PLUS: NVIDIA’s new server play, cracking the CUDA moat, and ChatGPT's group project
Good morning
A new startup has launched an app to create interactive digital avatars of the deceased, igniting a widespread and contentious debate about AI's role in processing grief.
The technology is being pitched as a way to create a "living archive," but the intense public reaction raises a significant question: are we prepared for AI to mediate our most personal experiences with loss, and where should the ethical lines be drawn?
In today’s Next in AI recap:
The AI app creating avatars of the deceased
NVIDIA’s plan to sell complete AI servers
Stanford’s new challenge to NVIDIA’s CUDA
ChatGPT's group project
An AI 'Afterlife'
Next in AI: A new startup called 2Wai has launched a controversial app that creates interactive digital avatars of deceased loved ones, prompting a viral debate over the ethics of AI and grief.
Decoded:
The company is pitching its tech as a way to build a "living archive of humanity," but the concept immediately drew comparisons to the unsettling Black Mirror episode, Be Right Back.
Users create what 2Wai calls "HoloAvatars," which are designed to look, talk, and share memories of the person they represent, and the app is now live on the Apple App Store.
A promotional video for the app ignited the intense online discussion, quickly crossing 4.1 million views on X and attracting labels from "demonic" to "nightmare fuel."
Why It Matters: This technology moves AI beyond simple productivity tools and into the deeply personal territory of memory and loss. The widespread and polarized reaction signals a growing public awareness of the complex ethical questions we face as AI integrates into core human experiences.
NVIDIA's Server Play
Next in AI: NVIDIA is reportedly planning a major strategic shift from selling individual GPUs to offering complete, fully-integrated AI servers. A new report from JP Morgan outlines this move toward vertical integration, which could reshape the AI hardware market.
Decoded:
The move is part of a broader vertical integration plan that gives NVIDIA control over the entire hardware stack, aiming to boost quality and profitability.
By providing a complete solution, NVIDIA simplifies the complex purchasing and setup process for enterprise clients and ensures seamless compatibility among all components.
This new strategy will reportedly launch with the Vera Rubin AI server, designed to meet the growing demands of intensive AI workloads.
Why It Matters: This shift allows NVIDIA to capture more value from the AI boom and strengthen its market leadership. It also signals a future where businesses acquire AI capabilities as turnkey solutions, not as a collection of parts.
Cracking the CUDA Moat
Next in AI: Stanford's Hazy Research lab just dropped the open-source library HipKittens, a new toolkit designed to unlock the full potential of AMD GPUs for AI. It simplifies writing high-performance code, directly challenging NVIDIA's long-standing software dominance.
Decoded:
For years, NVIDIA's CUDA software platform has created a “CUDA moat,” making it the default choice for AI development, even as AMD’s hardware offered competitive performance and memory.
The current AMD software ecosystem remains difficult to work with, failing to deliver peak performance despite backing from large-scale AMD deployments by major AI labs.
HipKittens demonstrates that its underlying programming concepts can work across different hardware, paving the way for a potential unified programming model that simplifies development for any AI accelerator, not just NVIDIA's.
Why It Matters: This project makes AMD's powerful hardware more accessible to a broader range of developers, not just a handful of experts. Increased competition in the GPU space could ultimately drive down costs and accelerate AI innovation for everyone.
ChatGPT's Collaborative Turn
Next in AI: OpenAI is piloting a new feature that transforms ChatGPT into a collaborative space for up to 20 users. This move allows teams, families, and friends to work with the AI in a single, shared conversation.
Decoded:
You can start a group chat from any new or existing conversation, and if you add people to an ongoing chat, a private copy is created for the group, keeping your original chat separate.
The feature is currently in a pilot phase, rolling out to both free and paid users in Japan, New Zealand, South Korea, and Taiwan.
The AI has been trained with new social behaviors to decide when to interject in the conversation, or you can directly mention "ChatGPT" to prompt a response.
Why It Matters: This pilot shifts ChatGPT from a solitary tool into a multi-user, collaborative platform. The move could reshape how teams brainstorm, plan projects, and make decisions with AI as an active participant.
AI Pulse
A ruling in the Authors Guild v. OpenAI lawsuit suggests detailed AI-generated book summaries may constitute copyright infringement, posing significant fair use questions for services that summarize content.
Microsoft partnered with Westinghouse Nuclear to use AI to accelerate the construction of new nuclear power plants, though a report from the AI Now Institute warns the push could compromise safety.
Activision confirmed it uses a variety of tools, including AI, to create gaming experiences in response to claims that generative AI was used for calling cards in Call of Duty: Black Ops 7.