• Analyst report claims primary smartphone market is expected to decline 14.8% in 2026
• Entry level smartphone prices have already risen more than 50% this year
• Refurbished smartphone sales grew 4% year-on-year during the first quarter of 2026

The global smartphone market is heading toward a difficult 2026 as rising component costs force manufacturers to increase device prices, new research has claimed.

New findings from FDM CCS Insight note the primary smartphone market is expected to decline by 14.8% in 2026 as memory shortages continue affecting production.

Known as ‘Weixin’ in China, WeChat has rolled out the AI agent on a phased basis.

Tencent is testing a new AI assistant on its ‘super app’ WeChat, as the company attempts to catch up with its Chinese and global contemporaries.

WeChat – known as ‘Weixin’ in China – is China’s most popular messaging platform with roughly 1.4bn users, and also has functions for social media, ride-hailing and payments. WeChat has rolled out the agent on a phased basis.

Users can interact with the AI agent, called ‘Xiaowei’, via text or voice, and complete tasks by tapping into mini apps. The agent assists with a wide range of tasks, including changing settings, sending messages, ordering food, hailing rides and generating images.

Xiaowei uses WeChat’s own large language model WeLM, while also tapping into DeepSeek to process some queries.

Tencent is closely associated with DeepSeek, reportedly leading a recent $7.4bn round into the AI start-up. The company was also reported to have proposed taking a 20pc stake in the start-up.

Meanwhile, The Information reported last week that Tencent was preparing to purchase Manus back from Meta after China blocked the $2bn acquisition. HSG and ZhenFund are also reportedly looking to buy back Manus using fresh capital.

Despite holding stakes in leading AI companies in the country, Tencent trails behind its peers ByteDance and Alibaba over adoption and advances in AI technology.

Alibaba has integrated travel, maps and e-commerce services into its Qwen AI app, while ByteDance has added agentic functions into its app called Doubao.

The Financial Times reported on Tencent’s plans to launch the embedded AI agent earlier this month, with added pressure from its well-performing contemporaries.

The publication reported that the company made the AI agent roll-out its highest strategic priority, while internal estimates suggest that a full roll-out of  Xiaowei will be very costly for the company. Tencent itself already has an embedded chatbot with search functions in WeChat called Yuanbao.

Don’t miss out on the knowledge you need to succeed. Sign up for the Daily Brief, Silicon Republic’s digest of need-to-know sci-tech news.

Disclaimer: Unless otherwise stated, any opinions expressed below belong solely to the author.

Singapore’s investment companies are riding the AI wave, hoping to capitalise on the boom in the industry. Just slightly over three weeks ago, both Temasek and GIC raised their investment in one of the most promising foundational AI companies, Anthropic, the developers of Claude AI.

It was the second round for Temasek and the third for GIC, after earlier rounds in Feb 2026 and Sept last year.

While we don’t know how much money each of them put in exactly, although GIC co-led two of the latest rounds and Business Times suggested that we’re talking about “billions,” what we do know is what the valuation of Anthropic was at each stage (all the figures are post-money):

• Series F (Sept 2025): US$183 billion
• Series G (Feb 2026): US$380 billion
• Series H (May 2026): US$965 billion

This is important because, next to SpaceX and OpenAI, Anthropic is the most anticipated trillion-dollar tech IPO of the year. Polymarket bets have the odds of it reaching as high as US$1.8 trillion at around 50% or more when it goes public sometime in the fall.

This means that institutional investors are looking at a nearly two-fold return just over the Series H, and as much as 10x since Sept 2025. Not bad for just one year.

This could very well be one of the best investments ever made by Singapore’s funds, which raises a question—why Anthropic and not OpenAI? Why Claude and not ChatGPT?

Less is more

After all, it was ChatGPT that launched the AI revolution in Nov 2022. Since then, all other companies have been seen more as followers than leaders of the race. Google famously suffered a string of embarrassments with its early Bard service, while OpenAI kept gaining steam, leaving competition in the dust.

Today, Google has caught up, after years of improvements and huge reach owed to its search engine monopoly. Both ChatGPT and Gemini are now serving close to 1 billion users each month. Meanwhile, Claude is estimated to be used by anywhere between 20 and 50 million people. Tens of times fewer.

And yet, in terms of revenue, it reportedly pulls in about twice what OpenAI does (close to US$50 billion vs US$25 billion in annualised run rate as of Apr/ May 2026).

The big shake-up came last year, when Anthropic had its own breakthrough moment—the launch of Claude Code.

The smart coding agent allowed it to capture the most lucrative part of the market: the enterprise customer. Corporations around the world quickly adopted CC to vastly improve and increase their in-house coding capabilities, as it was the first agentic AI tool that operated with high accuracy.

OpenAI may have millions of people paying 20 bucks to use ChatGPT, but Anthropic has thousands of companies paying thousands, if not millions, of dollars for intelligent enterprise automation.

The contrast couldn’t be starker: about 85% of Anthropic’s revenue comes from business customers, while 85% of OpenAI’s revenue comes from individual users.

This is why the former is already valued higher—US$965 billion vs US$852 billion—and is also expected to appreciate more following the IPO.

GIC’s and Temasek’s interest in and commitment to Anthropic is, therefore, hardly a surprise.

In fact, it’s quite likely that it had its source in first-hand experience, given how many different businesses both organisations have invested in. They see the tools that the corporations are using and how satisfied they are with them. And if it’s also reflected in the target company’s financials, then that seals the deal.

Anthropic’s competitors are trying to catch up, but so far, neither OpenAI nor Google have managed to create a worthy competitor. Claude Code is successful because it understands its audience: engineers living in the terminal, where Claude can do wonders, performing as a capable junior developer who needs occasional guidance and supervision, but otherwise can do the job himself.

What’s more, once an organisation adopts a technical solution, it’s very difficult to leave it. With a whole universe of plugins, untangling Claude in order to move to a different provider would only be justifiable if another company significantly outperformed it. So far, that doesn’t seem to be a risk.

Unlike fickle consumers, who can drop a subscription overnight, companies move slowly and avoid change unless it’s absolutely necessary.

While its competitors pursued mass market scale, Anthropic focused on the niche with the biggest needs and deepest pockets. This has turned out to be a masterstroke, which Singapore is about to profit from as well.

• Read other articles we’ve written on Singapore’s current affairs here.

Featured Image Credit: rafapress/ depositphotos

Alibaba Cloud on Sunday released HappyHorse 1.1, a major upgrade to its AI video generation model that the company says delivers production-ready video synthesis across core content creation scenarios. The model is now live on Alibaba Cloud Model Studio with full API access for enterprise customers and developers, accompanied by a 40% sitewide launch discount for the first two weeks.

The release arrives at a moment of remarkable upheaval in the AI video generation market — and Alibaba appears keenly aware of the timing. OpenAI discontinued Sora after it proved financially unsustainable. ByteDance indefinitely shelved the international rollout of Seedance 2.0 following a barrage of copyright complaints from Hollywood studios. For enterprise procurement teams that had been evaluating or integrating those tools into marketing, advertising, and content production workflows, the competitive landscape has contracted sharply in a matter of months.

That contraction creates both an opportunity and a test for Alibaba. HappyHorse 1.1 is not a research demo or a consumer toy — it is an API-first product built for integration into enterprise software stacks, priced for volume, and backed by a $52.7 billion global infrastructure buildout. Whether it can convert technical capability into enterprise adoption, particularly in Western markets navigating intensifying U.S.-China tech tensions, will determine whether Alibaba can establish itself as a serious player in the generative video market that analysts expect to reach tens of billions of dollars by the end of the decade.

How HappyHorse climbed from anonymous benchmark entry to top-ranked video model

HappyHorse first appeared in early April as an anonymous submission on the Artificial Analysis Video Arena, an independent benchmarking platform where real users compare model outputs in blind, side-by-side evaluations. The model immediately claimed the top position in both text-to-video and image-to-video rankings. Alibaba was subsequently confirmed as the creator, revealing it was built by the company’s ATH (Alibaba Token Hub) AI Innovation Unit — a team previously part of the Future Life Lab under the Taobao and Tmall Group before a strategic organizational restructuring.

According to Arena.ai, HappyHorse 1.0 now holds the No. 2 position across all three Video Arena leaderboards. The platform noted the model scores 1,444 in both text-to-video and image-to-video categories, leading Google’s Veo-3.1 (with audio) by 69 points in text-to-video and xAI’s Grok-Imagine-Video by 23 points in image-to-video. In Elo-based ranking systems like Arena’s, models gain or lose points based on whether users prefer their outputs in head-to-head comparisons, meaning persistent double-digit leads reflect a consistent quality gap as perceived by human evaluators — not a statistical fluke.

The model’s architecture helps explain why. According to community-compiled technical documentation, HappyHorse is built around a 15-billion-parameter unified self-attention Transformer that processes text, image, video, and audio tokens within a single token sequence. Unlike many competitors that stitch together separate models for video and audio, HappyHorse operates as a unified system that handles all modalities in a single generation pass, eliminating the need for third-party dubbing or post-processing audio tools. For enterprise buyers evaluating total cost of ownership, that architectural simplicity translates directly into fewer integration points, fewer vendor dependencies, and faster time to production.

What the 1.1 upgrade fixes — and why it matters for commercial video production

The 1.1 upgrade targets a set of pain points that enterprise video production teams know intimately. Alibaba Cloud described the release as “systematically optimized across core content generation scenarios,” and the specific improvements reveal a model that has been tuned for commercial deployment rather than viral social media demos.

The most consequential upgrade is multi-image reference capability, which Alibaba calls R2V (Reference-to-Video). The feature allows users to upload multiple character reference images and maintain consistent identity across generated video — directly addressing one of the hardest problems in AI video production, where subjects tend to drift in appearance between frames or shots. For brands producing advertising campaigns, product videos, or serialized marketing content, identity consistency is not a nice-to-have; it is a requirement that has historically forced teams back to traditional production methods.

Motion quality receives a significant overhaul, with what Alibaba describes as “strengthened motion modeling” that addresses prior limitations in speed and fluidity. The company also made targeted improvements to visual texture, specifically calling out the elimination of “facial oiliness,” “over-sharpening,” and “unnatural textures” — artifacts that have plagued commercial AI video since the technology emerged and that immediately signal to viewers that content is machine-generated.

Two additional upgrades round out the release. HappyHorse 1.1 improves audio-visual synchronization, including what Alibaba claims is “zero-drift lip sync” for dialogue scenes and context-aware speech pacing — building on the 1.0 version’s already notable ability to generate up to 15 seconds of 1080p video with synchronized audio output. The model also improves instruction-following for long and complex prompts, a critical differentiator for enterprise users who need to specify precise camera movements, lighting conditions, and narrative beats in a single generation pass rather than iterating through dozens of attempts.

Sora’s collapse and Seedance’s freeze leave enterprise buyers with fewer choices than ever

The competitive context surrounding this launch is unusually favorable for Alibaba, and it is worth understanding why.

OpenAI’s Sora web and app experiences were discontinued on April 26, with the Sora API set to follow on September 24. The shutdown came after the product proved financially untenable: Sora cost roughly $1 million per day to operate but generated only about $2.1 million in total revenue, while active users dropped from a peak near 1 million to under 500,000. For enterprise teams that had integrated Sora into production pipelines, the abrupt withdrawal underscored the risks of depending on AI products that lack a sustainable business model — a cautionary tale that procurement officers are unlikely to forget quickly.

ByteDance’s Seedance 2.0, which many considered Sora’s most formidable successor, ran into a different kind of wall. Netflix, Warner Bros., Disney, Paramount, and Sony sent legal threats to ByteDance over allegations of systematic copyright infringement after users generated viral clips featuring Hollywood intellectual property. ByteDance indefinitely postponed the international launch, and the global rollout remains suspended.

That leaves Google’s Veo 3.1 as the primary Western competitor in the enterprise video generation space. But Alibaba’s Arena rankings suggest HappyHorse is outperforming Veo on user-perceived quality, and the 40% launch discount on Alibaba Cloud Model Studio could make HappyHorse significantly cheaper at scale. At the 1.0 level, pricing through third-party API platforms ran roughly $1.82 per 10-second clip at 720p and $3.12 at 1080p. With the promotional pricing, HappyHorse 1.1 could bring production-quality AI video generation within reach of mid-market companies and agencies that previously considered the technology too expensive for anything beyond experimentation.

Alibaba’s $52.7 billion infrastructure bet gives HappyHorse a distribution advantage rivals can’t match

HappyHorse 1.1 does not exist in isolation. It sits atop a global infrastructure offensive that distinguishes Alibaba from pure-play AI model companies that build impressive technology but lack the physical and commercial machinery to serve regulated enterprise customers at scale.

Just five days before the HappyHorse 1.1 launch, Alibaba Cloud opened its first data centers in France, establishing its third European hub after Germany and the United Kingdom. The Paris region features two availability zones, bringing the company’s global footprint to 105 availability zones across 32 regions. “The expansion of our cloud infrastructure into France reinforces our ongoing commitment to empowering European businesses with sovereign, secure, and intelligent solutions,” said Dr. Feifei Li, Alibaba Cloud’s CTO and president of international business, in the company’s announcement. In Japan, the company opened its fifth data center in Tokyo on June 19.

As reported by Data Center Dynamics, CEO Eddie Wu has committed to investing $52.7 billion in building a “unified global cloud network,” with the company later considering increasing this to $69 billion. This year alone, Alibaba has launched new regions in Mexico, Thailand, Malaysia’s Johor, and France. The France deployment is also part of Alibaba Cloud’s plan to roll out enterprise-grade agentic AI services across Europe in the second half of the year, including AgentRun (a development platform for AI agents), STAROps (an intelligent operations platform), and ACS Agent Sandbox (which provides hardware-level security isolation for agent workloads).

The infrastructure buildout serves a dual purpose for a product like HappyHorse. Running a 15-billion-parameter video generation model with integrated audio is extraordinarily compute-intensive, and having local infrastructure reduces latency for enterprise API calls while keeping customer data within regulatory boundaries. For European buyers operating under the European Commission’s new tech sovereignty framework — published June 3 with the explicit goal of protecting the bloc’s “digital independence” — the ability to run AI video generation workloads on locally hosted infrastructure is not a luxury. It is increasingly a compliance requirement.

The Pentagon listing and geopolitical risk loom over Alibaba’s Western ambitions

Alibaba’s global push is unfolding under significant geopolitical headwinds that enterprise buyers cannot afford to ignore. The Pentagon added Alibaba, along with BYD and Baidu, to its list of Chinese military companies on June 8, preventing them from securing U.S. defense contracts. Alibaba rejected the designation, saying it is “not a Chinese military company nor part of any military-civil fusion strategy.”

The listing does not automatically trigger sanctions, and it does not directly restrict commercial transactions between private U.S. companies and Alibaba. But it adds a layer of reputational and regulatory complexity to procurement decisions, particularly for companies with U.S. government exposure, defense supply chain connections, or transatlantic operations. Enterprise technology purchases are rarely evaluated on technical merit alone — vendor risk assessments, board-level compliance reviews, and geopolitical scenario planning all factor into buying decisions for cloud infrastructure and AI tooling.

For European customers specifically, the calculus is layered in a different way. The continent’s growing emphasis on digital sovereignty cuts in two directions simultaneously: it creates demand for alternatives to the dominant U.S. hyperscalers (Amazon Web Services, Microsoft Azure, and Google Cloud control roughly 70 percent of European cloud infrastructure revenue, according to Synergy Research Group), but it also raises questions about whether a Chinese provider represents a meaningful improvement in strategic autonomy. Alibaba’s strategy of building sovereignty-compliant infrastructure in-market is a direct attempt to answer that question — but the Pentagon listing ensures it will be asked repeatedly.

What enterprise teams should watch as the AI video market consolidates

The practical implications of HappyHorse 1.1 for enterprise teams are substantial. HappyHorse supports four modes of generation — text-to-video, image-to-video, subject-to-video, and the newly added video editing — covering the full spectrum of commercial video needs from ideation through production to post-production, all with integrated audio at no additional cost. That breadth of capability, delivered through a single API endpoint, simplifies what has historically been a fragmented and expensive production pipeline.

The question going forward is whether Alibaba can convert benchmark dominance and competitive timing into durable enterprise relationships. The company plans to release HappyHorse through Alibaba Cloud Model Studio with full enterprise SLAs, security certifications, and regional compliance — the table stakes that separate research breakthroughs from production-grade services. Watch for customer disclosures, usage metrics, and whether third-party platforms like fal.ai and Atlas Cloud (which already host HappyHorse 1.0) update to the 1.1 version quickly, which would signal genuine developer demand beyond Alibaba’s own ecosystem.

The AI video generation market entered 2026 with three credible enterprise contenders. One is dead. One is frozen. And the one still standing is a Chinese company backed by $52.7 billion in infrastructure spending, ranked No. 2 across every major independent benchmark, and offering a 40% discount to anyone willing to place the bet. In enterprise technology, the best product does not always win — but it rarely loses when the competition has already left the field.

Auralink is seeking funds to accelerate product development and engage with visually impaired communities.

Founders behind the University College Dublin (UCD) student start-up Auralink have won the annual NovaUCD accelerator programme competition for emerging student entrepreneurs.

Auralink is developing an AI-powered assistive technology (AT) platform that allows visually impaired people to use smart glasses for better independence and safety. The platform is being developed by undergraduate student in economics and history Scott Nagle, and graduate entry medicine student Suyun Zheng.

Auralink combines smartphone software with smart glasses to provide users with real-time environmental sensing, object recognition and navigation support through audio feedback. UCD said that Auralink is being developed for 24/7 use across daily activities at home, as well as navigating public transport and unfamiliar environments.

The two students won the ‘One to Watch’ award from NovaUCD following the four-week-long accelerator programme for budding student entrepreneurs from the university.

The team, who were judged by a panel including Atlantic Bridge investment manager Dominik Leisi, New Frontiers programme manager Susanne L’Estrange and senior advisor for start-ups at Enterprise Ireland Michael O’Dea, also won €3,000 alongside the award.

“While existing solutions provide valuable support for people with visual impairments, they offer limited information about the surrounding environment,” said Zheng.

“Public transport, for example, remains challenging, especially when identifying platforms and reading timetables.

“We are focused on addressing this gap with software that runs on smartphones and smart glasses to make assistive technology more accessible, in line with the UN SDGs (sustainable development goals) of good health and well-being, and reducing inequalities.”

Physical and digital accessibility is far from being universally available across Ireland. Despite advancements in technology, including various AI tools and products such as smart glasses, some estimates say hundreds of thousands in Ireland could be living without the AT support that they need. AT users-turned-digital-coaches are attempting to bridge some of these gaps.

Auralink is seeking funding to help accelerate product development, user testing, pilot deployments and engagement with visually impaired communities in Ireland and internationally, Nagle said.

“We are also seeking mentors, industry connections and introductions to accessibility organisations, healthcare providers, transport operators and potential commercial partners,” he said.

The annual accelerator programme – now in its 12th year – supports the university’s undergraduate and postgraduate students in developing and growing business ideas into start-ups. The programme offers workshops, mentoring and pitching sessions.

This year’s cohort was made up of 13 early-stage student ventures and 17 participants. Over the last 12 years, some 105 early-stage ventures and 240 students have completed the programme.

Simon Factor, senior manager for new ventures at NovaUCD, said: “A key focus of this annual UCD accelerator programme is to provide the participating undergraduate and postgraduate students with the skills, the confidence and the opportunity to further refine their ideas, and hopefully in time launch start-ups at home and further afield.

“The pitches delivered at NovaUCD by the enthusiastic students, on a range of business ideas – from AI-powered assistive technology to ed-tech, to medtech, to sustainability, to robotics – were all excellent, and I would like to congratulate all the participants for successfully completing this year’s programme.

“I would especially like to congratulate Auralink on being named the overall winner, and I wish Scott and Suyun every success as they progress their new venture in the months ahead.”

Don’t miss out on the knowledge you need to succeed. Sign up for the Daily Brief, Silicon Republic’s digest of need-to-know sci-tech news.

Artificial intelligence is the transformative, strategic technology of the early 21st century. It is significantly reshaping practically every aspect of our lives, including in ways that probably no one anticipated. Its rate of adoption and impact have been unprecedented when compared with other technologies.

AI as a distinct field was formally established in 1956 at the Dartmouth Summer Research Project on Artificial Intelligence, proposed by John McCarthy, Marvin Minsky, Nathaniel Rochester, and Claude Shannon. In their August 1955 proposal for the research project, the scientists introduced the term artificial intelligence and envisioned machines capable of simulating human intelligence.

AI is the “science of making machines do things that would require intelligence if done by men,” as defined by Minsky. The professor received the ACM Turing Award, which is often called the “Nobel Prize in computing.”

Since AI’s humble beginnings 70 years ago, it has evolved significantly in its capabilities, gained prominence, and earned widespread adoption across many areas including business, education, finance, health care, industry, and the military.

IEEE’s contributions to the progress and adoption of AI throughout its journey are substantial and multifaceted.

As we celebrate AI’s 70th birthday, understanding its history, current status, limitations, and concerns is key to harnessing it for good.

The technology’s roller-coaster evolution

Although AI emerged as a distinct field in 1956, its intellectual roots extend back further. The ideas and theories that underpin AI predate modern computers such as the ENIAC, unveiled in 1946.

In 1943 Warren Sturgis McCulloch, a neurophysiologist and cybernetician, and Walter Pitts, a logician working in computational neuroscience, were inspired by the human brain. The two devised mathematical models of artificial neurons, demonstrating that artificial neural networks could perform logical computation.

Frank Rosenblatt, a Cornell psychologist, later advanced those ideas by developing the perceptron, an early neural network that laid the foundation for modern machine learning and deep learning.

A major milestone came in 1950, when celebrated computer scientist Alan Turing posed the question, “Can machines think?” In his 1950 landmark paper “Computing Machinery and Intelligence,” published in Mind, he explored the nature of machine intelligence. He introduced the “imitation game,” later known as the Turing test, as a practical means of evaluating it. The test remains an influential concept in AI and the philosophy of intelligence, as I discussed in my article “The Turing Test at 75: Its Legacy and Future Prospects,” published in IEEE Intelligent Systems.

Claude Shannon, recognized as the father of information theory, explored the potential of machines for complex reasoning tasks in his 1950 article “Programming a Computer for Playing Chess,” published in Philosophical Magazine.

In 1956 AI became a formal discipline, inspiring scientists to explore and advance it further. John McCarthy developed Lisp in 1958, and it became the dominant programming language for AI research and development. In 1959 Arthur Lee Samuel, a computer science professor at Stanford, introduced the term machine learning to describe programs that could improve their performance through experience.

In the early 1980s, renewed enthusiasm and government funding fueled the development of symbolic AI, a rule-based expert system (also known as a knowledge-based system) that encodes domain-specific knowledge as sets of rules. A notable example was MYCIN, designed to diagnose infectious diseases.

Although successful in limited domains, expert systems’ inherent limitations have restricted their broader adoption. Expert refers to a computer system that mimics human experts in a specific domain. It was popular in the early days of AI, and subsequently disappeared with advances in AI such as neural networks and machine learning.

AI’s journey was marked by periods of soaring expectations and disappointing progress, known as “AI winters,” during which funding, interest, and confidence declined. Analyses of the episodes revealed recurring causes and insightful lessons for the field.

A new phase of growth—often described as “AI spring”—emerged in the 2010s with advances in deep learning, the rise of large language models, the transformer architecture, and generative AI (GenAI).

“The imperative before us today is not only to advance AI’s capabilities but also to ensure that it remains human-centered, trustworthy, ethical, and dedicated to enhancing human well-being and societal progress.”

Unlike earlier approaches that processed information sequentially, a transformer model analyzes an entire sequence of text or audio, assessing the importance of each word or component relative to others, enabling dramatic advancements in GenAI and its applications.

Ashish Vaswani, a former computer scientist at Google, and his colleagues at Google Brain introduced the transformer architecture that underpins today’s generative AI systems in their influential 2017 paper “Attention Is All You Need.” Vaswani and Sam Altman—chief executive of OpenAI, which offers ChatGPT—are widely regarded as the masterminds behind the GenAI revolution.

AI reached new heights with the public release of ChatGPT in 2022, followed quickly by a wave of chatbots and generative AI tools that accelerated global interest.

More recently, the rise of agentic AI systems capable of increasingly autonomous operation has expanded AI’s capabilities and impact.

AI’s 70-year journey reflects an extraordinary interplay of vision, experimentation, setbacks, innovation, and impact.

For further information and diverse perspectives on AI history, check out my curated collection of articles.

Strengths and promises

AI’s pragmatic strength lies in its ability to process information, recognize patterns, and perform cognitive tasks at an unprecedented speed and scale. It can analyze vast amounts of data, extract insights, and identify trends or anomalies that are difficult for humans to detect. The programs can automate routine tasks and repetitive knowledge work, improve productivity, and reduce costs.

Chatbots and other forms of GenAI can answer queries and rapidly create text, images, videos, music, software code, educational materials, and other content on the fly in response to a user’s prompts, accelerating information-gathering, innovation, and decision-making. AI summarizes, translates, and rephrases text effectively and can assist in idea generation. It also facilitates natural-language interactions, making technology more accessible to nonexperts and the diverse global community. Its multimodal capabilities enhance its usefulness across diverse domains. Additionally, it can serve as a powerful collaborator, augmenting creativity and problem-solving capacity rather than replacing human intelligence.

AI is transitioning from standalone tools to autonomous, goal-driven systems. Agentic AI systems that can plan, act, and adapt with minimal human oversight are on the rise, enabling large-scale impact.

The 400-page AI Index 2026, published by the Stanford Institute for Human-Centered AI, reveals the technology’s enhanced capabilities and unprecedented adoption rates, outpacing those of the telephone, the television, the personal computer, and the Internet.

For a deep exposition on the current state of AI, read this analysis from IEEE Spectrum, which also published the “Great AI Reckoning” special report.

Weaknesses and concerns

Along with its benefits, AI presents significant risks and concerns. They include biased, discriminatory, and harmful responses; a lack of transparency and explainability in decision-making; privacy violations from data collected for AI training; and cybersecurity vulnerabilities including AI-powered attacks.

AI systems can hallucinate, generating confident but incorrect or fabricated information. Moreover, AI can facilitate and amplify the spread of misinformation, deepfakes, and manipulated content, undermining public trust and driving the algorithmic manipulation of public opinion. The flattering, people-pleasing, or affirming behavior known as AI sycophancy can be harmful as well.

Overreliance on AI could erode human judgment, critical thinking, and decision-making skills. And autonomous systems can make errors with serious consequences in critical domains including defense, health care, and transportation.

The technology’s development and deployment, therefore, must be guided by informed understanding, sound judgment, and responsible governance. In assessing AI’s suitability for any application, its capabilities, advantages, limitations, and risks must be carefully and holistically considered.

IEEE’s contributions

IEEE has not merely documented and disseminated AI’s progress. It has actively fostered, standardized, and guided it toward further advances and responsible use for the benefit of humanity. IEEE maintains a hub for information on its AI activities that is a valuable resource for researchers, developers, regulators, and users.

IEEE publishes 11 AI-focused journals that advance the frontiers of knowledge, including IEEE Intelligent Systems. In its AI at 70 commemorative issue, Intelligent Systems identified the 10 most influential AI articles published since 2000. The magazine, produced by the IEEE Computer Society, has inducted 10 pioneers into its AI Hall of Fame, honoring their contributions and impact on technology and society.

To foster AI research and development, since 2006, the magazine has recognized the field’s rising stars through its AI’s 10 to Watch awards. The biennial awards spotlight outstanding contributions of young researchers and professionals. Nominations for this year’s awards are open until 1 July.

Since the early days of AI, the IEEE Computer, Computational Intelligence, and Systems, Man, and Cybernetics societies have been among those that have fostered AI research and practice. The Computer Society offers a guide to becoming an AI developer.

IEEE and its societies sponsor more than 100 AI conferences annually. The conference archives are available in the IEEE Xplore Digital Library.

The IEEE Learning Network offers more than 200 courses across AI-related areas.

The IEEE Standards Association has developed more than 100 AI-related standards. Its CertifAIEd program promotes ethical design and deployment of autonomous intelligent systems.

The Institute has featured several IEEE members who have developed AI-driven applications, such as Abhishek Appaji, who has created tools to help detect psychiatric disorders.

Shaping AI’s future

The history of AI helps us understand the motivations behind developments and inspires and guides us toward the next phase of the technology’s innovation and revolution. AI’s trajectory is bound to be shaped by the collective choices we make now and in the future.

As Turing wrote in his 1950 landmark article, “We can only see a short distance ahead, but we can see plenty there that needs to be done.”

The imperative before us today is not only to advance AI’s capabilities but also to ensure that it remains human-centered, trustworthy, ethical, and dedicated to enhancing human well-being and societal progress.

The AI-driven memory shortage has now reached the oldest DRAM standard still in production. DDR2 contract prices rose 55 to 60 percent in the second quarter of 2026, according to Taiwanese market intelligence firm TrendForce, with a further 35 to 40 percent increase forecast for the third quarter.

The price surge is being driven by hardware makers downgrading their memory specifications to secure supply. TrendForce says some manufacturers are replacing DDR4 designs with DDR3, while others are swapping DDR3 components for DDR2, a standard that first shipped in 2003. The downgrades are a response to continued shortages in mainstream DRAM and rapidly rising contract prices across every memory generation.

The Register, which first reported TrendForce’s findings, noted that it is difficult to imagine modern PC processors supporting memory types this old. The downgrades are more likely affecting embedded systems, industrial equipment, networking hardware, and other devices where older memory standards remain in use.

The root cause is the same one that has been reshaping the entire memory market since late 2025. Samsung, SK Hynix, and Micron have redirected wafer capacity from consumer and commodity DRAM to high-bandwidth memory for AI data centres, where margins run at 70 percent or higher. Every wafer allocated to an HBM stack for an Nvidia GPU is a wafer denied to a consumer laptop, a smartphone, or an industrial controller.

The shortage cascaded downward through memory generations. DDR5 and DDR4 prices rose first, and as those components became scarce, buyers turned to DDR3. Now that DDR3 supply is tightening, the pressure has reached DDR2, a product so old that most of the industry had written it off as a low-margin afterthought.

The supply picture for DDR2 is especially fragile because only a handful of companies still make it. Taiwan’s Winbond and ESMT are the key suppliers. Winbond is gradually winding down DDR2 production and reallocating its capacity toward higher-margin products including DDR3, DDR4, and LPDDR4, according to TrendForce.

ESMT is moving in the opposite direction. The company plans to maximise DDR2 production within its existing wafer allocation at foundry partner Powerchip Semiconductor Manufacturing Corporation, concentrating resources on the segment to capture the demand that Winbond is leaving behind. The divergence means Winbond is removing DDR2 supply faster than ESMT can replace it.

The consequences of the broader memory crisis are already visible across consumer electronics. GoPro issued a going-concern warning after memory prices rose 80 to 115 percent, and PC prices have climbed by double-digit percentages. IDC projects that smartphones, PCs, and tablets could see price increases of 10 to 20 percent by the end of 2026.

Some relief is being planned, but it will arrive slowly. SK Hynix aims to double its silicon wafer output capacity over the next five years, a timeline its chairman announced at Computex in June, while Micron expects what it calls meaningful new capacity at its Virginia fabrication plant in 2027 and 2028. Neither commitment addresses the immediate shortage.

Chinese manufacturer CXMT has begun supplying DDR5 to Western brands including Corsair, offering a potential alternative source for mainstream memory. But CXMT is also converting roughly 20 percent of its own capacity to HBM because the margins are too attractive to resist, limiting how much consumer relief it can provide.

The fact that 2003-era memory components are now experiencing 60 percent quarterly price jumps illustrates how deeply the AI reallocation has distorted the semiconductor supply chain. The shortage is not confined to cutting-edge products. It has reached the bottom of the technology stack, affecting components that most of the industry assumed would remain cheap and abundant indefinitely.

from the good-deals-on-cool-stuff dept

MagStack is the perfect on-the-go wireless charging station that also transforms into a floating stand for smartphone FaceTime or video playback while charging. This 3-in-1 foldable design featuring 3 wireless charging spots, enables charging for up to 3 devices simultaneously, including iPhone, Apple Watch, AirPods Pro, AirPods with Wireless Charging Case, other Qi-compatible Android phones, and Bluetooth earbuds. With its versatile foldable design, MagStack also folds into a space-saving single-device charger for your phone or earbuds. It’s on sale for $45.

Note: The Techdirt Deals Store is powered and curated by StackSocial. A portion of all sales from Techdirt Deals helps support Techdirt. The products featured do not reflect endorsements by our editorial team.

Filed Under: daily deal

Cybersecurity researchers have identified nearly 7,000 fraudulent Amazon-themed domains registered in the six months leading up to Prime Day 2026, which begins on 23 June. Check Point Research tracked 6,843 new domains created between December 2025 and May 2026, with registrations peaking at 1,446 in April and remaining elevated at 1,267 in May.

Of the total, 9.2 percent were classified as malicious or suspicious. The rate accelerated sharply in early June: during the first week of the month, one in every 13 newly registered Amazon-themed domains was flagged, according to Check Point’s analysis.

Prime Day 2026 runs from 23 to 26 June across 22 countries, with four additional markets joining later in the summer, according to Amazon’s official event page. The extended four-day window and global reach make it a high-value target for phishing operations, which follow the same seasonal playbook that researchers documented around the FIFA World Cup, where over 13,000 fraudulent domains appeared in the months before kickoff.

The 💜 of EU tech

The latest rumblings from the EU tech scene, a story from our wise ol’ founder Boris, and some questionable AI art. It’s free, every week, in your inbox. Sign up now!

The phishing infrastructure includes fake Amazon storefronts designed to harvest credit card numbers, spoofed login pages that steal account credentials, and email campaigns with subject lines such as “Refund Due, Amazon System Error” that direct recipients to counterfeit sites. Check Point flagged one campaign using a sender address mimicking Amazon’s customer service domain closely enough to bypass casual inspection.

A notable cluster targeted Spanish-speaking shoppers. Check Point identified 46 domains registered under the “amazoncredito” pattern, all linked to a single registrant and aimed at Latin American markets where Amazon has been expanding its Prime membership. Five of six “amazon-prime” top-level domain variants were already classified as malicious at the time of the report.

The tactics are not new, but the scale keeps growing. Google recently sued a Chinese cybercrime ring that used AI to generate phishing code and operated one million fraudulent domains, illustrating how cheap and automated domain-based fraud has become. Check Point’s findings suggest that Amazon-themed operations are following the same industrial pattern, with thousands of domains registered months in advance and activated as shopping events approach.

Check Point recommended that shoppers type amazon.com directly into their browser rather than clicking links in emails or ads, enable two-factor authentication on their Amazon accounts, and treat any unsolicited refund notification as suspicious. The company also advised looking for HTTPS and padlock icons, though it noted that fraudulent sites increasingly use valid SSL certificates to appear legitimate.

The timing is significant because Prime Day has become one of the largest online shopping events globally, generating billions in revenue and drawing millions of first-time deal hunters who may be less familiar with phishing tactics. Amazon has not publicly commented on Check Point’s findings.