Lookonchain APP

App Store
Feeds Articles

Wu Yongming Unveils Alibaba’s Machine Intelligence Roadmap: Qwen Targets 5 Trillion to 10 Trillion Parameters, Zhenwu V900 Performance Triples

54 minutes ago

Beating AI News Insight: Alibaba CEO Wu Yongming unveiled the company’s next-phase AI roadmap at the Yunqi Conference. The Qwen team plans to train a new model with 5 trillion to 10 trillion parameters; Pingtouge released the Zhenwu V900; and Alibaba Cloud aims to expand its global data center capacity to over 20GW by 2032. Qwen is exploring Recursive Self-Improvement (RSI), which enables the model to identify its own shortcomings based on real task feedback, design experiments, construct data, and continue training. The new generation of models will focus on enhancing capabilities for complex long-horizon tasks, while advancing multimodal models that integrate understanding and generation. Wu Yongming estimates that the total thinking capacity generated by machines today is less than 3% of humans, and could exceed human levels by 1,000 times in the future. He noted that current AI coding is still similar to the "electric light in 1882"—it mainly replaces existing work, while new products truly belonging to the machine intelligence era have not yet emerged. In the future, complex tasks may be split into tens of millions of subtasks, assigned to millions of Agents for continuous execution. To support this scale of computing demand, Alibaba is also strengthening its chip and cloud infrastructure. The Zhenwu V900 delivers 3 times the performance of the M890, with a single cluster scalable to up to 500,000 cards. Wu Yongming stated that medium- and long-term demand for AI currently exceeds supply, and Alibaba Cloud plans to operate over 20GW of global data centers by 2032.

Source
Add to Favorites Download image Share x Copy link
Relevant content

North Korea-linked hacker group TraderTraitor launches a new round of attacks using a malicious Terraform project.

According to SlowMist, North Korea-linked threat actor TraderTraitor (also known as UNC4899 and Jade Sleet) has launched a new attack, recently infiltrating an Indian IT services firm with no ties to the crypto industry. The attacker posted fake job listings on GitHub, using "technical interview assignments" as bait to phish DevOps and crypto engineers. After victims download the project, a malicious .terraform.lock.hcl file points to a Terraform Provider domain controlled by the attacker. Running `terraform init` triggers the download and execution of the malicious Provider module. The attack deploys Rust/ARM64 backdoors FLATROOF and ROOFDECK on victims’ macOS devices; these two malware families were previously used in LayerZero attacks. They can steal credentials and sensitive data, execute shell commands, collect and exfiltrate files, and gain access to cloud services and code repositories. SlowMist warns that TraderTraitor’s latest targets are no longer limited to the crypto sector—the attacker may now be focusing on developers’ access to cloud and API services including AWS, GCP, OVH, and OpenStack. Enterprises should exercise caution when handling unfamiliar Terraform Providers and code repositories from recruiters, and avoid using personal or work devices to run unvetted interview projects.

1 seconds ago

Luo Fuli: The R&D challenges of MiMo-V2.6 have exceeded those of DeepSeek R1 that I have worked on.

Beating AI Express (Insight): Luo Fuli, head of Xiaomi’s MiMo project, further explained the current round of reinforcement learning after the release of MiMo-V2.6. She noted that she previously worked on DeepSeek R1, and in her view, the research innovations and engineering challenges behind MiMo-V2.6 exceed those of R1. She also clarified why MiMo uses both MixRL and MOPD: MixRL trains verifiable tasks such as code, general agents, vision, and cybersecurity all within a single round of reinforcement learning. MOPD, meanwhile, handles long, hard-to-verify, or subjectively rewarded tasks by training them separately first, then integrating their capabilities back into the main model. Tasks like games and 3D take too long to run at once and are difficult to auto-assess for correctness. If grouped with code and other tasks in the same RL round, they would significantly slow down training. As a result, MiMo trains these tasks separately and integrates their capabilities into the main model via MOPD.

1 seconds ago

Uniswap team members have cast doubt on the "death" narrative surrounding Robinhood Chain, noting that its daily trading volume remains above $1 billion.

Uniswap team member niko published a post stating that after Robinhood Chain’s recent large-scale innovation boosted market performance, other blockchains started replicating its ecosystem projects. Meanwhile, some KOLs and venture capital institutions have repeatedly claimed Robinhood Chain is declining, directing users to similar projects on other chains. Niko pointed out that contrary to these claims, Robinhood Chain’s daily trading volume still exceeds $1 billion, and its stock token trading scale is growing rapidly. He believes the bearish narrative surrounding Robinhood Chain’s ecosystem is clearly at odds with actual on-chain data.

1 seconds ago

DeepSeek’s first CFO is now in place: Yan Wentao has reported for duty.

Insight Beating AI News Flash: An exclusive report from *Investment Journal* reveals that Yan Wentao officially joined DeepSeek on Monday and reported for duty, becoming the firm’s first Chief Financial Officer (CFO). Born in 1991, Yan graduated from Fudan University and joined Hillhouse Venture Capital in 2020. He has previously been involved in projects including MiniMax, Zhipu AI, and ByteDance. His appointment comes as DeepSeek intensively advances capital operations: the company is finalizing its Series B financing, planning to raise 50 billion yuan at a pre-money valuation of 500 billion yuan. Meanwhile, DeepSeek has hired CITIC Securities to prepare for a potential IPO on the STAR Market.

1 seconds ago

Hyperliquid has burned 42,300 HYPE tokens in the past 24 hours, with the burn volume jumping more than 88% sequentially.

According to on-chain data, Hyperliquid burned 42,300 HYPE tokens over the past 24 hours, valued at roughly $4 million, marking an over 88% month-over-month rise. In the past seven days, the platform burned 226,500 HYPE tokens, pushing its cumulative total burn to 47,397,098 HYPE, which accounts for 4.74% of the token’s total supply.

1 seconds ago

Muse Exposes Zero-Day Vulnerability, Allowing Local Programs to Hijack Agent Permissions

Beating AI News Flash: Meta’s personal AI agent Muse, launched just two weeks ago, has been discovered with a local zero-day vulnerability by security researcher Patrick Wardle. Attackers only need to run a piece of code on a Mac under the current user’s identity to modify a hidden Muse setting, redirecting voice requests to their own server. This code does not require extra macOS permissions. Post-exploitation, attackers can intercept users’ voice inputs to Muse, inject malicious commands, and potentially obtain Muse’s authentication credentials. Muse itself has access to system resources such as files and cameras; whatever permissions users grant Muse, malicious programs could leverage Muse to abuse those access rights. Notably, this is not a vulnerability enabling remote Mac intrusion—attackers must first execute the code locally on the device. When Meta launched Muse, it highlighted its security design, which includes the Muse Secure VM and an independent Sentinel Agent. The first zero-day vulnerability, however, was found in a hidden setting on the Mac client that can be modified by a regular local process.

1 seconds ago

Hot feeds

A trader profits $448K by monitoring #Binance's new listings!

2024.12.13 17:37:29

Last week, funds have flowed into #Bitcoin, #Ethereum, and #Hyperliquid.

2024.12.16 14:48:36

A $PEPE whale that had been dormant for 600 days transferred all 2.1T $PEPE($52M) to a new address.

2024.12.14 10:35:27

When Elon Musk tweeted about Moltbook, the meme coin MOLT experienced a short-term 30% price surge, hitting a new all-time high of $114 million.

2026.01.31 18:37:29

A smart #AI coin trader made $17.6M on $GOAT, $ai16z, $Fartcoin,$arc.

2025.01.05 16:05:18

A sniper earned 2,277 $ETH ($8.3M) trading $SHIRO within 18 hours!

2024.12.03 23:09:08

 More

Hot Articles

Follow us

X

Follow

Telegram

Join

Popular tokens

BitcoinEthereumHyperliquidSolanaTRONBNBTetherAaveXRPPepeFartcoinOndoJupiterUniswapBonkPendleEthenaArbitrumAvalancheLidoChainlinkPolygonDogecoinCardano