A.I News
Seedance Competition
Developers generally use an API key—similar to a password—to access and interact with AI models.
Alibaba’s latest move comes after the launch of Seedance 2.0, a video-generation model released earlier this year by TikTok parent ByteDance, as Chinese tech companies race to capitalize on growing demand in the AI video market.
The model, named HappyHorse 1.0, has held the top position on third-party benchmarking site Artificial Analysis’ text-to-video leaderboard since its debut earlier this month.
However, the identity of the model’s creator remained unclear until Friday, when Alibaba confirmed it was behind the project.
“HappyHorse is currently undergoing internal beta testing, and we expect to roll out API access in the near future,” a spokesperson for Token Hub, Alibaba’s newly established AI division, said.
Earlier this year, ByteDance introduced Seedance 2.0, intensifying competition among Chinese internet firms seeking an early lead in AI-generated video. Alibaba soon followed with its own offering.
The Hangzhou-based company has also invested in AI video startups, including ShengShu Technology, the developer of the popular video-generation model Vidu.
New AI model generates 45-minute lip-synced video from one photo
ARRI has introduced LPM 1.0, an AI model that generates real-time video of a speaking, listening, or singing figure from a single image.
The model processes text, audio, and reference images simultaneously. It produces lip-synchronized speech, facial expressions, and emotional transitions. It can connect directly into voice-audio AI models from ChatGPT to create a visual conversation partner in real time.
LPM 1.0 works across different image styles, photorealistic faces, anime, and 3D game characters, without any additional training. The entire video generation runs as a streaming process in real time rather than rendering a finished video all at once. Videos up to 45 minutes long should remain stable.
LPM 1.0 uses "multi-granularity identity conditioning. Alongside a main image, the model also receives reference images from different angles and with varying facial expressions. This means it doesn't have to invent details like teeth, wrinkles tied to specific emotions, or profile views on its own — it can pull them directly from the reference material.
The future of AI
At TechCrunch, several influential leaders took the stage to discuss the future of AI and computing infrastructure. Among them were Christophe Fouquet, head of the Dutch semiconductor equipment giant responsible for the extreme ultraviolet lithography machines essential to modern chip production; Francis deSouza, who is leading one of the largest infrastructure investments in corporate history; Qasar Younis, whose $15 billion physical AI company expanded from simulation technology into defense applications; Dmitry Shevelenko, representing the AI-driven search and agent platform; and Eve Bodnia, a former quantum physicist challenging traditional AI architecture through her startup, Logical Intelligence. Earlier this year, former Meta chief AI scientist Yann LeCun also joined the company as the founding chair of its technical research board.
One major theme from the discussion was the growing strain on the AI industry’s infrastructure. The rapid expansion of AI is colliding with significant physical and manufacturing limitations, particularly in semiconductor production. Fouquet emphasized that chip manufacturing is accelerating at an unprecedented pace, but warned that supply will still fall short of demand for the next several years. As a result, major tech companies such as Google, Microsoft, Amazon, and Meta are unlikely to secure all the chips they need, despite their massive investments.
Elon Musk and OpenAI
Elon Musk’s lawsuit against OpenAI could ultimately come down to one central question: Has the company’s push toward profit weakened its original mission of developing AI that benefits humanity?
During a federal court hearing Thursday in Oakland, former OpenAI employee and board member Rosie Campbell testified that the organization gradually shifted away from its research and safety-first culture toward a more product-driven approach.
Campbell, who joined OpenAI’s AGI readiness team in 2021, said conversations about AGI safety were once a major focus inside the company. But by the time she left in 2024—after her team was dissolved—OpenAI had become increasingly centered on launching products. Around the same period, the company also shut down its Super Alignment team.
While Campbell acknowledged that large-scale funding is likely essential to achieving AGI, she argued that building highly advanced AI systems without strong safeguards would conflict with OpenAI’s founding mission.
She also referenced a situation in which Microsoft released a version of GPT-4 in India through Bing before OpenAI’s Deployment Safety Board had completed its review. Although she said the model itself posed limited risk, Campbell stressed the importance of establishing reliable safety procedures as AI systems become more powerful.
Under questioning from OpenAI’s legal team, Campbell also conceded that, in her opinion, OpenAI’s safety practices remain stronger than those at xAI, Elon Musk’s competing AI company, which was acquired by SpaceX earlier this year.
Kodiak AI
Kodiak AI shares plunged 37% in after-hours trading Thursday after the autonomous trucking company revealed it had raised $100 million through a discounted stock sale — signaling that investors remain interested in the company, but at a lower valuation.
According to a filing with the SEC, Kodiak sold shares at $6.50 each, significantly below its previous closing price of $9.10. The financing package also included warrants, giving investors the option to purchase additional shares later at prices as low as $6.
The funding round included participation from existing investor Ares Management along with several unnamed institutional firms.
The new capital arrives as Kodiak continues the costly process of expanding its autonomous trucking operations across industrial sites and public highways, while working toward long-term profitability.
Financial results highlight the challenge. Kodiak generated $1.8 million in first-quarter revenue, up from $1.4 million during the same period last year. However, the company posted an operating loss of $37.8 million — double the loss reported a year earlier.
Those figures help explain why investors reacted negatively to the discounted raise. While the financing provides additional runway, it does little to ease concerns about the company’s rapid cash burn in the near term.
Kodiak has recently announced several business developments, including a commercial partnership with Roehl Transport, a pilot program with West Fraser Timber Co. in Alberta to test autonomous log-hauling trucks, and a collaboration with General Dynamics Land Systems to develop autonomous military ground vehicles.
As part of the Roehl agreement, Kodiak-equipped trucks will autonomously transport freight between Dallas and Houston on four weekly round trips. Although the vehicles operate autonomously throughout the route, a human safety driver remains behind the wheel as a precaution.
