Today on Don’t Fear AI

• From Oil to AI: Saudi Arabia’s $940B Vision 2030 Bet on Humain

• Why Klarna’s AI Pivot Failed: Klarna Rehiring Humans

• Audible Embraces AI: A New Chapter for Audiobooks and Translations

• Transformer Lab: 100% Open Source Toolkit for Large Language Models

From Oil to AI: Saudi Arabia’s $940B Vision 2030 Bet on Humain

Saudi Arabia has launched Humain, a new artificial intelligence company under the Public Investment Fund (PIF), as part of its Vision 2030 strategy to diversify its economy and become a global AI leader. Chaired by Crown Prince Mohammed bin Salman, Humain will operate across the entire AI value chain, developing advanced technologies such as Arabic-language large language models (LLMs), AI infrastructure, and next-generation data centers.

Key points:

• Strategic Goal: Humain aims to build a world-class AI ecosystem, reduce reliance on imported tech, and foster innovation across sectors like healthcare, logistics, defense, and education.

• Global Positioning: With $940 billion in PIF backing, Saudi Arabia wants to shape the global AI landscape and ensure the Arabic-speaking world is represented.

• International Collaboration: Launch timing aligns with the Saudi-US Investment Forum, reinforcing Saudi Arabia's openness to international partnerships. The kingdom plans to invest $600 billion in the U.S. over four years.

• Business Implications: For international businesses, especially in tech, cloud, data, and education, Humain presents significant investment and partnership opportunities. However, success requires understanding local regulations and cultural dynamics.

• Global Recognition: Saudi Arabia topped the 2024 Global AI Index in government AI strategy.

• Vision 2030 Alignment: Humain is a cornerstone of the Kingdom's shift from oil-based to innovation-driven economic growth, targeting job creation, IP development, and a knowledge-based economy.

In essence, Humain is both a national AI engine and an international invitation to collaborate, invest, and innovate in Saudi Arabia's growing tech landscape.

Saudia PIF article

Why Klarna’s AI Pivot Failed: Klarna Rehiring Humans

After heavily promoting itself as an “AI-first” company, Swedish fintech Klarna is reversing course and rehiring human customer service agents due to the shortcomings of AI-led support. CEO Sebastian Siemiatkowski admitted the company had over-prioritized cost savings through AI, which ultimately led to a decline in service quality. Klarna had previously claimed its AI chatbot powered by OpenAI replaced 700+ agents, handling millions of chats across 35+ languages. In 2023, it paused all hiring and cut headcount by 22%.

Now, Klarna is launching a pilot program to reintroduce human agents via a flexible, gig-style model, particularly targeting students, rural workers, and passionate customers. The move reflects a broader industry pattern: an IBM survey found only 1 in 4 AI projects deliver expected ROI, and just 16% are scaled. Despite aggressive AI adoption, many companies are encountering quality and trust issues, with customers still demanding human interaction

Experts note that AI should enhance not replace human service, emphasizing that poor service experiences can lead to significant customer loss. Klarna maintains it is still committed to AI, with the bot now doing the work of 800 roles, but acknowledges that “investing in quality human support” is now part of its future strategy.

Link to article

Audible Embraces AI: A New Chapter for Audiobooks and Translations

Amazon’s Audible has announced the rollout of its new AI-powered audiobook production and translation tools, aiming to make more books accessible in audio format and multiple languages. The technology will be offered to select publishing partners via two options:

1. Audible-managed, end-to-end production

2. Self-service tools for publishers to produce audiobooks independently.

Key features include:

• Access to 100+ AI-generated voices in English, Spanish, French, and Italian.

• Voice upgrades over time as the technology evolves.

• Beta launch of AI translation tools later in 2025, enabling both text-to-text and speech-to-speech translation. Human linguist review will be available to ensure translation quality.

Audible believes AI can expand audiobook availability globally, aligning with its vision of providing “every book in every language.”

Link to articles

Transformer Lab: 100% Open Source Toolkit for Large Language Models

If you're delving into large language models (LLMs) and seeking a comprehensive, open-source solution for training, fine-tuning, and deploying models locally or in the cloud, look no further than Transformer Lab.

🔧 Key Features:

• One-Click Model Downloads: Access hundreds of models from Hugging Face, including DeepSeek, Llama 3, Mistral, Phi-4, and Gemma.

• Interactive Chat Interface: Engage with AI models, adjust responses, and save conversations seamlessly.

• Training & Fine-Tuning: Utilize platforms like Apple MLX, Hugging Face, and GPUs for model training.

• Reinforcement Learning from Human Feedback (RLHF): Implement techniques like DPO, ORPO, and SIMPO.

• Dataset Management: Build and manage datasets with an intuitive drag-and-drop UI.

• Model Conversion: Easily switch between formats such as Hugging Face, MLX, and GGUF.

• Plugin Support: Extend functionality with built-in or custom plugins.

• Local or Cloud Deployment: Run models on your machine or deploy to remote/cloud servers.

Compared to tools like LM Studio, Transformer Lab offers enhanced control over model training, experimentation, and deployment, making it a versatile choice for developers and researchers alike.

Link to repo