🎙️Each episode of AI Assembled is a curated collection of the most important AI news stories that apply to you. Chris Doelle compiles the best stuff from dozens of daily AI announcements, and uses AI to create a custom podcast.

🤖The hosts you hear are all AI generated!

• OpenAI’s ChatGPT is rolling out a new feature allowing it to remember past conversations for pro users initially. This “greatly improved memory” will enable more personalized AI companions by referencing your entire chat history.
• This gets around the short context window limitation. Potential benefits include more efficient research assistants and coding buddies.
• This feature isn’t available in the EA, UK, Switzerland, Norway, Iceland, or Likenstein yet, likely due to data privacy regulations.
• Users will have options to opt out of the memory feature or use temporary chats for conversations they don’t want to be remembered.
• Reportedly, OpenAI has closed a funding deal valuing the company at $300 billion, nearly doubling in just six months. This shows incredible investor confidence in AI’s future and OpenAI’s leading position.
• This valuation is driven by the long-term vision for AI, rapid progress, new applications, Gen AI becoming embedded across industries, and OpenAI’s enterprise focus. This funding provides significant resources for R&D, talent, and infrastructure.

Google’s Agent 2 Agent Protocol (A2A):

• Google unveiled the Agent 2 Agent Protocol (A2A), a proposed open standard for AI agents to communicate securely, share information, and coordinate tasks across different company systems.
• A2A aims to create a more connected AI agent ecosystem, allowing agents built by different companies to cooperate. This could lead to a new level of automation, for example, in customer service.
• This is a collaborative effort with over 50 tech partners and service providers involved, including Atlassian, Salesforce, SAP, Accenture, and Deloitte.
• A2A complements Anthropic’s MCP by focusing on how capable agents interact, while MCP focuses on improving context for a single model.
• The design principles of A2A include embracing how agents work, building on existing web standards, security first, support for long tasks, and modality agnostic handling.
• Interactions involve a client agent asking a remote agent to perform a task, discovering capabilities via “agent cards,” managing the task through messaging, and even collaborating on user experience.
• A candidate sourcing example illustrates how an AI agent could use A2A to ping other specialized agents on platforms like LinkedIn and GitHub to quickly curate a list of candidates.

Amazon’s AI Focus (Andy Jassy’s Shareholder Letter):

• Andy Jassy’s shareholder letter heavily focused on AI, stating that Gen AI will reinvent pretty much every customer experience.
• Amazon is already using Gen AI internally for productivity, cost savings, customer service, and workflows, but sees much wider applications.
• Jassy expects AI to transform coding, search, shopping, personal assistance, and healthcare, emphasizing the speed of these changes.
• Amazon is investing massively, with over a thousand Gen AI applications being built across the company.
• AWS provides the building blocks, including custom Tranium 2 chips for training, Sage Maker and Bedrock for building models, Amazon Nova models, and tools for creating agents.
• Jassy is optimistic that AI costs will decrease, highlighting the better price performance of Trainium 2 for training and the importance of reducing inference costs.
• Alexa Plus is positioned as a major upgrade, as smart as top chatbots but also capable of real-world actions like playing music, controlling devices, ordering, and making reservations. With over 600 million Alexa devices, this could become a super intelligent personal assistant.

Adobe’s Agentic AI Vision:

• Adobe believes AI should empower humans by taking away grunt work, allowing them to focus on creative tasks.
• They emphasize that AI agents are tools for sparking ideas and speeding up content creation, not replacements for human creativity.
• Examples include Acrobat analyzing documents for sales pitches and Express generating animated flyers.
• For professionals using Photoshop and Premiere, the goal is to integrate AI agents into the workflow for tasks like suggesting edits in Photoshop or assisting with rough cuts in Premiere Pro.
• Features like distraction removal in Photoshop and media intelligence in Premiere are steps towards more comprehensive agent help.
• The upcoming actions panel in Photoshop will analyze images and recommend context-aware edits, moving towards software that anticipates user needs.
• Media intelligence in Premiere will allow agents to assist with more complex video editing decisions.

Thinking Machines Lab (Mira Murati’s Startup):

• Mira Murati, former CTO of OpenAI, has a new startup called Thinking Machines Lab.
• The company is reportedly trying to raise a $2 billion seed round, potentially valuing it at $10 billion or more, despite being in early stages with no product or revenue.
• This extraordinary valuation is based on the talent being assembled, including top AI researchers and advisors from OpenAI like Bob Mcgru and Alec Radford.
• It’s a bet on their experience and expertise to push beyond current AI limits, aiming for more widely understood, customizable, and capable AI.

Other Notable Developments:

• Accenture and Google Cloud are deepening their partnership to accelerate enterprise AI adoption, focusing on customer experience, mainframe modernization, and edge computing.
• Startup Gen Layer raised $7.5 million to build trust infrastructure for autonomous AI agents using blockchain technology.
• Their concepts include intelligent contracts (AI-powered agreements) and optimistic democracy (a consensus system using multiple LLMs as validators).
• Gen Layer is launching a test net exploring AI-run supply chains, influencer marketing, DeFi, DAO, and AI insurance, aiming to enable new autonomous AI business models.

This week’s news highlights the rapid and multi-directional advancements in AI, making it more intelligent, connected, personal, and autonomous.

If you would rather get this as an email newsletter, head here: https://www.freshmediaworks.com/assembled