Discussion | Link

Table of Contents FastAPI for MLOps: Python Project Structure and API Best Practices Introduction What You Will Build and Learn Why Software Engineering Comes First in MLOps Best Practices Where This Fits in the Overall Curriculum Python Project Structure Best… The post FastAPI for MLOps: Python Project Structure and API Best Practices appeared first on PyImageSearch.

In the pursuit of powerful and stable quantum computers, researchers at Chalmers University of Technology, Sweden, have developed the theory for an entirely new quantum system – based on the novel concept of ‘giant superatoms’. This breakthrough enables quantum information to be protected, controlled, and distributed in new ways and could be a key step towards building quantum computers at scale.

Discussion | Link

Data protection provider Commvault Systems Inc. today announced new artificial intelligence capabilities that it says will allow organizations to accelerate AI adoption with confidence while maintaining control over data, agents and recovery. The new capabilities allow enterprises to activate AI safely, discover and govern AI agents and build and control agentic workflows from within Commvault Cloud. As […] The post Commvault rolls out AI capabilities to secure agentic workflows and data appeared first on SiliconANGLE.

Overall, investors put $27.4 billion to work across seed- through growth-stage financings for Asian companies in Q1, per Crunchbase data. That’s up about 20% from the prior quarter and nearly double year-ago levels, with total funding also hitting its highest level in more than three years.

As major news outlets cut off the Wayback Machine, journalists and advocacy groups are rallying to protect the Internet Archive’s vast collection of web pages.

Many companies are taking a slower, more controlled approach to autonomous systems as AI adoption grows. Rather than deploying systems that act on their own, they are focusing on tools that assist human decision-making and keep control over outputs. This approach is especially clear in sectors where errors carry real financial or legal risk. One […] The post Companies expand AI adoption while keeping control appeared first on AI News.

The developers of Pixel Societies are using AI agents to simulate social interactions. It's an attempt optimize the process of choosing new colleagues, friends, and even romantic partners.

Discussion | Link

PLUS: Generate editable infographics in 15 minutes with AI

Moovila, the leading AI-driven project automation platform for managed service providers (MSPs), has been named to the 2026 CRN AI 100 list in the AI Software category. This marks the second consecutive year Moovila has been recognized on the prestigious list, which highlights companies driving innovation in AI across the... The post Moovila Earns Multiple AI Industry Honors, Recognised in Second CRN AI 100 first appeared on AI-Tech Park.

Westend61 / Getty Images New, more powerful artificial intelligence (AI) models are announced pretty regularly these days: the latest version of ChatGPT or Claude or Gemini always has new features and new capabilities that its makers are eager for customers to try out. But now Anthropic has announced a new model with great fanfare, but is only giving access to a select handful of users. In what the New York Times calls a “terrifying warning sign” of the model’s power, the company has instead started an initiative called Project Glasswing to use the model for good instead of evil. Why? Early reports indicated that the model, with instruction, had been able to move outside a contained testing “sandbox” and send an email to a researcher. A little alarming, perhaps. But more significantly, Anthropic claims Mythos has uncovered software vulnerabilities and bugs “in every major operating system and every major web browser”. Finding hidden vulnerabilities In one remarkable example, the model found a flaw in OpenBSD, a security-focused operating system used in firewalls and routers, which had gone undetected for 27 years. According to Anthropic, it also found a 16-year-old vulnerability in FFmpeg, a little-known but widely used behind-the-scenes piece of software that helps computers, apps, and websites handle audio and video files. Anthropic also says Mythos found several vulnerabilities in the kernel of the Linux operating system, and chained them together in a way that could give an attacker complete control of a machine. Anthropic’s internal testing (which has not been independently verified) showed the Mythos model was far more successful than earlier models at turning software bugs into working exploits. Anthropic Anthropic’s internal assessment of the model highlights both its technical promise and the need for vigilance. The report outlines a hypothetical risk that an advanced AI might exploit its access within an organisation, but concludes that the model poses a very low threat of harmful autonomous actions. In other words, it is unlikely to “go rogue” – but may follow human directions to do things that cause harm. Why Anthropic is keeping Mythos off‑limits Anthropic says it decided not to release the model publicly because of its capabilities and the potential risks it poses. At the same time, the company launched Project Glasswing. The effort brings together a broad coalition of tech companies such as Microsoft, Amazon, Google, Apple, Cisco and NVIDIA, open-source organisations such as the Linux Foundation, and major financial actors such as JPMorganChase, to channel Mythos towards cyber defence rather than misuse. The idea is to give defenders a head start to find and fix weaknesses in critical software before similar AI capabilities become widely available to attackers. Reading between the lines of Anthropic’s messages This is not the first time an AI firm has decided a model was too powerful to release widely. In 2019, years before the ChatGPT era, OpenAI did something similar with its (now quite primitive-looking) GPT-2 model. (Dario Amodei, now chief executive of Anthropic, was a key OpenAI researcher at the time.) However, this doesn’t mean these announcements should not be taken seriously. Anthropic has published unusually detailed material for a model it is not widely releasing. Reports suggest US authorities convened major US bank CEOs in Washington to discuss the cyber risks associated with Mythos. However, we should exercise caution about Anthropic’s claims, because outsiders cannot yet verify most of the underlying evidence. Anthropic says more than 99% of the vulnerabilities it found are still undisclosed because they have not yet been patched. That is responsible disclosure, but it also means the public is being asked to trust a great deal it cannot fully inspect. What Mythos could mean for the future of cybersecurity Cybersecurity failures can have real effects on individuals. In Australia, the Optus breach exposed the personal information of about 9.5 million people. In another case, stolen Medibank records included sensitive health information, and some of the data was later released on the dark web. These were not just database problems. They became crises of privacy, identity and trust. That is why Mythos matters. Mythos and other AI models like it could change the basic economics of cybersecurity. In the past, serious vulnerabilities have often stayed hidden simply because nobody found them. And this in turn was because finding them took rare skill, patience, and time. If models like Mythos can scan the hidden plumbing of the internet – operating systems, browsers, routers, and shared open-source code – at an unprecedented scale, then what is now specialised hacking could become a routine and automated process. For organisations and software development firms, Mythos is a double-edged sword. It could rapidly uncover hidden flaws in their own code, but it also raises the fear attackers could find the vulnerabilities first. The implications reach well beyond tech companies. Much of that underlying, invisible software supports many of the services people rely on every day, from electricity and water to airlines, banking, retail and hospitals. What now? So far, cybersecurity and software companies have been remarkably quiet in public about Anthropic’s Mythos. Many firms appear to be waiting and watching, unwilling to signal their stance in case the model exposes weaknesses in their own systems. But developments like Mythos are a reason to stop treating cybersecurity as somebody else’s problem. For now, for individuals, the response is simple: basic cyber hygiene matters more than ever. Update phones, laptops, browsers and routers. Replace unsupported devices. Use a password manager. Turn on multi-factor authentication. Do not ignore patch notices. Those are the immediate steps. Beyond them lies a harder set of questions about AI and cyber security – about who gets access to powerful AI models, who oversees their use, and who decides what counts as the “right hands”. Stan is an member of the Association for Information Systems. Saeed Akhlaghpour does not work for, consult, own shares in or receive funding from any company or organisation that would benefit from this article, and has disclosed no relevant affiliations beyond their academic appointment.

Recognized for advancing real-world AI safety through continuous adversarial testing of AI systems Adversa AI announced today that it won the Artificial Intelligence Excellence Awards 2026 in the Safety and Alignment category. The award is judged by leading AI experts and business executives and recognizes companies with real-world impact in AI innovation. The... The post Adversa AI Wins Artificial Intelligence Excellence in Safety and Alignment first appeared on AI-Tech Park.

Cloudflare brings OpenAI’s GPT-5.4 and Codex to Agent Cloud, enabling enterprises to build, deploy, and scale AI agents for real-world tasks with speed and security.

Discussion | Link

Artificial intelligence is often celebrated as the future of work. It is efficient, innovative and neutral. Yet, for many women in Indonesia’s gig economy, AI feels like a source of mounting pressure. In my recent research on female gig workers in Indonesia, I examine what I call AI colonialism. This term describes how colonial influence persists today through technology and digital systems that maintain control. This concept captures how powerful actors use AI – often based in the Global North – to exploit workers in the Global South. Much like historical colonialism, this digital iteration relies on the extraction of data, labour and resources to cement unequal power relations. In Indonesia, AI-driven platforms like ride-hailing and e-commerce draw on informal labour but push the risks and responsibilities back onto workers. But women pay the highest price because algorithms fail to recognise the realities of care work, safety concerns and social norms. AI and the gendered restructuring of work Indonesia’s labour market has long been defined by informality. Millions are working without formal contracts or social protections. Tech companies like Gojek, Grab, Maxim and Shopee didn’t formalise this workforce – they only digitised it. Drivers are classified as partners rather than employees. This means no minimum wage, no sick pay and no maternity leave. Income is dictated entirely by completed tasks and algorithmic ratings. For women, this structure collides with the so-called “double burden” since they are responsible for paid work and unpaid care. Lia, a 33-year-old food delivery rider, wakes before sunrise to cook and get her children ready for school. It is only after she has cleared her domestic duties that she finally logs into the app. “The system doesn’t know I have children,” she told me. “It only knows whether I am online.” Platform algorithms reward constant, uninterrupted availability. Incentive schemes demand a specific number of trips within narrow time windows – a high bar for those with domestic ties. If Lia logs off to pick up her children, she risks losing potential bonuses. If she reduces her hours due to menstrual pain or fatigue, her performance metrics drop. Neoliberal capitalism relies on a massive amount of unpaid “invisible labour”, such as childcare and housework, but refuses to pay for it or provide a safety net for those who do it. Far from correcting this imbalance, AI systems make things worse. When Cinthia, a female food delivery rider and a single mother of a one-year-old, fell ill and turned off her app for several days, she noticed fewer job offers upon returning. “It felt like the system punished me,” she said. “Now I’m afraid to stop working.” The algorithm does not explicitly discriminate. However, it operates on the assumption of a worker without caregiving constraints – a norm that systematically disadvantages women. Discrimination behind a ‘neutral’ interface The digital economy often claims neutrality. But gender bias persists. Yanti, a 43-year-old ride-hailing driver in Yogyakarta, regularly messages male passengers before pickup: “I am a woman driver. Is that okay?” Many cancel immediately. The app records cancellations. It does not record gender bias. Because Yanti avoids working late at night for safety reasons, she misses out on rush-hour incentives. The system, however, doesn’t account for safety – it simply interprets her absence as lower productivity. Scholars, like Virginia Eubanks, have pointed out that automated systems often mirror and amplify social inequalities rather than eliminate them. In Indonesia’s platform economy, discrimination isn’t necessarily hard-coded. It is a byproduct of a design logic that favours efficiency over equity. In India, women drivers also report earning less on average than their male counterparts, partly due to safety-driven choices regarding timing and route selection. The algorithm does not account for risk in its calculations. It only measures raw output. Safety, surveillance and algorithmic discipline For women drivers, safety is a constant negotiation. Around 90% of the women in our focus group discussions chose food delivery because it felt safer than ride-hailing. Even so, harassment persists in delivery work. Lia shared how a male colleague targeted her with inappropriate comments as they waited for orders. “It’s not only customers,” she said. “Sometimes it’s other drivers.” During the COVID-19 pandemic, gig workers were labelled “essential”. Yet their income dropped dramatically by as much as 67% in early 2020. To cover the loss, many worked 13 or more hours per day. Platforms maintained their rigid performance metrics throughout the crisis. Drivers who are forced to stop working due to illness often see their ratings decline. Health vulnerability was translated directly into an algorithmic penalty. This reflects labour discipline through digital infrastructure: control shifting from foreman to code. AI colonialism is more than just foreign ownership. It is about the way extractive logics are woven into everyday digital systems. Workers bear the burden of labour, data, time and risk – yet the platforms hold all the power over algorithmic governance. Coping, solidarity and everyday resistance Female gig workers have built dense networks of solidarity through WhatsApp and Telegram groups. They share information about policy changes, warn each other about unsafe customers and exchange strategies for navigating algorithmic shifts. If an account becomes “gagu/silent” (receiving few orders), experienced drivers “warm it up” by temporarily boosting its activity. They lend money for fuel. They pool resources for vehicle repairs. When someone faces harassment, others circulate the information quickly to protect fellow drivers. They visited the platform office together when a member was suspended. Rather than waiting to be formally acknowledged as employees, these women build protection among themselves. This “solidarity over recognition” emerges from shared vulnerability as mothers, caregivers and workers in male-dominated spaces. Their mutual aid turns care into a strategy and a form of “everyday resistance” – subtle acts that challenge dominant systems, while reflecting a distinctly feminist ethic of survival through relational solidarity. Beyond innovation narratives AI is not colonial by design. But when embedded in platform capitalism within unequal societies, it can reproduce colonial patterns of exploitation and loss of ownership. If we are serious about building just digital futures, we must move beyond innovation narratives and listen to workers, especially women and vulnerable groups in the Global South. Their stories are a vital reminder that behind every “efficient” algorithm is a human being navigating the delicate balance of survival, dignity and hope. Suci Lestari Yuana menerima dana dari Hibah Penulisan Dosen, Fakultas Ilmu Sosial dan Ilmu Politik, Universitas Gadjah Mada

<blockquote cite="https://bcantrill.dtrace.org/2026/04/12/the-peril-of-laziness-lost/"><p>The problem is that LLMs inherently <strong>lack the virtue of laziness</strong>. Work costs nothing to an LLM. LLMs do not feel a need to optimize for their own (or anyone's) future time, and will happily dump more and more onto a layercake of garbage. Left unchecked, LLMs will make systems larger, not better &mdash; appealing to perverse vanity metrics, perhaps, but at the cost of everything that matters.</p> <p>As such, LLMs highlight how essential our human laziness is: our finite time <strong>forces</strong> us to develop crisp abstractions in part because we don't want to waste our (human!) time on the consequences of clunky ones.</p></blockquote> <p class="cite">&mdash; <a href="https://bcantrill.dtrace.org/2026/04/12/the-peril-of-laziness-lost/">Bryan Cantrill</a>, The peril of laziness lost</p> <p>Tags: <a href="https://simonwillison.net/tags/bryan-cantrill">bryan-cantrill</a>, <a href="https://simonwillison.net/tags/ai">ai</a>, <a href="https://simonwillison.net/tags/llms">llms</a>, <a href="https://simonwillison.net/tags/ai-assisted-programming">ai-assisted-programming</a>, <a href="https://simonwillison.net/tags/generative-ai">generative-ai</a></p>

In this article, we fine-tune the DeepSeek-OCR 2 model using Unsloth for Hindi language OCR. We create a simple Gradio application to run inference and check the diff between the original and the inference result. The post Fine-Tuning DeepSeek-OCR 2 appeared first on DebuggerCafe.

<p>Thanks to a <a href="https://twitter.com/RahimNathwani/status/2039961945613209852">tip from Rahim Nathwani</a>, here's a <code>uv run</code> recipe for transcribing an audio file on macOS using the 10.28 GB <a href="https://huggingface.co/google/gemma-4-E2B">Gemma 4 E2B model</a> with MLX and <a href="https://github.com/Blaizzy/mlx-vlm">mlx-vlm</a>:</p> <pre><code>uv run --python 3.13 --with mlx_vlm --with torchvision --with gradio \ mlx_vlm.generate \ --model google/gemma-4-e2b-it \ --audio file.wav \ --prompt "Transcribe this audio" \ --max-tokens 500 \ --temperature 1.0 </code></pre> <p><audio controls style="width: 100%"> <source src="https://static.simonwillison.net/static/2026/demo-audio-for-gemma.wav" type="audio/wav"> Your browser does not support the audio element. </audio></p> <p>I tried it on <a href="https://static.simonwillison.net/static/2026/demo-audio-for-gemma.wav">this 14 second <code>.wav</code> file</a> and it output the following:</p> <blockquote> <p>This front here is a quick voice memo. I want to try it out with MLX VLM. Just going to see if it can be transcribed by Gemma and how that works.</p> </blockquote> <p>(That was supposed to be "This right here..." and "... how well that works" but I can hear why it misinterpreted that as "front" and "how that works".)</p> <p>Tags: <a href="https://simonwillison.net/tags/uv">uv</a>, <a href="https://simonwillison.net/tags/mlx">mlx</a>, <a href="https://simonwillison.net/tags/ai">ai</a>, <a href="https://simonwillison.net/tags/gemma">gemma</a>, <a href="https://simonwillison.net/tags/llms">llms</a>, <a href="https://simonwillison.net/tags/speech-to-text">speech-to-text</a>, <a href="https://simonwillison.net/tags/python">python</a>, <a href="https://simonwillison.net/tags/generative-ai">generative-ai</a></p>