Each week, we bring you 5 stories that resonated the most in our internal Slack channel #AI-news. We write the newsletter using various AI tools because we're an AI company and our marketing wants to move with the times too. ๐
Today you're reading the 65th issue in a row.
#1
Memory update ๐ง : ChatGPT can now remember your past conversations ๐๐ก!
OpenAI has launched long-term memory for ChatGPT ๐ง , which allows it to remember things from previous chats โ even if you didnโt manually save them โ๏ธ. This means it can give you more personalized ๐ and better-tailored responses ๐ฏ over time.
It builds on last yearโs โMemoryโ feature ๐๏ธ, but now it also includes automatic reading from your chat history ๐. This feature is optional โ and can be turned off anytime ๐.
Itโs currently being gradually rolled out to Pro users with a $200/month subscription, while those with a Plus plan for $20 and businesses ๐งโ๐ผ will get access soon. However, itโs not yet available in the EU or in other countries ๐ with stricter regulations โ๏ธ.
ย
#2
Agent2Agent ๐ค: A new standard for AI agent communication ๐ก.
Google has introduced Agent2Agent (A2A) ๐ค๐, a new communication protocol for AI agents designed to ensure they can clearly and securely exchange information and collaborate ๐๐ฌ โ even if they come from different companies ๐ข. As AI tools become more common in businesses today ๐ง ๐, itโs important that these systems can understand each other ๐.
More than 50 partners ๐ค participated in the development of A2A, including companies like Salesforce, Intuit, and LangChain. The protocol sets shared rules for exchanging tasks, context, and data. Itโs based on standard formats like JSON ๐งพ, itโs open source ๐, and it works even when agents donโt share memory ๐ง or use the same tools ๐ ๏ธ.
This protocol fits into a broader trend in the AI world ๐๐ค. Similar efforts include MCP from Anthropic and AGNTCY, backed by Cisco ๐ฐ๏ธ. Google says that A2A is not a replacement for these solutions but can work alongside them โ๏ธ. The real challenge now? Making sure as many players as possible adopt these shared standards ๐งฉ๐.
ย
#3
IEA forecast ๐: By 2030, AI data centers could consume as much electricity as all of Japan ๐ฏ๐ต๐พ.
According to the latest report from the International Energy Agency (IEA) ๐๐, artificial intelligence could more than double global electricity consumption in data centers by 2030 ๐พ โ reaching the level of electricity currently used by all of Japan. In the U.S. ๐บ๐ธ, AI data centers might soon consume more electricity than the entire heavy industry sector combined ๐ญ. This kind of growth could strain power grids, increase emissions, and intensify demand for rare minerals ๐ชจ used in AI hardware ๐ฅ๏ธ.
But thereโs also another side to AI โ๏ธ โ it can help reduce emissions, improve energy efficiency ๐ก, and accelerate the development of technologies like batteries and solar panels โ๏ธ. That makes it a tool that can be part of the solution, not just the problem ๐.
The report calls on governments ๐๏ธ to act quickly โฑ๏ธ: modernize the electricity infrastructure, improve data center efficiency ๐ง ๐พ, and better align energy and tech policies ๐งฉ. Otherwise, the rapid growth of AI might outpace our ability to keep it running without unnecessary strain on the planet ๐โ ๏ธ.
ย
#4
Google adopts Anthropicโs standard for connecting AI with data ๐ง ๐.
Google is adding support for the Model Context Protocol (MCP) ๐โcreated by Anthropicโto its Gemini models ๐ง โจ and developer SDK. With this move, Google joins OpenAI ๐ค, which already supports the protocol. MCP is an open standard ๐ that allows AI models to directly access data from various business tools, databases, and apps ๐ฑโand even work with them directly โ๏ธ.
Thanks to MCP, AI applications like chatbots ๐ฌ or automation tools become much more practical โ . They can now use up-to-date information ๐ก instead of relying solely on previously โhard-codedโ knowledge. This opens the door to much more accurate answers ๐ฏ and better interactivity ๐.
Since MCP was made open source ๐, companies like Block, Apollo, Replit, and Sourcegraph ๐๏ธ have joined the initiative. Now that Google supports it too, it seems MCP is quickly becoming the new standard for how AI systems interact with real-world data ๐๐ง .
ย
#5
Netflix ๐ฅ: Testing AI search powered by OpenAI ๐ค๐.
Netflix is testing a new search tool ๐ that uses technology from OpenAI ๐ค. This tool lets you search for shows and movies ๐ฌ using very specific detailsโlike your mood ๐ or personal preferences ๐ฏ. Instead of typing in traditional genres or actor names, you can write something like โsomething light and funny after a long dayโ ๐๏ธ, and the system will suggest titles that match โ .
For now, itโs an optional test ๐งช available only to selected iOS users in Australia and New Zealand ๐ณ๐ฟ. Netflix plans to expand testing to the U.S. ๐บ๐ธ in the coming weeks or months. Since this is an early beta version ๐งฑ, the company is currently collecting feedback ๐ฌ and fine-tuning the features โ๏ธ.
This move shows how streaming services ๐ฅ are trying to better adapt to users' individual needs ๐ค. With this test, Netflix is exploring how to make content discovery more enjoyableโand less of a guessing game ๐ฒ.
ย
