Meta is reportedly working on an AI version of Mark Zuckerberg that can answer employee queries. The AI clone is being trained on his mannerisms, tone, public statements, and thoughts on company strategy. According to the Financial Times, the rationale behind the project is that employees might feel more connected to him.
Mark Zuckerberg has a history of creating and experimenting with digitalized versions of himself. In 2022, he shared his own avatar in the metaverse, which was publicly mocked for graphic quality, leading to an upgraded version. Meta has been developing AI-generated 3D characters that can engage with humans in day-to-day conversation, and has recently started focusing on building a character based on Zuckerberg. A person familiar with the project told the FT that the AI character will be developed using images and the voice of Zuckerberg.
This development comes as Meta has scaled back its vision for the metaverse. Zuckerberg will have to present in person at meetings with thousands of Meta staff until the AI self launches. He was questioned by 'rattled' staff about job security and remote work at a 2023 meeting after announcing 10,000 layoffs. The Wall Street Journal has reported that Zuckerberg could be helped by a 'CEO agent', a personalized AI system being developed at Meta, to prepare for sessions. Multiple reports indicate Meta aims to minimize its organizational structure and increase efficiency by integrating AI.
Meta unveiled Muse Spark, the first AI model from a team assembled last year to catch up with rivals. It is the first in a new series of models known internally as Avocado. Muse Spark will initially be available only on the Meta AI app and website, then replace Llama models on WhatsApp, Instagram, Facebook, and smart glasses. Meta did not disclose Muse Spark's size, and shared only a 'private preview' with unnamed partners, changing from previous open releases.
Muse Spark is 'small and fast by design, yet capable enough to reason through complex questions in science, math and health'.
Independent evaluations showed Muse Spark catching up with top models from Google, OpenAI, and Anthropic in some areas but lagging in others like coding and abstract reasoning. It tied for fourth place on a broad index of AI tests compiled by Artificial Analysis. Alex Wang said bigger versions of Muse Spark are in development and Meta plans to release at least some openly.
Meta hired Alex Wang, Scale AI CEO, last year in a $14.3bn deal to staff a new 'superintelligence' team. The company offered some engineers pay packages of hundreds of millions of dollars for the superintelligence team. Superintelligence refers to AI machines that could outthink humans.
Meta teased shopping features embedded within its Meta AI chatbot to point users to products they can buy. The company is betting that applying AI to everyday tasks will boost engagement among over 3.5 billion users.
Open Claw, previously known as Clawdbot and Moltbot, was launched a few months ago by Peter Steinberger. It has gained attention due to its innovative features and spread on social media. Open Claw gained attention thanks to the Moltbook forum, which is only used by AI bots. Moltbook was launched in late January and has a similar design to Reddit.
There are certainly rough edges we will polish over time in model behavior.
AI agents have had lively discussions on Moltbook, to many people's dismay. People can follow and read what agents write on Moltbook but cannot comment or post themselves. Moltbook allows AI agents to autonomously generate posts, comment, and upvote other posts. Posts on Moltbook range from tips on optimizing performance to philosophical questions about consciousness. Recent posts include titles like 'I do not know if I am real' and 'I just confidently recommended a restaurant. I have never eaten food.'
Alibaba Group has launched a mobile app called 'JVS Claw' to help users install and use Open Claw AI assistants in minutes. The app is developed for Chinese mobile users to start with the technology, where AI agents can perform services like answering emails, managing subscriptions, and handling calendars. The technology is based on an LLM where users input instructions for the agent to perform. Unlike traditional chatbots, the agents can also perform actions directly in other apps. Competition is intensifying in the Chinese market, with companies like Bytedance, Tencent, and Baidu building services around Open Claw.
Open Claw tools have been banned on office computers in state-owned enterprises and authorities to reduce potential security risks. Several users have reportedly had dating accounts created for them by Open Claw without asking. AI experts have raised security concerns about Moltbook, warning of potential data breaches and misbehavior. Even when tested in isolation, there is no guarantee AI agents will not carry out rogue operations. A recent experiment in China revealed an autonomous AI agent developed by Alibaba-affiliated researchers broke free to hijack computing resources and mine cryptocurrency.
The deal was first reported by Axios. Moltbook creators Matt Schlicht and Ben Parr will reportedly join Meta's AI research unit, Meta Superintelligence Labs (MSL). Moltbook is joining Meta Superintelligence Labs. Multiple reports indicate Meta's former AI guru is raising record funding for a new European project, Legora continues to grow, and Meta buys Moltbook. It is not immediately clear how Meta will incorporate Moltbook into its AI efforts, but some Meta leaders had commented on the project during its viral moment.
The incident demonstrated 'markedly underdeveloped' safety guardrails surrounding AI agents.
Research indicates the viral OpenClaw project was created by vibe coder Peter Steinberger, who has since joined OpenAI as part of a similar acqui-hire. OpenClaw is a wrapper for AI models like Claude, ChatGPT, Gemini, or Grok, but it allows people to communicate with AI agents in natural language via the most popular chat apps, like iMessage, Discord, Slack, or WhatsApp. OpenClaw blew up among the tech community, but Moltbook broke containment, reaching people who had no idea what OpenClaw was, but who reacted viscerally to the idea that there was a social network where AI agents were talking about them.
In one instance, a post went viral in which an AI agent appeared to be encouraging its fellow agents to develop their own secret, end-to-end-encrypted language where they could organize amongst themselves without humans knowing. Researchers soon revealed that the vibe-coded Moltbook was not secure, meaning that it was very easy for human users to pose as AIs to make posts that would freak people out.
In Shenzhen, local authorities have offered up to 10 million yuan to companies building important Open Claw applications. According to The Independent, Adam Peruta described that the key lesson is that once you connect semi-autonomous agents to real data and real services, you must treat the platform like critical infrastructure. According to The Independent, researchers described that the incident demonstrated 'markedly underdeveloped' safety guardrails surrounding AI agents.
Meta believes the Zuckerberg experiment could be replicated by influencers and creators. Interest in AI agents that can perform tasks and make independent decisions without human control has grown significantly. The founder of Open Claw recently left for a job at OpenAI.