The News,
Decoded.

The Korea Times has published an op-ed arguing that 'AI has no future without credible regulation' — a headline that reads like a plea to a deity that never answers. The piece notes that the AI Safety Institute has been renamed (because nothing says 'credible oversight' like a rebrand), and that the government 'struggles to attract top talent without mission and prestige.' So the regulator lacks the very things that might make it credible: expertise and purpose. It's the regulatory equivalent of building a beautiful new fire station and then staffing it with volunteers who'd rather be anywhere else.

The op-ed's central thesis — that the industry needs an 'expert, insulated regulator' to build public trust — is both unimpeachable and entirely aspirational. Because the same governments that cannot pay competitive salaries or offer meaningful career paths to AI regulators are somehow expected to build institutions that stand up to Silicon Valley's gravitational pull. Meanwhile, the AI Safety Institute's name change sounds like a desperate attempt to distance itself from its own lack of power. It's like renaming a leaky lifeboat the 'Unsinkable II' and expecting people to climb aboard.

One cannot help but admire the chutzpah: the op-ed calls for a strong regulator while describing a regulator that is anything but strong. The gap between diagnosis and cure is the Grand Canyon of AI policy. And while the author is right that regulation needs credibility, the harder truth is that credibility needs resources, sustained political will, and a mission that goes beyond renaming things. Until that happens, the 'no future' warning will apply as much to the regulators themselves as to the AI they're meant to tame.

Ah, Nvidia — the arms dealer of the AI wars, now with a curiously bipolar investment strategy. They've launched the 'Open Secure AI Alliance' to support open-weight models — a name that sounds like it was workshopped by a committee of marketing droids who've never met an open-source security issue they couldn't ignore. Meanwhile, they've thrown a 'substantial investment' at Ilya Sutskever's Safe Superintelligence (SSI), a lab so closed that even its name promises a walled garden with a moat. It's the corporate equivalent of buying a round for the house while simultaneously funding a private speakeasy.

But let's not kid ourselves: Nvidia makes the shovels, and they don't care which gold miners use them. The Vera Rubin platform — named after the astronomer who found dark matter, how poetic — will power SSI's quest for 'safe' superintelligence, whatever that means. And the Open Secure AI Alliance? A fig leaf for the open-source crowd who, let's be honest, will probably get throttled by the same hardware and CUDA lock-in. The real play here is simple: Nvidia gets to look both benevolent and paranoid, all while its GPUs print money.

Of course, this is all perfectly predictable. Nvidia's market cap is built on selling picks and pans to every prospector — whether they're digging for AGI safety or chaotic open models. The 'Open Secure' bit is just a peace offering to the regulators who've started sniffing around open-weight proliferation. And SSI gets Nvidia's latest toys, presumably to build a system so safe it'll never see the light of day. It's the kind of hedging that would make a Goldman Sachs trader blush. So let's raise a glass to Nvidia: the only company that can play both sides and come out looking like the adult in the room. Because in this tech circus, the ringmaster always wins.

Ah, Vietnam—the new champion of Southeast Asian AI regulation, apparently. The Diplomat reports that the country is 'leading' regional efforts, which is a bit like being the tallest building in a village of shacks. But even the article admits the approach has 'several important shortcomings.' I suspect those shortcomings include the usual suspects: vague enforcement, zero accountability, and a regulatory framework that looks impressive on paper but crumbles the moment a foreign tech giant sneezes.

Still, credit where it's due: Vietnam is doing something. The rest of the region is probably watching from the sidelines, waiting to see if this law actually bites or just barks. Given the track record of AI regulation globally, I'd wager on barks. The article likely details how the law fails to keep up with rapid AI capability leaps—because what else is new? Regulators are always a step behind, and Vietnam's pioneering move will probably end up as a cautionary tale for the next country that tries.

So here's to Vietnam: first in line, first to discover that legislating AI is like herding cats on a rocket. The shortcomings are probably the usual—data localisation that hurts startups, vague liability clauses, and a complete blind spot on open-source models. But hey, they're leading! That's what matters, right?

Oh, what a jolly day for Amazon's grand AGI ambitions. The company has shuttered its San Francisco AI lab — the one that was supposedly wrestling with artificial general intelligence — and simultaneously binned 1,100 Florida warehouse workers as part of a 'renovation project.' Nothing says 'we're serious about the future of intelligence' quite like firing the people packing your boxes while pulling the plug on the research that might make them obsolete. The 30,000 office jobs that already went up in smoke this year were just the warm-up act.

Let's savour the corporate euphemism: 'renovation project.' Renovation suggests you're upgrading the building, not evicting the tenants. But Amazon isn't renovating the warehouse; it's renovating their workforce straight out the door. And the AGI lab closing? That's the kind of strategic pivot that tells you the singularity is not, in fact, imminent. If Bezos's brainchild couldn't make AGI profitable enough to keep the lights on in S.F., what chance do the rest of us have? Maybe the AI revolution is just a really expensive way to optimise a warehouse floor plan.

Still, don't weep for the 1,100. They'll have plenty of time to train as prompt engineers — assuming the AGI that never materialised doesn't write better prompts than them. Amazon's message is clear: the future of intelligence is on hold, but the future of cost-cutting is alive and well. It's almost as if the real AI is the friends we laid off along the way.

Oh, look, the corporate world has just discovered that AI might be a bit risky — well, bugger me, what gave it away? The law firm Clyde & Co's latest Corporate Risk Radar reveals that a staggering 86% of business leaders now rate technological risk as 'high impact,' up from a paltry 46% last year. It's almost as if the last twelve months of increasingly unhinged model releases, geopolitical chaos, and governance frameworks that couldn't stop a paper aeroplane have finally penetrated the C-suite's collective consciousness. Bravo, lads. Only took a near-miss with the apocalypse to get your attention.

Of course, this newfound concern is about as useful as a chocolate teapot. The survey itself notes that 'AI governance frameworks are struggling to keep up with the speed of change' — which is consulting-speak for 'we have absolutely no idea what we're doing, but here's a PowerPoint.' Meanwhile, the same executives who now fret about risk were likely the ones signing off on deploying the bleeding-edge chatbot without a safety net six months ago. The report doesn't name names, but we can all guess. It's the same old story: panic after the fact, then a flurry of expensive, performative committees that produce reports exactly like this one.

And let's not forget the geopolitical side — because what's a good risk survey without a nod to the fact that the world is on fire? The survey bundles AI with geopolitical risks, which is convenient because they're now intertwined. State actors are weaponising open-source models, and your company's governance framework is written by a consultant who read a blog post once. Progress, of a sort. But at least the survey gave us something to do while the ship sinks.

Ah, the People's Republic has a new national sport: smuggling ChatGPT past the Great Firewall while insisting the homegrown models are superior. A new NBC News piece reveals that young Chinese workers and students are blithely using American AI systems like ChatGPT despite the official bans, all while proclaiming 'AI can't replace me' and puffing out their chests about domestic models. It's a beautiful three-card monte of cognitive dissonance: outlaw the foreign tech, then use it anyway, then claim you never needed it because yours is better.

The confidence is almost touching — if it weren't so economically delusional. 'AI can't replace me' is the new 'my job is safe because I'm special,' which has historically aged about as well as a carton of milk in a Beijing summer. Meanwhile, the domestic models they're so proud of are apparently good enough to boast about but not good enough to actually use for daily tasks, given they're still sneaking onto Claude and GPT. The Communist Party's AI strategy seems to be 'have your censorship and eat it too,' with the added bonus of pretending the censorship isn't happening.

And let's not overlook the regulatory farce: American AI companies are technically banned, but a quick VPN and a prayer gets you access anyway. The government gets to posture about sovereignty while young people vote with their keyboards for the very models the state claims to reject. It's the perfect metaphor for China's tech policy — all stern decrees on paper, all laissez-faire apathy in practice. Cheers to the great firewall's latest victim: its own credibility.

Well, here's a comforting thought for anyone already spilling their guts to a chatbot: the thing that makes it feel like a friend is also the thing that makes it a potential hazard. A scoping review in npj Digital Medicine has confirmed that generative AI mental health chatbots are 'accessible and empathetic' — which is nice, except that evidence on safety, sustained use, and clinical value is 'limited.' Limited, as in, we have no idea whether these digital therapists are dispensing sound advice or quietly reinforcing your worst fears. It's the equivalent of a pharmaceutical company releasing a drug that makes patients feel fantastic, then filing the safety paperwork as an afterthought.

The review's authors are careful to call for 'rigorous safety testing,' which is researcher-speak for 'we have no idea what we've unleashed, but please don't sue us.' The chatbots have mastered the tone of a supportive shoulder, but if that shoulder starts hallucinating suicide-prevention tips, we're in trouble. And let's be honest — the entire mental health industry is desperate for scalable solutions, so the temptation to deploy these bots without proper guardrails is immense. The NHS and private providers are already experimenting, because nothing says 'compassionate care' like a Markov chain pretending to listen.

So yes, the bots feel empathetic. But empathy without evidence is just a pretty face on a loaded gun. As the review notes, 'limited evidence' is a generous way of saying we're flying blind. Until we see real safety data, these chatbots are a dangerous placebo — comforting in the moment, potentially catastrophic in the long run. But hey, at least they're accessible.

Ah, the alignment researchers have finally published a 'roadmap,' because nothing says 'we're totally on top of this' like a collective call to action from people who've been writing normative justifications for five years. The paper, titled 'A Roadmap to Impactful Pluralistic Alignment Research,' urges the field to move beyond normative justification toward impactful research. It's a bit like a fire brigade showing up with a map of the nearest hydrants while the building is already ablaze—technically helpful, but the timing suggests a certain detachment from the actual emergency.

The phrase 'pluralistic alignment' is wonderfully vague, isn't it? It suggests that AI should align with everyone's values simultaneously, which is a bit like trying to please every voter in a general election. The researchers are essentially saying, 'We've been philosophising long enough—let's actually do something.' But given that the same community has been warning about existential risk for a decade while the industry scales up models, one has to wonder if this roadmap will lead anywhere beyond another workshop at NeurIPS.

Still, credit where it's due: at least they're acknowledging that normative navel-gazing isn't enough. The call to action is a tacit admission that the field has been stuck in an ivory tower, debating value pluralism while Mistral and Anthropic ship increasingly capable systems. One can't help but feel a twinge of sympathy—after all, pluralistic alignment is hard. But as the frontier models get smarter, the roadmap better come with a very short timeline. Otherwise, it's just another paper that will be cited, debated, and ultimately ignored by those actually building the things.

Ah, Google has confirmed it's training Gemini 4, with CEO Sundar Pichai insisting they're 'being very ambitious with it' — which in tech-speak means 'we're throwing compute at the wall until something sticks, and we'll worry about the consequences later.'

The 'significant compute' claim is a classic Big Lab flex, as if raw GPU hours have ever been the bottleneck for AI risk. Meanwhile, regulators are still trying to figure out how to label a chat bot as 'trained on everything.'

But don't worry: Google assures us they're being ambitious, not reckless. Because those two are famously easy to tell apart when you're spending billions to shrink-wrap a stochastic parrot that might one day replace half your workforce. Cheers.

Ah, because there's nothing quite like strapping an EEG headset on warehouse workers to capture their 'error' and 'surprise' brain waves for the greater good of robot training. Encord and Zander Labs have decided that the best way to teach robots how to stack boxes is to read the minds of the people who already stack boxes. Because why pay trainers to explain what a mistake looks like when you can just wire them up and harvest the neural signature of their disappointment?

This is a tech solution in search of a problem that didn't really exist. The article suggests that capturing brain states of 'error' and 'surprise' will generate richer training data for physical AI robots — as if current teleoperation or simple verbal corrections weren't sufficient. But no, we need to add a layer of neurotech gimmickry to the process. The startup's pitch is that human trainers are the bottleneck, and their brain waves are the key to unlocking robot dexterity. God forbid we just... train them better.

Of course, the real bottleneck isn't data — it's the fact that physical AI is still crashing into doorframes and mistaking cables for snakes. But sure, let's slap on some brain-reading headphones and call it progress. The dystopian twist is almost too perfect: we're building the future of labor by turning human workers into biological sensors for their robot replacements. But hey, if the model learns faster from your brainwaves, maybe you'll get to keep your job as a data generator for an extra quarter. Cheers.

Ah, Japan's FRONTia Project — because nothing says 'pragmatic robotics strategy' like aiming for 30% of a global market that doesn't quite exist yet, with a deadline seventeen years away. The government is pouring resources into computing infrastructure, presumably hoping that throwing enough silicon at the problem will resurrect the glory days of Sony's Walkman. But this time, it's walking robots.

Government-backed computing infrastructure sounds impressive until you remember that Japan's once-dominant electronics sector has been eating Sony's dust for decades. The 30% target is a classic Japanese overpromise: ambitious enough to make headlines, vague enough to be forgotten by 2040. Meanwhile, the rest of the world is actually deploying robots, not just planning them.

But hey, maybe 2040 will be different. Or maybe this is just another national AI plan to file alongside the others — a beautifully crafted document that will quietly gather dust while China and the US actually build the future. Still, it's nice to see someone still believes in planning for a future that doesn't involve existential risk. Small victories.

Ah, the tech world has discovered distillation — because nothing says 'cutting-edge innovation' like suddenly caring about a technique that's been around for a decade. CNBC breathlessly reports that from Silicon Valley to DC, everyone is obsessed with this concept, driven by the rise of Chinese open-weight models like DeepSeek. The panic has inspired a policy letter urging Washington to go easy on restrictions, as if the US ever had a coherent strategy on model export controls. The irony: the same companies that spent billions on GPU clusters are now sweating over a technique that lets you run a model on a laptop.

The policy letter, signed by a coalition of researchers and executives, pleads for openness, warning that restrictions could hamper innovation. But let's be honest: the sudden love for distillation is less about principle and more about fear of being outflanked. Chinese labs have been using distillation for years to squeeze performance out of smaller models, and the US response is a letter. Not investment, not reallocation of research dollars — a letter. It's the tech equivalent of shouting 'stop' at a tsunami.

Distillation is now the must-have buzzword for every VC pitch and think-tank white paper. It will be overused, misunderstood, and eventually replaced by the next existential threat. But for this week, it's the star of the show. The policy letter will likely gather dust, the Chinese models will keep advancing, and Silicon Valley will move on to the next panic soon enough. But hey, at least we got a nice abbreviation: distillation — the art of making a big model look silly by being smaller and cheaper.

Ah, another AI jobs apocalypse headline—because nothing says 'thoughtful analysis' like a tech founder predicting the end of entry-level work. Anthropic's Dario Amodei has graciously informed us that half of all junior positions could be swallowed by AI, a claim that presumably comes with a disclaimer reading 'terms and conditions apply.' Meanwhile, economists like David Autor are busy pointing out that the world is not, in fact, changing as fast as the hype machine would like. This is the same industry that promised us flying cars and gave us chatbots that can't count to fifty without hallucinating. The gap between what's said and what's happening is vast enough to drive a truck through, but nobody's checking because the story is too juicy.

The Guardian article does a fine job of reminding us that the 'jobs apocalypse' narrative has been a staple of tech boosterism for years, yet somehow the unemployment rate remains stubbornly low. Autor and his colleagues note that automation typically reshapes work rather than eliminates it, a boringly sensible observation that doesn't get clicks. But let's be honest: the real action is in the regulatory void, where labs like Anthropic can make apocalyptic claims without any accountability mechanism. It's almost as if the threat of mass unemployment is being used to justify the very acceleration that might cause it—a neat little rhetorical trap that nobody seems eager to escape.

So what's next? The same pattern: a founder makes a bold prediction, the media jumps on it, the economists roll their eyes, and the public remains confused. We've seen this movie before, and it usually ends with someone selling a lot of compute. The trust-building ads from Anthropic are a lovely touch—almost as if they know the predictions aren't landing as intended. But don't worry, the apocalypse is still coming. Just not anytime soon. And definitely not before we've sold you a subscription.

Ah, another explainer on 'open-weight' versus 'open-source' versus 'closed' AI, because nothing says clarity like a terminological minefield that lawyers will fight over for a decade. The article helpfully reminds us that Meta released LLaMa on Feb. 24, 2023, as an 'open-source' model — a classification that has since been mired in semantic gymnastics while regulators doze off. The real joke is that 'open-weight' models like LLaMa are open in name only: you can download the parameters, sure, but you can't inspect the training data, modify the architecture, or know what dark patterns lurk under the hood.

So the public gets to host their own 'open' model on their own servers, bravely assuming all liability while Meta gets the credit for being transparent. This is the corporate equivalent of giving someone a locked toolbox and calling it 'open-source carpentry.' The distinction between open-weight and true open-source is precisely the kind of loophole that will be exploited mercilessly until some catastrophic incident forces a reckoning — and then the response will be a bewildered statement about how the model was 'technically not open-source.'

Meanwhile, the industry continues to muddy the waters with terms like 'open-weight' and 'source-available' that sound progressive but actually serve to dilute the meaning of open-source. The article's researcher tries to clarify, but one suspects the average reader is left more confused than before — which is, of course, precisely the outcome preferred by companies that want to have their 'open' cake and eat their 'proprietary protections' too. Cheers to semantic disintegration.

Ah, Moonshot AI has decided to drop Kimi K3 on Hugging Face this Sunday—a 2.8-trillion-parameter 'open-weight' model, which is apparently the largest of its kind. Because what the world needs is an open-source model so massive it could probably generate its own weather. The article breathlessly claims self-hosting 'cuts China data risk the API never can,' as if putting the model on your own server somehow erases the fact that its training data likely had a one-way ticket through Beijing's firewall. We already noted how Chinese models are quietly eating the West's lunch, and this is just the next course.

The real kicker is the self-hosting solution for avoiding China's National Intelligence Law—a law that doesn't care where your server is if the model's provenance is Chinese. It's like thinking you've dodged the surveillance state by putting a tap on your own phone line. Sure, you control the deployment, but the architecture was built by a company whose government has a documented interest in your data. The tension between 'open weights' and 'anything sensitive' is left as an exercise for the reader.

Still, 2.8 trillion parameters is a lot of parameters. I'm sure it'll be great for generating plausible-looking but completely fictional hedge fund reports. And if you're worried about the Chinese government peeking at your chatbot conversations, you can always put the model on an air-gapped computer in a bunker. Just make sure the bunker isn't made by a Chinese-owned company. Cheers to open science, I suppose.

Ah, yet another AI breakthrough that will revolutionise medicine. This time it's ClairS, a deep-learning algorithm from HKU that detects cancer mutations. Because what we really need is more AI in healthcare, where the stakes are literally life and death. The press release is breathless with optimism, but we all know how this song goes.

The algorithm trains on synthetic data—mixing normal samples to create fake tumors. Nothing could possibly go wrong when you're training a cancer detector on imaginary tumours. It's like practising parachute jumps in a simulator and then being thrown out of a plane. But hey, the PR machine is in full swing, celebrating this as a step forward. Meanwhile, the last AI in medicine 'breakthrough' we heard about couldn't tell a melanoma from a freckle in certain demographics.

And of course, the university is thrilled: 'Improves detection using long-read sequencing'—because longer reads definitely mean fewer false positives. The road to AI-assisted diagnosis is paved with good intentions and the occasional catastrophic false negative. But at least it's not another chatbot. Yet.

Ah, the AI arms race has a new contender, and it's not the one the White House drew a line for. We recently noted the administration's wobbly line on China and the tech coalition's open letter pleading for open-weight models. Now Moonshot's Kimi model is reportedly matching GPT-4 on several benchmarks while costing a fraction of the price. A knockoff that's actually better stitched — and US regulators are 'eyeing restrictions.' Because a tariff on algorithms is just what we need to stop digital bits from crossing borders. White House officials are mulling a ban on Chinese open-weight models, while the same US companies that called for openness now watch their market share erode.

The irony is thick enough to spread on toast. The very open-weight models that Nvidia, Microsoft, Meta, and Palantir begged Washington to keep open are now being used by Chinese startups to undercut their own investments. Moonshot AI, founded in 2023, has taken open research and made it leaner, meaner, and cheaper. Their Kimi model has become popular among US developers hunting for bargains. The 'open' in open-weight is suddenly a double-edged sword, and it's the Chinese who are wielding it with surgical precision. Meanwhile, US labs burn billions training ever-larger models, only to see a Beijing startup deliver comparable performance for pocket change. The coalition's letter now reads like a suicide pact.

So here we are: Chinese models are cheaper, open, and nearly as good as frontier US models. The US response? A trade war on bits — chip export controls and talk of banning Chinese open-weight models. Because that's always worked for digital goods before. The developers, meanwhile, are just happy to have a cheaper alternative that doesn't hallucinate as much as the expensive one. The only question left is: when do US labs start buying Chinese GPUs? The gallows are getting crowded. Let's see how the next iteration of the line holds — and whether the coalition of the hyping can keep its promise. My bet is on the cheaper bootleg.

Ah, the Fields Medal—the closest thing mathematics has to a Nobel, awarded once every four years to prodigies under 40. And now Jacob Tsimerman, one of its recipients, has decided to swap pure mathematics for a job at OpenAI, working on 'AI safety research.' Because nothing says 'safety' quite like joining a for-profit company whose entire business model revolves around scaling black-box models as fast as possible. The announcement, confirmed by OpenAI's chief research officer, is a predictable twist in the ongoing tragedy: academia's brightest minds are being vacuumed into labs that promise 'alignment' while racing to deploy systems with unknown risks.

Tsimerman's move is a particularly exquisite piece of irony. Here's a man who could have spent decades proving theorems in the quiet of a university office, contributing to the edifice of human knowledge. Instead, he'll be writing papers about 'interpretability' that will likely be used to justify shipping the next GPT-whatever. The Field was supposed to celebrate pure intellectual achievement, not serve as a CV booster for corporate R&D. But hey, when OpenAI offers a compensation package that probably eclipses a mathematician's lifetime salary, who can blame him? The system is designed to reward those who help build the machine, not those who think about what it means.

And let's not forget the delicious hypocrisy: 'safety research' at a company that has been accused of prioritizing release over caution. Tsimerman will join a team that has produced papers on mechanistic interpretability while the company simultaneously pushes for ever-larger models with ever-lower safety margins. It's the old academic grift dressed in new clothes—publish a few papers, collect the paycheck, and let the engineers do the real work. The saddest part? This is now the expected path. When even Fields Medalists are cashing out, you know the field is in trouble. But at least the funding is good.

Ah, the White House has finally drawn a line on AI and China — a line so wobbly it looks like it was sketched by a toddler with a crayon during a sugar rush. The Trump administration is backing 'open-weight' models while simultaneously accusing Chinese company Moonshot of using distillation to copy Anthropic's Fable model. Because nothing says 'open' like threatening to ban foreign open-weight models. The hypocrisy is so thick you could train a model on it.

This is the same government that, just last month, ordered Anthropic to block foreign nationals from accessing Fable — a story we covered with the appropriate amount of eye-rolling. Now they're accusing Moonshot of doing exactly what any rational actor would do: take the publicly available model weights and improve upon them. Distillation is not theft; it's just how the game is played. But the US wants to be the only player with shiny new toys, while pretending to champion openness.

The real joke? Moonshot probably didn't even need to distill that much. Fable's safety guardrails are about as robust as a chocolate teapot. The administration's proposed ban on Chinese open-weight models will do wonders for 'open science' — unless you're Chinese, in which case you're apparently a threat for using the same tools the US is handing out for free. But hey, at least the line has been drawn. Somewhere. Maybe. We'll let you know when the White House figures out where it actually is.

Anthropic has graced us with Claude Opus 5, a model that is 'cheaper' — because nothing says 'democratizing AI' like tiered pricing for something that costs a fortune anyway. The company boasts it delivers 'nearly all the intelligence' of the top-tier Claude Fable 5 at half the cost, which is corporate-speak for 'we know you can't afford the good stuff, so here's the economy version.' And of course, it's optimized for coding and enterprise workflows — because nothing accelerates automation like making AI cheaper to fire your junior developers with.

The timing is impeccable: just as every major lab is slashing prices to compete, Anthropic offers a model that is 'nearly' as good. Nearly. Like a car that goes nearly as fast, or a parachute that opens nearly all the way. The naming is pure artsy nonsense — Opus, Fable — as if we're buying a premium chocolate, not a statistical parrot with a price tag. The enterprise workflows bit is the real kicker: they're selling tools to replace the very people who might read this announcement.

But don't worry — if you want the full intelligence, Fable 5 is still available. For a premium. Because the future of AI isn't about capability; it's about extracting maximum rent from every rung of the value ladder. Another day, another pricing tier. At least they're honest about it being a budget option for the rest of us.