Google Announced 🧞‍♂️Genie: Generative Interactive Environments

DeepMind just unleashed something straight out of sci-fi with 'Genie,' an AI that's literally crafting interactive video games out of thin air. Imagine pulling a game out of a single image or prompt – that's Genie for you, powered by a mind-blowing 200,000 hours of video game footage from the wilds of the internet, all without a hint of supervision.

Here's where it gets wild: Genie doesn't need anyone to tell it what's what in a game. No action labels, no text to guide it; it figures out who the hero is all on its own and hands over the reins for you to dive into these freshly minted virtual realms. The magic behind this? A cocktail of tech wizardry involving a Latent Action Model, a Video Tokenizer, and a Dynamics Model – and trust me, we'll dig into those goodies in tomorrow’s deep dive.

But hold your horses, it's not game time for everyone just yet. This groundbreaking tech is still in the lab coat phase, courtesy of Google DeepMind, so no public playdates (cue the eye-roll memes). And sure, it's currently rolling games at a nostalgic 1FPS, but remember, this is just the starting line. The potential here is bonkers – we're on the brink of a world where anyone can whip up their own digital dreamscapes. Buckle up; we're just getting started in this wild, wild world of AI-generated gaming.

Mistral AI releases new model to rival GPT-4 and its own chat assistant

Alright, folks, buckle up because Mistral AI out of Paris is shaking things up big time in the AI playground. They're not just stepping into the ring; they're looking to throw some serious punches at heavyweights like OpenAI and Anthropic. With a shiny new toy called Mistral Large and a chatty sidekick, Le Chat, they're rolling out the red carpet for anyone itching to dive into some deep conversational wizardry.

This isn't your average garage band startup. We're talking about a dream team of brainiacs from the elite leagues of Google's DeepMind and Meta, who've just bagged a cool $415 million, with Andreessen Horowitz writing the checks. They've got their eyes on the prize, pivoting from their open-source jam to a more OpenAI-style gig, slinging access to Mistral Large with a price tag, ready to rumble with the likes of GPT-4 and Claude 2 in a battle of wits across languages from English to Italian.

And then there's Le Chat, Mistral AI’s answer to ChatGPT, currently strutting its stuff in beta, free for the masses with flavors ranging from Mistral Small to the sleek Mistral Next. But don’t get too comfy with the freebies; they've got plans to roll out the red tape for enterprise folks, packing in all the bells and whistles from central billing to tailor-made moderation tools. Plus, they've cozied up with Microsoft, slotting their AI magic into Azure's cloud kingdom, which could seriously spice up Mistral AI's game, giving them a leg up in the sprawling AI market while keeping Azure's clientele snugly in Microsoft's embrace. So yeah, Mistral AI isn't just making waves; they're whipping up a storm in the AI world.

Related: Microsoft strikes deal with Mistral in push beyond OpenAI

Xiaomi 14’s AI Portrait feature is equal parts amazing and creepy

The Xiaomi 14 series introduces an innovative AI Portrait feature, revealed at MWC 2024, that allows users to insert photo-realistic versions of themselves or others into generated images, provided they have at least 20 photos for model creation. This feature, alongside a generic AI image generator, enables the crafting of personalized scenes, such as "snowy mountains" or "at the beach," with the user's AI model taking center stage. The inclusion of such technology in smartphones highlights the evolving capabilities of AI in personalizing user experiences, albeit raising ethical concerns regarding the potential misuse for creating deceptive imagery of individuals.

However, the excitement surrounding Xiaomi's AI Portrait comes with its share of apprehensions about privacy and the ethical implications of generating fake images, particularly of celebrities and public figures. The possibility of misuse by malicious actors underscores the darker aspects of this technology, despite assurances that content restrictions will be in place to prevent the creation of offensive imagery. This dual nature of AI Portrait—its capacity to delight and disturb—underscores the complex relationship society has with advancing AI technologies.

The Xiaomi 14 and Xiaomi 14 Ultra, priced at €999 and €1,499 respectively in Europe, bring not only AI Portrait but also features like AI image expansion, subtitling, translation, and photo search. While these devices will not be available in the U.S., their introduction marks a significant step in smartphone innovation, offering an alternative for those seeking new features beyond what is offered by established brands like Samsung, Apple, and Google. The integration of advanced AI features in Xiaomi's latest offerings highlights the growing trend of AI in enhancing mobile user experiences.

How to Use ChatGPT’s Memory Feature

OpenAI recently introduced a Memory feature for ChatGPT, allowing the chatbot to store and recall personal details shared during conversations, enhancing its ability to provide personalized interactions. This experimental feature is currently available to a select group of users, with plans for a broader rollout. The Memory feature automatically activates, but users have the option to opt-out or manage what the chatbot remembers via the Settings and Personalization menu. Users can add or remove details from ChatGPT’s Memory with ease, either through direct conversation prompts or by adjusting settings.

The Memory feature, while not replacing the custom instructions feature, offers a new level of interaction, allowing ChatGPT to remember details such as names, preferences, and other personal information. Users concerned about privacy can adjust their settings to prevent their data from being used to train future AI models, although OpenAI assures that some information might still influence model development. For those sharing accounts, such as ChatGPT Plus users, disabling the Memory feature might be advisable to prevent mixing details from different users.

OpenAI’s introduction of the Memory feature to ChatGPT suggests a future where AI can engage in more lifelike and personalized conversations. However, OpenAI emphasizes that certain sensitive information, like Social Security numbers or passwords, won't be stored, and reminds users that deleting conversations does not automatically erase the associated memories. This development represents a significant step towards creating more engaging and personalized AI interactions, though it also raises important questions about privacy and data usage.

Other stuff

• Figure 01 Completing real-world end-to-end autonomous tasks

• As Nvidia hits $2 trillion, billionaire Marc Rowan’s asset manager Apollo calls AI a ‘bubble’ worse than even the dot-com era 

• A lone developer just open-sourced a tool that could bring an end to Nvidia's AI hegemony

• Defense Department AI chief: Keeping humans in the loop is not always possible anymore

• Wall Street Traders Are Too Scared to Fight the AI Rally

• Google hopeful of a fix for Gemini’s historical-image diversity issue within weeks

• Treating a chatbot nicely might boost its performance — here’s why

Hope you enjoyed today's newsletter

⚡️ Join over 200,000 people using the Superpower ChatGPT extension on Chrome and Firefox.