Nordic Data & AI Weekly: Google's launch of Gemini, ChatGPT's 1st birthday, a big step forward for the EU AI Act, various takes on how to build AI applications, & all recent Nordic investment rounds
Recent funding, news, and resources for builders
Quick reads on value creation
Things to know, whether you are building 🛠 or investing 💰
💡 Greylock Partners on why a vertical approach is key to building enduring AI applications.
💡 A great read from NFX - looking at the 5-level AI application spectrum, AI in horizontal markets, a deeper look at AI leapfrogging, the AI-first app layer, and their lens for evaluating AI companies.
💡 a16z shared their predictions and “Big ideas in tech 2024”.
💪 Humanloop with an opinion on “how to build the right team for generative AI”. Key takeaways include the idea that you probably don’t need ML engineers, product managers and domain experts are increasingly important, and generalist full-stack engineers can outperform AI specialists.
📃 The state of generative AI in SaaS.
Recent AI funding in the Nordics & Baltics
🇸🇪 🇳🇴 🇩🇰 🇫🇮 🇪🇪 🇱🇻 🇱🇹 🇮🇸
🇸🇪 FinetuneDB: Easy to use fine-tuning platform to improve any LLM - create datasets collaboratively and monitor performance. Raised an undisclosed amount from Antler.
🇸🇪 bifrost security: Behaviour-based adaptive security for cloud software, securing your containerised software with tailor-made security profiles that continuously adapt to every next build. Raised $901k from LU Innovation, Quinary Investment, Fast Track Capital, and angels (link).
🇸🇪 Prompt Studio: No-code tools allowing domain experts to build and test AI product features, faster. Raised $108.1k from Heartfelt (APX).
🇸🇪 Superagent: The open framework for building AI-assistants - make your applications smarter and more capable with AI-assistants, and build unique ChatGPT-like experiences with custom knowledge, brand identity, and external APIs. Raised $500k from Y Combinator.
🇸🇪 Vesiro: Database optimisation software enhancing search performance in order to reduce server costs & energy consumption. Raised $450k from First Gate Invest, Chalmers Ventures, Länsförsäkringar, Norrsken VC, and E14 Invest (link).
🇸🇪 Playmaker AI: Football analytics made easy - get access to AI-evaluated football data from hundreds of leagues around the world, always up to date, always available – simple, understandable and meaningful. Raised $363.7k from undisclosed investors.
🇪🇪 humbl.ai: Modular suite of tools that use automation and AI to empower iGaming and e-commerce clients in the areas of paid media, affiliation, SEO, and compliance. Raised $291.87k from undisclosed investors.
If there is some relevant Nordic AI news or funding that is missing from this newsletter, feel free to share by replying directly or emailing nordicaiweekly@substack.com and I’ll look to spread the word next week. Thanks!
News from the wider world 🌍 🌎 🌏
🆕 Google launched Gemini - its largest and most capable AI model, trained on text, images, and audio. Launch video here. Further demo video here. And from TechCrunch.
Here’s what Google wants you to know:
Gemini comes in three sizes: Gemini Ultra (for highly complex tasks), Gemini Pro (for scaling across a wide range of tasks), and Gemini Nano (for on-device tasks).
The benchmarks presented by Google suggest that Gemini outperforms GPT-3.5 in six of eight benchmarks, and surpasses GPT-4 in 30 out of 32 benchmarks, demonstrating its superior multimodal understanding and interaction. So this represents the first real challenge to OpenAI’s dominance. Full technical report here.
The biggest breakthrough seems to be that Gemini models are trained on multimodal data from the ground up.
And of course, with Google’s products and vast distribution, we can expect to see these models rolled out in Bard (Gemini Pro available now, Gemini Ultra from the start of the year), and mobile devices (Gemini Nano now running on the Pixel 8 Pro).
It’s not all good though… some less positive early impressions on TechCrunch, while it has been widely suggested that the demo video was faked - see social media, and more from TechCrunch, while according to Bloomberg, the video demo “wasn’t carried out in real time or in voice.“
Aravind Sriniva, CEO of Perplexity AI, with the more balanced view below…
🎂 ChatGPT turned one year old, here with a look at how it changed the world. And in other OpenAI news…
📹 Pika Labs introduced Pika 1.0 - the idea-to-video platform that brings your creativity to life, enabling you to create and edit videos with AI. It’s a pretty impressive leap forward. The company also announced a $35m Series A led by Lightspeed Venture Partners.
💰 Mistral - the French AI company and rival to OpenAI - is reportedly closing in on a €450m funding round that will value the company at $2bn.
💰 In the quest to challenge OpenAI and the rest, Elon Musk’s xAI raised $135m and wants to raise up to $1bn.
🆕 Amazon came with a few big announcements…
Amazon Q - “a new generative AI–powered assistant that is specifically designed for work and can be tailored to your business to have conversations, solve problems, generate content, and take actions using the data and expertise found in your company’s information repositories, code, and enterprise systems.” More here.
A new text-to-image AI model called Titan, which can produce “realistic, studio-quality images.” Unlike other AI image generation apps like ChatGPT and Midjourney, Titan won’t be available as a standalone app. Instead, companies will be able to integrate Titan into their apps and websites to generate images.
💎 Researchers at Google DeepMind and Lawrence Berkeley National Laboratory have developed a new AI system that has discovered over 2 million new materials that could be used in batteries, solar panels, and computer chips.
On safety, regulation, and sustainability ⛑ 👩🏻⚖️ 🌎
🏭 Making an image with generative AI uses as much energy as charging your phone.
🇪🇺 The EU agreed a political deal on a risk-based framework for regulating artificial intelligence, meaning, after being originally proposed back in April 2021, and after months of negotiations, a pan-EU AI law is definitively on the way. Politicians were very happy to share the news here and here, but many doubt whether these regulations will be able to keep up with the technology's rapid pace of development.
✍️ The United States, Britain and more than a dozen other countries unveiled an agreement on how to keep AI safe from rogue actors, pushing for companies to create AI systems that are "secure by design."
✍️ IBM and Meta have launched an international AI Alliance along with over 50 other organisations to advance open and responsible AI innovation.
🔬 Sequoia with a look at what physics can tell us about AI risk, highlighting the fundamental limitations of AI systems.
If you know someone that would be interested in the contents of this newsletter, please forward it onwards.