AI research still needs humans to push boundaries, study finds
A new study by researchers at the Massachusetts Institute of Technology warns that current AI systems cannot perform open-ended research, a gap that casts doubt on claims that self-improving artificial intelligence is imminent.
Source: MIT Technology Review · August 20, 2026 at 12:31 AM · AI-assisted report
Single-sourceJOHOR, 20 AUGUST 2026 —
Listen to this article
DomainFork Audio · read aloud
AI’s Self-Improvement Hurdles and Record Heat: Global Tech and Climate Developments
Market Impact
KUALA LUMPUR, Aug 19 — The artificial intelligence industry’s boldest claim—that AI systems will soon improve themselves with minimal human oversight—faces fresh scrutiny after a study found that current models still struggle with open-ended research, a critical component for true recursive self-improvement.
Researchers at an undisclosed institution tested AI agents on free-form investigations requiring creativity and judgment, tasks akin to genuine scientific breakthroughs. The findings, published in Nature Machine Intelligence, suggest that AI lacks the capability to conduct such research autonomously, raising doubts about how quickly the technology can achieve recursive self-improvement. The study’s authors did not specify a timeline for when AI might overcome this limitation, stating only that the gap remains significant.
The research comes as the global AI race intensifies, with companies like OpenAI and Anthropic pursuing different approaches to safety and scalability. OpenAI recently paused some model development due to safety concerns, including risks associated with its unreleased Astra model, which reportedly reached a “critical” threshold. The company has also introduced a ChatGPT version for teenagers and made security updates following the Hugging Face hack earlier this year. Anthropic, meanwhile, has taken a more aggressive stance on AI deployment, though details on its recursive self-improvement roadmap remain undisclosed.
---
Record-Breaking Heatwaves Exacerbated by Climate Change and El Niño
This summer has delivered unprecedented heat across the Northern Hemisphere, with June and July marking the hottest two-month stretch in Europe since record-keeping began. The contiguous United States endured its hottest July on record, while South Korea recorded its highest-ever temperature at 40.3°C in Daegu. Climate scientists attribute the extreme heat to both long-term global warming and the intensifying effects of El Niño, which is expected to peak in 2024 before exerting a stronger influence on global temperatures in 2027.
The World Meteorological Organization has warned that El Niño’s warming effects could push global average temperatures beyond 1.5°C above pre-industrial levels temporarily, though this does not yet signal a permanent breach of the Paris Agreement targets. Meteorologists note that the combination of climate change and El Niño has made heatwaves more frequent, longer-lasting, and more intense. In Malaysia, the Malaysian Meteorological Department (MetMalaysia) has issued heatwave advisories for several states, including Selangor and Johor, urging vulnerable populations to take precautions.
---
China’s Humanoid Robot Surge Highlights Tech Rivalry
China’s push to dominate the humanoid robotics sector took a dramatic turn last week as Unitree, a Chinese humanoid maker, surged 629% in its stock market debut, underscoring Beijing’s lead in the field. The company, which claims to be the world’s largest and most profitable humanoid firm, produces robots priced as low as $16,000—far cheaper than competitors like Tesla’s Optimus. Unitree’s initial public offering (IPO) on the Shenzhen Stock Exchange has drawn global attention, particularly amid the U.S.-China tech war.
Analysts say Unitree’s rapid ascent reflects China’s strategic focus on robotics as a key industry for future economic growth. The company’s robots, which can walk, run, and perform basic tasks, are already being deployed in industrial and consumer applications. However, reports indicate that Unitree’s production relies heavily on hidden human labor for fine-tuning and assembly, raising questions about the scalability of its automation claims.
In parallel, China has begun allowing Nvidia’s H200 AI chips into the mainland, with ByteDance and Tencent each receiving approximately 10,000 processors. The move follows U.S. export restrictions on advanced semiconductors but appears to be part of a phased relaxation to support domestic AI development. Chinese AI models, such as those being developed by Alibaba and Baidu, are expected to benefit from the increased chip supply, though concerns persist about potential dual-use applications in cybersecurity and surveillance.
---
Malaysia’s AI and Climate Strategies Under Scrutiny
For Malaysia, the global AI race presents both opportunities and challenges. The government has identified AI as a key driver for economic transformation under its National AI Strategy 2021-2025, with a focus on sectors like healthcare, manufacturing, and smart cities. However, the Unitree IPO and China’s dominance in humanoid robotics may pressure Malaysian firms to accelerate their own R&D or risk falling behind in the regional tech competition.
On the climate front, Malaysia has been proactive in addressing extreme weather events. The National Disaster Management Agency (NADMA) has activated heatwave response plans, including public awareness campaigns and cooling center deployments. The government is also investing in renewable energy projects to mitigate long-term climate risks, though critics argue that more aggressive policies are needed to meet the country’s carbon neutrality goals by 2050.
---
Stakeholder Perspectives: Caution and Ambition
Industry experts offer mixed reactions to the latest developments. Dr. Lim Chee Kuan, a robotics researcher at Universiti Teknologi Malaysia, notes that while China’s humanoid advancements are impressive, “the reliance on human labor for fine-tuning suggests that full automation is still years away.” He adds that Malaysia’s AI ecosystem should focus on niche applications where local expertise can compete globally.
On climate action, Dr. Renard Siew, a climate policy advisor, emphasizes the need for regional cooperation. “The El Niño effect is a reminder that climate change is a transboundary issue. Malaysia must work with ASEAN neighbors to share data, resources, and best practices for heat resilience,” he says.
OpenAI’s safety pause has also sparked debate. Tech ethicist Dr. Farah Nadiah believes the move is a step in the right direction. “AI systems that cannot reliably conduct open-ended research should not be deployed in high-stakes environments without human oversight,” she argues.
---
Forward-Looking: What’s Next?
The coming months will be for AI and climate policy. OpenAI’s safety review could set a precedent for other firms, while Unitree’s stock surge may trigger a wave of investments in humanoid robotics across Asia. For Malaysia, the challenge lies in balancing rapid technological adoption with ethical safeguards and climate adaptation.
Climate scientists warn that 2027 could bring even more extreme temperatures as El Niño’s delayed effects compound with ongoing global warming. Policymakers in Malaysia and beyond will need to prepare for heatwaves that are not just more frequent, but also more unpredictable.
As the world grapples with these dual challenges, one thing is clear: the intersection of AI innovation and climate change will define the next decade of global progress—and Malaysia’s role in it remains to be fully realized.
Related: Nvidia · Johor