ai

Anthropic aims to decode AI ‘black box’ within two years​

Anthropic CEO Dario Amodei has unveiled an ambitious plan to make AI systems more transparent by 2027. In a recent essay titled ‘The Urgency of Interpretability,’ Amodei highlighted the pressing need to understand the inner workings of AI models.

He expressed concern over deploying highly autonomous systems without a clear grasp of their decision-making processes, deeming it ‘basically unacceptable’ for humanity to remain ignorant of how these systems function.

Anthropic is at the forefront of mechanistic interpretability, a field dedicated to deciphering the decision-making pathways of AI models. Despite these advancements, Amodei emphasized that much more research is needed to fully decode these complex systems.​

Looking ahead, Amodei envisions conducting ‘brain scans’ or ‘MRIs’ of advanced AI models to detect potential issues like tendencies to deceive or seek power. He believes that achieving this level of interpretability could take five to ten years but is essential for the safe deployment of future AI systems.

Amodei also called on industry peers, including OpenAI and Google DeepMind, to intensify their research efforts in this area and urged governments to implement ‘light-touch’ regulations to promote transparency and safety in AI development.​

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot! 

YouTube is testing AI-generated video highlights

Google is expanding its AI Overviews feature to YouTube, bringing algorithmically generated video highlights and search suggestions to the platform. Initially rolled out to a limited number of YouTube Premium users in the US, the experimental tool uses AI to identify and surface the most relevant clips.

The AI-generated results are currently focused on shopping and travel content, offering viewers a new way to discover videos and related topics without watching entire clips.

Google says the feature is designed to streamline content discovery, though it arrives with some scepticism following the rocky debut of AI Overviews in Google Search last year. That version, introduced in May 2024, was widely criticised for factual errors and bizarre “hallucinations” in responses.

Despite its troubled track record, Google is pushing ahead with AI integration across its platforms. The company’s blog post emphasised that the YouTube trial remains limited in scope for now, while promising future refinements.

Whether the move improves user experience or adds confusion remains to be seen, as critics question the reliability of AI-generated summaries on such a massive and diverse video platform.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot! 

Alphabet surpasses expectations with strong Q1 performance and major buyback plan

Google’s parent company, Alphabet, announced a $70 billion share buyback after posting first-quarter profits that exceeded Wall Street forecasts. The company’s shares surged 4% after-hours trading, boosting its market value by around $75 billion. Alphabet reported $90.23 billion in revenue, mainly driven by steady growth in its core digital advertising business, which offset a slight slowdown in its cloud computing segment.

Despite concerns over economic uncertainty linked to US trade policies, Alphabet’s ad revenue, making up 75% of its total income, rose 8.5% to $66.89 billion, surpassing analyst expectations. CEO Sundar Pichai highlighted strong engagement in Google Search, particularly with AI-powered features, attracting 1.5 billion monthly users.

Meanwhile, Google Cloud saw a 28% revenue increase, narrowly missing projections but still reflecting solid growth. The tech giant also ramped up capital spending by 43% to $17.2 billion as part of its $75 billion annual investment plan, focusing on expanding data centres and AI infrastructure.

Despite rising costs and global competition in the AI sector, Pichai emphasised the need for heavy investment to enhance services like Search and develop AI tools. Alphabet’s positive results lifted other digital ad players, with Meta, Amazon, and Snap seeing gains in extended trading.

While Big Tech remains committed to AI spending, signs of caution emerge as some companies begin to scale back data centre expansions amid economic pressures.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot! 

AI tool improves accuracy in detecting heart disease

A team of researchers at Mount Sinai Hospital in New York has successfully calibrated an AI tool to more accurately assess the likelihood of hypertrophic cardiomyopathy (HCM) in patients.

By assigning specific probability scores, the AI model now offers clearer guidance to clinicians and patients regarding disease risk.

HCM, a thickening of the heart muscle that affects around one in 200 people globally, can lead to serious complications such as heart failure or sudden cardiac death.

The Viz HCM algorithm, already approved by the US Food and Drug Administration, previously provided vague classifications like ‘suspected HCM.’ Thanks to model calibration, clinicians can now give patients more precise estimates—for instance, a 60% probability of having the condition.

Researchers ran the algorithm on nearly 71,000 patients who had undergone electrocardiograms between March 2023 and January 2024. Out of these, 1,522 were flagged by the AI, with further review of medical records and imaging confirming diagnoses.

The results validated that the newly calibrated probabilities closely reflected real-world outcomes, improving the tool’s accuracy and practical utility.

Experts say this advancement enhances clinical workflows by helping doctors prioritise patients based on their actual risk levels.

Beyond technological innovation, the study marks a step forward in integrating AI responsibly into everyday clinical practice—making healthcare more personalised, interpretable, and effective.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

AI investments lift Alphabet despite cloud slowdown

Alphabet’s shares climbed over 5% in premarket trading after the company reported strong earnings that reassured investors of its AI strategy.

Despite a slight deceleration in advertising and cloud growth, Google’s parent company beat expectations, signalling that its major bets on artificial intelligence are starting to pay off.

Advertising revenue, which forms the backbone of Alphabet’s business, rose 8.5% in the first quarter to $66.89 billion—outperforming analyst projections.

Although this marks a slowdown from the previous quarter’s growth, it reinforces investor confidence in Alphabet’s ability to monetise AI across its services. Meanwhile, Google Cloud revenue grew by 28%, falling just short of forecasts and indicating some cooling in the segment.

The company is pressing ahead with its ambitious infrastructure plans, reaffirming a $75 billion investment in expanding data centre capacity.

Alongside Microsoft’s even larger plans, these efforts contribute to Big Tech’s anticipated $320 billion AI investment in 2025. However, growing trade tensions and fears of an economic downturn have led to questions about the sustainability of such capital spending.

While Alphabet remains a key player in the AI race, legal challenges loom large. Ongoing antitrust actions in the United States could compel the company to divest core assets like Chrome, as regulators seek to limit Google’s market dominance.

Nevertheless, many analysts remain optimistic, with several brokerages raising their price targets, pointing to Alphabet’s ability to deliver GenAI-powered products at scale despite headwinds.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

Fibocom launches AI-powered 5G hotspot

Fibocom has introduced a next-generation 5G AI Mobile Hotspot, combining ultra-fast 5G connectivity with intelligent voice interaction features.

Designed to serve both individual users and business environments, the device marks a shift from conventional mobile hotspots to smart, AI-enabled communication hubs.

The hotspot is powered by a custom-built AI app that offers voice-activated multilingual control, allowing users to manage services hands-free.

With real-time content access and an AI translation engine supporting over 20 languages at up to 95% accuracy, the device is particularly suited to international settings, such as global meetings and travel.

Built on the Qualcomm QCM4490 platform, the device uses an 8-core 4nm CPU running up to 2.4 GHz. It supports behavioural analytics, dual SIM standby, USB 3.1 and Ethernet, and offers 15-W fast and reverse charging.

Compatible with both Android and Linux systems, it also accommodates external cameras and 1080P displays.

Incorporating Qualcomm’s advanced Wi-Fi 7 chipset, the hotspot supports concurrent dual-band modes, enabling download speeds up to BE5800.

With full 3GPP Release 16 compliance and support for Sub-6-GHz frequencies, it promises reliable, high-performance wireless access for long-term and future-ready use.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

Auto Shanghai 2025 showcases cutting-edge AI robots

At Auto Shanghai 2025, running from April 23 to May 2, nearly 1,000 companies from 26 countries showcase their innovations.

A major highlight of the event has been the introduction of AI humanoid robots.

Among the most talked-about innovations is Mornine Gen-1, an AI humanoid robot developed by Chinese automaker Chery.

Designed to resemble a young woman, Mornine is set for various roles, from auto sales consultation to retail guidance and entertainment performances.

Also drawing attention is AgiBot’s A2 interactive service robot. Serving as a ‘sales consultant,’ the A2’s smart, interactive features have made it a standout at the event.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

Politeness to AI is about us, not them

In his thought-provoking blog post ‘Politeness in 2025: Why are we so kind to AI?’, Dr Jovan Kurbalija explores why nearly 80% of users in the UK and the USA instinctively say ‘please’ and ‘thank you’ to AI platforms like ChatGPT.

While machines lack feelings, our politeness reveals more about human psychology and cultural habits than the technology itself. For many, courtesy is a deeply ingrained reflex shaped by personality traits such as agreeableness and lifelong social conditioning, extending kindness even to non-sentient entities.

However, not everyone shares this approach. Some users are driven by subtle fears of future AI dominance, using politeness as a safeguard, while others prioritise efficiency, viewing AI purely as a tool undeserving of social niceties.

A rational minority dismisses politeness altogether, recognising AI as nothing more than code. Dr Kurbalija highlights that these varied responses reflect how we perceive and interact with technology, influenced by both evolutionary instincts and modern cognitive biases.

Beyond individual behaviour, Kurbalija points to a deeper issue: our tendency to humanise AI and expect it to behave like us, unlike traditional machines. This blurring of lines between tool and teammate raises important questions about how our perceptions shape AI’s role in society.

Ultimately, he suggests that politeness toward AI isn’t about the machine—it reflects the kind of humans we aspire to be, preserving empathy and grace in an increasingly digital world.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

Google boosts AI in coding and cloud growth

More than 30% of all code at Google is now written with the help of AI, according to CEO Sundar Pichai during Alphabet’s Q1 2025 earnings call.

Developers are embracing AI-generated suggestions in nearly one out of every three code changes, thanks to improved models and the rollout of agentic workflows—AI systems designed to manage complex, multi-step tasks.

Pichai noted that AI-assisted coding is expanding across the company, with customer service teams leading the way in adoption.

Alphabet reported strong financial results, with quarterly revenue climbing 12% year-over-year to $90.2 billion. Net income rose sharply by 46% to $34.5 billion, and earnings per share jumped 49%.

While there was a slight quarterly dip in revenue from the previous quarter, Google’s core advertising business, YouTube, and Google Cloud all contributed to year-on-year growth. Cloud revenue, in particular, increased 28% due to surging demand for AI and infrastructure products.

The recently released Gemini 2.5 Pro model was described by Pichai as ‘state-of-the-art’, outperforming rivals across benchmarks and landing the top spot on Chatbot Arena. Gemini models are now integrated across 15 Google products used by more than half a billion people.

Features like Gemini Live and AI-powered camera tools are being rolled out on Android and Pixel devices, while Google Assistant will also receive a Gemini upgrade later this year.

Instead of slowing down, Google is accelerating its AI development with initiatives like Gemini Robotics Models, the AI Co-Scientist for research, and the continued success of AlphaFold, used by over 2.5 million scientists.

With over 200% growth in users of AI Studio and Gemini API, and more than 140 million downloads of Gemma models, Google is clearly positioning AI at the centre of its future across products, platforms, and research.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

ChatGPT expands Deep Research to more users

A new feature introduced by ChatGPT in February, called Deep Research, is gradually becoming available across its user base. This includes subscribers on the Plus, Team, and Pro plans, while even those using the free ChatGPT app on iOS and Android can now access a simplified version.

Designed to carry out in-depth reports and analyses within minutes, Deep Research uses OpenAI’s o3 model to perform tasks that would otherwise take people hours to complete.

Instead of limiting access to paid users alone, OpenAI has rolled out a lightweight version powered by its o4-mini AI model for free users. Although responses are shorter, the company insists the quality and depth remain comparable.

The more efficient model also helps reduce costs, while delivering what OpenAI calls ‘nearly as intelligent’ results as the full version.

The feature’s capabilities stretch from suggesting personalised product purchases like cars or TVs, to helping with complex decisions such as choosing a university or analysing market trends.

Free-tier users are currently allowed up to five Deep Research tasks each month, whereas Plus and Team plans get ten full and fifteen lightweight tasks. Pro users enjoy a generous 125 tasks of each version per month, and EDU and Enterprise plans will begin access next week.

Once users hit their full version limit, they’ll be automatically shifted to the lightweight tool instead of losing access altogether. Meanwhile, Google’s GeminiAI offers a similar function for its paying customers, also aiming to deliver quick, human-level research and analysis.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!