior Site Reliability Engineer at Google, will present the evolution of Service Level Objectives (SLOs) for the GCE Compute API over the past eight years. He'll start with the initial 30 SLOs, move through a phase with around a thousand, and end with millions of per-customer SLOs. He'll share anecdotes, techniques for handling low-QPS (continuous over discrete metrics), and strategies for aggregating data to enhance leadership visibility. He'll also give practical tips for running and improving this system in production. #observability #engineering #devops #sre

8

1 Comment

Recently I spoke at Lenny Rachitsky's London meetup. Here's a slice of what I shared about hiring in tech: As an Amazon Interview Bar Raiser for 4+ years, I interviewed hundreds of candidates across PM, Marketing, Sales, Engineering, and Design roles for Amazon Web Services (AWS), Devices, and more. There's lots of talk online about the Leadership Principles, I won't rehash that here. 𝘉𝘶𝘵 𝘩𝘦𝘳𝘦'𝘴 𝘸𝘩𝘢𝘵 𝘺𝘰𝘶 𝘮𝘶𝘴𝘵 𝘬𝘯𝘰𝘸 about interviewing methodologies: Some companies evaluate candidates based on individual 𝙘𝙤𝙢𝙥𝙤𝙣𝙚𝙣𝙩𝙨 (this is how Amazon does it). Others evaluate candidates as a 𝙘𝙤𝙢𝙥𝙤𝙨𝙞𝙩𝙚 of competencies, with each interview stage being a re-evaluation, by yet another person, of multiple competencies. Have a look through the slides to see how candidate performance can lead to various hiring outcomes. Navigating this as a candidate can be incredibly challenging and stressful--it's why preparation is key in this current tech job market. Comment "video" below if you'd like a video explaining this in more depth. I'm organising next month's meetup on October 9th, exclusive for subscribers of Lenny's Newsletter. Come say hi 👋🏻

10

2 Comments

There are now three new additions to the Gemma 2 family: 1) Gemma Scope: a model interpretability tool for transparency⁠ 2) Gemma 2 (2B): a new model with an updated 2 billion (2B) parameter size⁠ 3) ShieldGemma: a suite of safety content classifiers for user protection Learn more now ↓

2

Amazon launched new state-of-the-art foundation models delivering frontier intelligence and industry-leading price performance at AWS ReInvent last week! A couple of key initiatives me and our team has been lucky to be part of, special kudos to my collaborators Xuandi FU Sai Muralidhar Kanthashree Mysore Sathyendra Srikanth Ronanki Ruchita Gupta Goeric Prudhvee Narasimha Saket Dingliwal Yilin Xiang Sai Spurthi Melanie Srini and many others, inspirational leadership from Shubham Katiyar: 𝑳𝒐𝒏𝒈 𝑪𝒐𝒏𝒕𝒆𝒙𝒕 𝑺𝒖𝒑𝒑𝒐𝒓𝒕: Amazon Nova Micro supports context length of 128K input tokens, whereas Amazon Nova Lite and Amazon Nova Pro support context length of 300K tokens, or 30 minutes of video processing. In early 2025, Amazon will support context length of over 2M input tokens. 𝐅𝐚𝐬𝐭 𝐚𝐧𝐝 𝐂𝐨𝐬𝐭 𝐄𝐟𝐟𝐞𝐜𝐭𝐢𝐯𝐞: All Amazon Nova models are fast, cost-effective and have been designed to be easy to use with a customer’s systems and data. Amazon Nova Micro, Amazon Nova Lite, and Amazon Nova Pro are at least 75 percent less expensive than the best performing models in their respective intelligence classes in Amazon Bedrock. They are also the fastest models in their respective intelligence classes in Amazon Bedrock. Please give the models a try and let us know what you think! #amazonNova #AWS #LLMs #GenAI #SOTA

148

2 Comments

Working at Google is like swimming in a sea of other people's opinions. The tsunami of opinions very quickly hit me when I joined. I was a bit off balance and put off by it at first. I found it all a bit... full on (and then some) compared to other places I'd worked. Looking back though, I can see that learning to swim in these waters has been a very good thing. I've got stronger bones and a broader outlook as a result. I can make waves of my own too 🌊, and I frequently do. Opinions can be strong currency for building your influence. People doing really well at this tend to: ✅ Assert their opinions respectfully, concisely and generally with supporting data ✅ Challenge other people's opinions with a disarming combination of good grace, sharp intellect and collaborative intent It's fair to say, plenty aren't winning at this every time 😬 But, the few who *really* are, give us lots of stellar examples we can learn from. I'm getting ready to fly to SFO tomorrow for our annual leadership summit. There'll be some leaders there with that jaw dropping capacity to effortlessly sway opinion in the room with the quality and relevance of their own. And no, they're not necessarily the highest paid people in the room either 😉. I particularly look forward to seeing them in action. Like last year, I'm setting myself some personal goals. Bringing some level of intention to how I show up tends to help shape my contribution. In a large crowd of powerful people, it's all too easy to slip into "attendee" mode. What's the point in traveling just to do that? My personal goals for 2023's summit were: ⏩ Be present ⏩ Focus outward ⏩ Listen with curiosity My personal goals this time are: ⏩ Listen with curiosity ⏩ Form and share my own opinions ⏩ Courageously challenge others' opinions, no matter whose they are I'll share my reflections throughout the trip, as I did last year. Follow me to hear more. Silicon Valley, here I come! 🛫

23

2 Comments

Capgemini’s research shows 80 percent of organizations have increased their #genAI investment in the last year. But even though gen AI is full of promise, it is tough for companies to roll out and scale projects. Everyone wants to explore innovative ways to harness gen AI and get business value but it is not always a clear path to success. Capgemini's partnered approach with Google Cloud is paying dividends for our clients. The collaboration helps us understand how best to harness the power of the technology and accelerate time-to-market. We’re innovating on gen AI with: ✅ Hackathons to drive innovation, use cases, and gen AI MVPs in a compressed time frame ✅ A Capgemini global center of excellence focused on bringing gen AI resources together ✅ Capgemini's #RAISE (Reliable AI Solution Engineering) built using #GoogleCloud native services to accelerate gen AI at scale ✅ A specialization designation for Google's gen AI platform It is one of the reasons Capgemini was named #GoogleCloud Global Industry Partner of the Year for gen AI services. Read this blog to find out how you can conquer the hurdles and find gen AI value: https://bit.ly/3XJ3pGt #genAISolutions #genAIInnovation

1

1 Comment

How Forging Partnerships Multiply Your Market Share by 50x! Aligning with the right collaborators, being in the industry for longer →Mentioning case studies where PayPal and eBay made alliances PayPal spotted a gap in eBay's payments, offered a value proposition hence driving mutual growth. →Our partnerships with Ashley Wright, starting brands almost as low as $20K GMV to scaling 50x! For a Single Brand →Harnessing the power of 2300 affiliates it’s not just possible; it’s happening! →Just like how Pure Peak used our bot to pass $10M in sales, showcasing tech's power. Start leading the market! Comment "Lead" and access your exclusive trial before you decide!

4

1 Comment

I ran across this post in the Changelog newsletter: https://lnkd.in/eCdjVekq - I like the name slop for unwanted AI-generated content. I love LLMs. They are part of my daily workflow, and I've been exploring them more seriously while on my sabbatical. I wouldn't misrepresent generated content without critical review as my own, but the web seems to be filled with more generated garbage (or 'slop')—content that looks fine at first but isn't accurate or helpful. LLMs (and other generative AI) should help us to make things of higher quality, not just churn out slop more quickly.

11

1 Comment

Principal Engineer Staff Engineer Distinguished Engineer (my favourite) I’m going to say it and it’s likely a little rhetorical: I dislike all these terms immensely. Please can someone give me the canonical description for what they are supposed to mean? (I know I can use a search engine or AI to do it… I want to know what people think they mean). The industry I work in is in danger of disappearing up its own backside. What’s wrong with: Software engineer Why do we need to differentiate? The only reason I can see is signposting to people with differing levels of experience or different sets of responsibilities. If we need to do that then, I’m not sure our teams are set up correctly in the first place. #software #softwareengineering #softwaredevelopment #productdevelopment

96

105 Comments