Cohere sees valuation soar to $5.5B after new funding round (2 minute read)
Cohere, a Canadian AI startup focused on large language models, has reached a $5.5B valuation following a $500M Series D funding round. The new funds aim to enhance its enterprise-grade AI tech for global business efficiency. Key investors include PSP Investments, Cisco, Fujitsu, AMD Ventures, and EDC.
Figma AI Update (3 minute read)
Figma briefly rolled back its 'Make Designs' AI feature from its limited beta after realizing it generated UI designs that too closely resembled existing apps. The feature uses off-the-shelf AI models like GPT-4 and Amazon's Titan and it needs refinement to ensure originality. Figma aims to re-enable the feature with improved QA processes to better assist designers in leveraging AI for efficient design creation.
ElevenLabs Turbo 2.5 model (2 minute read)
ElevenLabs has introduced a new model, Turbo 2.5, that unlocks high-quality low-latency conversational AI for nearly 80% of the world - Hindi, French, Spanish, Mandarin, and 27 other languages. For the first time, it supports Vietnamese, Hungarian, and Norwegian text-to-speech. English is now 25% faster compared to Turbo v2.
Inside the fight over California's new AI bill (6 minute read)
California state Senator Scott Wiener's Safe and Secure Innovation for Frontier Artificial Intelligence Models bill requires companies training 'frontier models' that cost more than $100 million to do safety testing and be able to shut off their models in the event of a safety incident. The bill has received heavy criticism from the tech industry. It will affect everyone who does business in California, not just companies that develop their models in California. This article contains an interview with Wiener about the bill and its critics.
Using LLMs For Evaluation (33 minute read)
As large language models become more capable, evaluating them accurately remains challenging due to their diverse tasks. While human evaluation is the gold standard, it's costly and slow. Using LLMs themselves as evaluators offers a scalable, cost-effective alternative, despite potential biases such as positional and verbosity bias, which can be mitigated through techniques like randomizing output positions and using multiple evidence calibrations.
Three Archetypes of AI Application Startups (7 minute read)
AI applications are evolving into three common archetypes: AI Copilots that assist with tasks, AI Colleagues that autonomously manage specific tasks alongside human workers, and AI-Native Services which offer end-to-end services that integrate AI with human input. GitHub Copilot and Devin exemplify AI Copilots and Colleagues, aiding in coding and engineering, respectively. AI-Native Services, such as Pilot for bookkeeping, compete with traditional service firms by offering automated solutions in areas like law and accounting.