📈 How are companies using AI?

➜ Typetone, a Dutch genAI content agency, had trouble finding the right people for their team so they built those people – Sarah is allegedly the first AI Digital Worker of her kind, it can do market research, create strategies, run content operations from scratch to published content, and it improves over time.

LinkedIn

➜ The NBA is introducing ‘movie mode’, a genAI-powered feature that transforms the game into a movie-like experience of your choosing in real-time – for example, it can turn the game into a comic book movie for you.

X (Twitter)

➜ NICE Actimize, the largest provider of financial crime, risk and compliance solutions for finance and gov, has introduced three genAI-based solutions – two of them designed for financial institutions, they promise to cut manual labor time by up to 70% for specific activities, and one is a co-pilot for fraud analysts.

NICE Actimize

➜ Fleet management company Geotab introduces Geotab Ace, a co-pilot that makes complex datasets straightforward, simplifies data exploration, and democratizes access to data by enabling data analysis with natural language.

Geotab

➜ Riskonnect, an integrated risk management provider, has boosted their Business Continuity & Resilience solution with genAI – it streamlines the creation and updating of business continuity plans, business impact analyses, and incident templates, significantly reducing the time required to draft or update such docs.

Riskonnect

➜ Penny is Priceline’s AI co-pilot for travelers – the company has introduced over 30 new features to its Trip Intelligence suite, which will save their users time and money in planning and booking travel, including such solutions as itinerary creation or 24/7 AI assistance.

Priceline

➜ In collaboration with Infosys, the Australian Open is using genAI extensively to improve the game experience for players, coaches, and fans.

Infosys

➜ Among other AI enhancements, Otter AI Chat introduces "AI Chat in Channels," offering instant access to past discussion insights in a channel, and generation of summaries and other content – the goal is to let you ”tap into the collective wisdom of ALL your past meetings, regardless of platform.”

Otter 1, Otter 2

➜ Available as a paid add-on for English Enterprise plans, Slack AI is here, and it can save workers 97 minutes per week on average – it offers personalized search answers, channel recaps, thread summaries, personalized response generation, with full data control and privacy, adhering to high security and compliance standards.

Slack

➜ Local AI computing is clearly the way of the future, and Nvidia is at the forefront – Chat With RTX is their new demo app that lets you personalize a genAI model connected to your own content (docs, notes, videos, or other data), provides answers to your questions based on the context of your machine, and runs 100% locally on your Windows RTX PC or workstation.

Nvidia

➜ Seems like the beginning of a new business model – similarly to the GPT Store, where creators of custom ChatGPTs can earn extra money, Eleven Labs now lets you add a 30-minute sample of your voice and earn commissions when anybody uses your voice for their project.

X (Twitter)

💡 What are the nerds up to?

➜ Finally, some good f*****g genAI video – until now, almost every genAI video product has been disappointing, but OpenAI came through and delivered the first actually impressive solution, Sora (well, almost, it’s in early access and not publicly available yet). It’s still not 100% perfect, but it’s the closest to being perfect that the industry has seen so far.

OpenAI

➜ Aya is a “massively multilingual model” which covers 101 languages – trained on the most extensive assembly of multilingual instruction fine-tuning datasets to date, open-source, it achieves state-of-the-art performance across many multilingual benchmarks.

Cohere

➜ A startup that’s taking on Google DeepMind and winning – WindBorne Systems has achieved a breakthrough with WeatherMesh, its AI forecasting model, delivering the most accurate global weather forecasts to date. Capable of running on a gaming laptop, the model relies on data from WindBorne’s autonomous balloon network for global atmospheric data collection.

WindBorne

➜ Google has open-sourced Magika, an AI-powered system for precise file-type identification – it can be used via a command-line tool or a Python library, relies on a deep-learning model for quick and accurate detection of both binary and textual file formats, works well even on CPU-based systems, outperforms existing tools by about 20%, and it boosts malicious document detection on platforms like Gmail and Drive by 50%.

Google

➜ OS-Copilot is a framework to build generalist agents that can autonomously do most of the things that humans can do on computers – an example agent, FRIDAY, outperforms other methods by 35%, with a strong ability to adapt to unseen applications, and learns to control and self-improve on Excel and Powerpoint with minimal supervision. 

OS-Copilot

➜ The Large World Model (LWM) – trained on text and video, it’s one of the largest context size transformers that sets new benchmarks in long video understanding, simplifies vision-language training, and comes as a fully open-sourced family of 7B parameter models capable of processing long text documents (LWM-Text, LWM-Text-Chat) and videos (LWM, LWM-Chat) of over 1M tokens.

GitHub

Thanks for reading!

I share these stories because I believe it's important for all of us to keep up with AI. To support my mission, share this newsletter on LinkedIn.