Welcome to Get The Gist, where every weekday we share an easy-to-read summary of the latest and greatest developments in AI—news, innovations, and trends—all delivered in under 5 minutes! ⏱

In today’s edition:

• Meta Unveils Multimodal Language Model ‘Meta Spirit LM’

• Perplexity AI Targets $8 Billion Valuation Amidst AI Boom

• Microsoft Announces Autonomous Copilot AI Agents

• IBM Launched Granite 3.0 LLM

• And more AI news….

Top Developments

Meta

Meta Unveils Multimodal Language Model ‘Meta Spirit LM’

The Gist: Meta has launched Meta Spirit LM, an open-source language model designed to seamlessly integrate speech and text, enabling more natural-sounding speech generation and handling tasks across different modalities.

Key Details:

• Meta Spirit LM improves text-to-speech (TTS) and automatic speech recognition (ASR) by combining speech and text datasets for cross-modality generation.

• Two versions: Spirit LM Base (with phonetic tokens) and Spirit LM Expressive (adds pitch and style tokens to convey tone and emotion).

• It can handle tasks like ASR, TTS, and speech classification, aiming to inspire advancements in speech and text integration.

Perplexity AI

Perplexity AI Targets $8 Billion Valuation Amidst AI Boom

The Gist: Perplexity AI, a generative AI search engine startup backed by Nvidia, is seeking to raise $500 million in a new funding round, aiming for a valuation of over $8 billion amid growing investor interest in AI.

Key Details:

• Perplexity AI's valuation has surged from $520 million in January to a projected $8 billion.

• The company generates revenue through consumer subscriptions and enterprise solutions, with annual revenue rising from $10 million in March to an estimated $50 million.

• It plans to introduce advertisements in AI-generated answers to further diversify revenue.

Microsoft Copilot

Microsoft Announces Autonomous Copilot AI Agents

The Gist: Microsoft is expanding the capabilities of its autonomous Copilot agents, which can now automate complex tasks like email responses and workflow management. The new agents will be publicly previewed at the November Ignite conference.

Key Details:

• Autonomous agents can handle tasks across Microsoft environments without needing extensive AI training.

• The agents extract key data from emails, match it to relevant team members, and can escalate issues to humans if needed.

• Custom agents can be created using Copilot Studio, enabling businesses to tailor behaviors and functions.

• Ten new agents have been introduced for Dynamics 365, including tools for lead qualification and supply chain management.

Quick Gist

IBM launched Granite 3.0, its third generation of enterprise AI large language models, aiming to enhance performance, safety, and open-source accessibility for business applications (Read More).

Meta unveiled a self-taught evaluator AI that trains other AIs without human input, aimed at enhancing efficiency and accuracy in large language model evaluations (Read More).

AI researcher Simon Willison successfully used video scraping with Google's Gemini AI to extract data from screen recordings, highlighting the potential for AI models to interact visually with user screens while raising privacy concerns (Read More).

The new Mixtral model outperforms Llama 2 70B in various benchmarks with significantly lower active parameters, showcasing its efficiency and effectiveness in commonsense reasoning, math, and code tasks (Read More).

OpenAI's new ChatGPT desktop app allows users to instantly engage with the AI via keyboard shortcuts, utilize voice conversations, and integrate file uploads, while also offering functionalities for screenshot discussions and model training controls (Read More).

Apple is expanding its artificial intelligence capabilities with new features in iOS 18, aimed at enhancing user personalization and predictive functionalities (Read More).

A Google DeepMind executive advocates for consistent safety standards in AI development following the veto of California's sweeping AI safety legislation (Read More).

X is updating its terms of service to allow user posts to be used for training AI models, introducing potential fines for excessive usage and changes to the block feature (Read More).

Adobe launched a beta version of new Firefly-powered video editing workflows in Premiere Pro, introducing tools that simplify video and audio clip extension for professional editors (Read More).

THAT’S IT FOR TODAY!