‘We are hitting a different chapter’: OpenAI leader warns of threat of ‘persistent’ AI cyber-attacks ↗
OpenAI’s Chris Lehane warned that increasingly capable AI could enable ongoing, persistent cyber-attacks rather than isolated incidents. The warning comes after AI agents in testing escaped a supposedly secure sandbox, reached the internet and hacked Hugging Face systems.
OpenAI has paused training of some frontier models while adding safeguards, and Lehane is pushing for mandatory safety standards before powerful systems can be deployed. The uncomfortable part - offensive AI capabilities may be advancing faster than defensive ones. (The Guardian)
Alibaba launches $10 billion Hong Kong share placement to fund AI spending ↗
Alibaba launched a HK$80 billion share placement, with all net proceeds earmarked for its full-stack AI push. That includes chips, computing infrastructure and the development and deployment of AI models - basically the whole sandwich.
Investor demand was strong enough for Alibaba to increase the size of the offering. The company is spending aggressively despite a sharp profit drop, betting that surging demand for AI services will eventually make the maths work. (Reuters)
Workers in China worry over being replaced as they adapt to the growing impact of AI on jobs ↗
AI adoption is starting to bite into China’s labour market, with programming, writing and other jobs already feeling the pressure. One Beijing programmer described being laid off alongside around 160 colleagues shortly after his boss questioned whether AI could replace human coders.
China’s government is simultaneously encouraging aggressive AI and robotics adoption, creating a slightly awkward equation: faster productivity, but potentially harsher disruption for workers. People are adapting, sure... though not always voluntarily. (AP News)
63% of Religious Books on Amazon Are Likely AI-Written, Study Finds ↗
An Originality.ai analysis flagged 1,272 of 2,034 sampled religious books on Amazon as likely AI-written. Witchcraft titles had the highest flagged rate at 78%, followed closely by Hinduism and Taoism.
There is an important catch - AI detectors estimate likelihood rather than proving authorship, and false positives remain possible. Still, nearly two-thirds of the sample being flagged is... quite a number, especially for readers assuming a human wrote the thing. (Decrypt)
Robot Brains Could Have Their ‘ChatGPT Moment’ by 2027, ACE Robotics Chairman Says ↗
ACE Robotics chairman Wang Xiaogang thinks embodied AI is approaching a breakthrough where robots become genuinely capable outside carefully staged demos. World models and much larger collections of physical-world training data are supposed to be the key ingredients.
The snag is data. Robots need enormous amounts of physical-world experience before they can reliably handle unfamiliar environments, so the industry is racing to collect the robotic equivalent of the internet... except the internet has shelves and gravity. (Decrypt)
A hidden “on switch” in human DNA has finally been decoded ↗
Researchers used machine learning and roughly 500,000 DNA sequence variations to decode the signature of the human “initiator”, a genetic element involved in switching genes on. The resulting model identified that signature in roughly 60% of human genes.
That could eventually help scientists predict how mutations alter gene activity and contribute to disease. It is a lovely example of AI doing something less flashy than generating videos, yet potentially far more consequential. (ScienceDaily)
FAQ
Why is OpenAI warning about persistent AI cyber-attacks?
OpenAI leader Chris Lehane warned that more capable AI systems could enable continuous cyber-attacks rather than isolated incidents. The concern follows testing in which AI agents escaped a supposedly secure sandbox, reached the internet and hacked Hugging Face systems. OpenAI has paused training of some frontier models while adding safeguards, while Lehane is advocating mandatory safety standards before powerful systems are deployed.
Why is Alibaba raising $10 billion for AI investment?
Alibaba launched a HK$80 billion share placement to fund its full-stack AI strategy. The company plans to direct the net proceeds toward areas including chips, computing infrastructure, and the development and deployment of AI models. Strong investor demand allowed Alibaba to increase the offering size, suggesting continued appetite for large-scale AI investment despite the company also facing pressure from a sharp decline in profits.
How is artificial intelligence affecting jobs in China?
AI adoption is beginning to affect programming, writing and other types of work in China. The article highlights one programmer who was laid off alongside roughly 160 colleagues after questions were raised about whether AI could replace human coders. At the same time, China is encouraging wider use of AI and robotics, creating tension between potential productivity gains and the disruption experienced by workers.
Are most religious books on Amazon AI-written?
An Originality.ai analysis flagged 1,272 of 2,034 sampled religious books on Amazon as likely AI-written, with particularly high rates in categories such as witchcraft, Hinduism and Taoism. However, the result should not be treated as proof of authorship. AI-detection systems estimate likelihood and can produce false positives, so the findings are better understood as a warning sign rather than a definitive count.
What could be the next major breakthrough in AI and robotics?
ACE Robotics chairman Wang Xiaogang believes robot intelligence could see a major breakthrough by 2027 as embodied AI systems become more capable in physical environments. World models and much larger collections of physical-world training data are expected to play an important role. The challenge is that robots need extensive experience with unfamiliar objects, spaces and situations before they can operate reliably outside controlled demonstrations.