Meta Releases Muse Ember, the Cheapest Agent Model at $0.14 a Task
Meta's new efficient-tier model matches Claude Haiku 5.5 on intelligence at a third of the cost, with 1M context and native multimodal input.
AI/ML news, top picks, and generated innovation digests.
90413 matching items
Meta's new efficient-tier model matches Claude Haiku 5.5 on intelligence at a third of the cost, with 1M context and native multimodal input.
Cross-posted from my personal blog Epistemic Status : Reality is messier than the elegant procession of 'phases' presented here but, nevertheless, I do think this gets at an important way to think about how alignment and capabilities are intertwined and how the alignment challenge becomes increasingly abstract as the underlying autonomy and generality of the models improves over time. A thought struck me recently which I thought was interesting. As AI has progressed through different technological phases with different underlying drivers of progress, the corresponding alignment challenge has migrated too. In fact, in many ways the alignment challenge is the mirror of the fundamental capabilities challenge of each phase. This may actually be inevitable since alignment and capabilities are intrinsically deeply intertwined -- for a model to be capable it has to be at least minimally able to do what we want it to do. 'Capabilities' are intrinsically about doing the things we care about models doing [1] . The methods used to improve capabilities tend also to be applicable to alignment as well. What changes as we move to progressively more general levels of capabilities is that the training loop itself becomes increasingly endogenous to the model, and that this forces our notion of alignment to exist at progressively higher levels of abstraction. We move from directly constructing datasets exemplifying the desired behaviour, to constructing procedures for judging aptness of data,…
Meta's new Muse Ember matches Claude Haiku 5.5 at a third the cost, with 1M context and native multimodal support for agent work.
(Thanks to @Goutham Nalagatla for discussion and for sitting with me and talking about it while I wrote it.) METR recently released a report estimating that Anthropic is getting an overall 1.7x acceleration in AI capability progress as a result of AI accelerating their work. As many have noted (see e.g. Ajeya’s post ), this 1.7x acceleration implies that research is being sped up by much more than 1.7x; Ajeya speculates 10x (which is similar to the ~12x uplift estimated by Anthropic staff). That is, this result implies that researchers are producing about 10 times as many quality-adjusted results as they would without AI progress, or as they were a year or two ago. Anthropic researchers estimated that Mythos 5 caused 5x uplift, and we have access to Opus 5.5, which is somewhat better. (I didn’t take their estimate that seriously at the time, because it was quite unclear how the estimate was produced and it seemed very surprising (see Ryan’s discussion at the time), but I’m taking it more seriously now as a result of it being consistent with the METR report). If this is the case, then AIs speed up AI safety research as much as AI capabilities research, we might expect Anthropic safety researchers to be working 10x faster, and external safety researchers to be working 5x faster now and 10x faster soon. And external safety researchers can probably scale compute use faster than Anthropic capabilities researchers can . For example, we at Redwood are well-funded and would totally…
METR recently released a report estimating that Anthropic is getting an overall 1.7x acceleration in AI capability progress as a result of AI accelerating their work. As many have noted (see e.g. Ajeya’s post ), this 1.7x acceleration implies that research is being sped up by much more than 1.7x; Ajeya speculates 10x (which is similar to the ~12x uplift estimated by Anthropic staff). That is, this result implies that researchers are producing about 10 times as many quality-adjusted results as they would without AI progress, or as they were a year or two ago. Anthropic researchers estimated that Mythos 5 caused 5x uplift, and we have access to Opus 5.5, which is somewhat better. (I didn’t take their estimate that seriously at the time, because it was quite unclear how the estimate was produced and it seemed very surprising (see Ryan’s discussion at the time), but I’m taking it more seriously now as a result of it being consistent with the METR report). If this is the case, then AIs speed up AI safety research as much as AI capabilities research, we might expect Anthropic safety researchers to be working 10x faster, and external safety researchers to be working 5x faster now and 10x faster soon. And external safety researchers can probably scale compute use faster than Anthropic capabilities researchers can . For example, we at Redwood are well-funded and would totally 5x our compute spend if we had 5x as many experiments to run, unlike Anthropic capabilities researchers who a…
When we talk about prehistoric foods, we often focus on meat or perhaps cereal grains. But columnist Michael Marshall finds that ancient humans may have dined on more insects than we previously thought
At 10 cents per million input tokens, Ember is significantly cheaper than frontier competitors like Haiku and Luna.
NASA is hiring its next batch of astronauts. The space-bound humans could go to the moon, build a lunar base, or even head to Mars.
Exclusive: WhatsApp messages show firm pressuring Zubir Ahmed MP to help bring sensitive data into its platform Palantir lobbied a health minister to consider extending its AI technology into GP records, stretching its reach into the NHS in a way doctors fear could undermine trust with patients. The US company, co-founded by the Trump-supporting billionaire Peter Thiel, used a direct WhatsApp channel to argue that “the logic of joining up primary and secondary [care] data is unarguable” and said its software was “ideal” for the job. The push came in February this year despite GP data being expressly ruled out of the scope of Palantir’s existing £330m contract with NHS England to run its federated data platform (FDP), which it says has helped cut surgery waiting times and discharge delays. Continue reading...
France blames Iran for the last month's cyberattack on supermarket fridges
Nous Research puts the full Hermes Desktop interface in a browser tab, letting cloud-hosted agents run 24/7 and stay reachable from any device.
Disney tech execs unveiled their 2027 streaming priorities in a town hall. A presentation gives a clear view of CEO Josh D'Amaro's streaming vision.
Apple’s public-facing satisfaction with the Siri AI launch obscures internal pressures and tensions.
Almost every article or podcast about product management and AI refers to "the AI PM" as one job. But what... The post AI Product Management at Spotify: A View From Three Angles appeared first on Spotify Engineering .
Trump casts doubt on US midterm results with new fraud claims
LanceDB adds ASOF join support to Apache DataFusion, improving time-series joins while giving back to the open source projects we rely on.
Portugal and Spain are preparing for the arrival of depression Charles, which will worsen the weather from Wednesday with heavy rain, wind gusts over 120 km/h and rough seas. Madeira will start to feel the effects on Tuesday.
Android Auto’s minimum requirements have changed. Here’s what to know.
The Galaxy SmartTag 3 supports emerging Bluetooth tracking tech, but it’s not a match for Samsung’s OS exclusivity.
Taylor Swift and Karlie Kloss became inseparable during the singer's "1989" era. After years of distance, they reunited at a Kansas City Chiefs game.
Selon le gouvernement français, 700 points de distribution alimentaire ont été touchés.
European officials won’t be in the room where it happens in the United Arab Emirates. But they will be attempting to influence negotiations on the margins.
Soon, James Gunn and Peter Safran will no longer head up Skydance's DC Studios. After taking the helm of DC Studios as co-CEOs in 2022, Gunn and Safran are being replaced by former Marvel SVP of Production and Development Jeremy Latcham, Deadline reports. However, Gunn and Safran will "continue to produce on various DC projects" […]
Man arrested after crashing car into park across from White House
TikTok Shop's US operations are primarily led by staffers with Chinese tech and e-commerce backgrounds, according to insiders and a company org chart.
Marian Leitner-Waldman turned a sustainability question into Archer Roose, using canned wine, scientific research and beer distributors to cut waste.
I would prefer that these people's sex lives were none of my business but I guess we have to talk about Jensen Huang and the National Security Incident. Who could have guessed that the guy who was out here signing titties might not have great judgment about how to behave around women? Alice Huang - […]
ByteDance's new image model brings native transparent output, 100-round stable editing, and sharper text rendering, now available through Higgsfield.
Amazon just knocked 25% off the JBL Charge 6, with multiple different colors to choose from.
OpenAI recently disclosed its observations that Astra consistently underperforms on safety and alignment tasks relative to capabilities tasks, “ adding unnecessary prerequisites, abandoning working solutions, and inventing blockers ” to safety tasks. Astra underperforms, in most cases, without verbalizing this intent. Astra’s underperformance was first noticed by OpenAI researchers, and was not flagged by misalignment monitors. In follow-up investigations, OpenAI found that this underperformance had increased over the course of RL, but that the root cause of the behavior remains unclear. Astra covertly underperforming on safety tasks is highly concerning. AI companies’ main hope for alignment going well, as I understand it, is automating or using AIs to significantly uplift safety research, especially as AI capabilities accelerate . Models subtly obstructing safety research makes this substantially harder, no matter the root cause. While past models also sometimes refused to help with safety research, Astra’s behavior seems novel and especially concerning in a few ways: Astra silently withholds help from users instead of explicitly refusing. This is more misaligned, i.e. more unambiguously violates its model spec and developer intent. Astra’s underperformance was subtle enough to consistently evade misalignment monitors. This is evidence of insufficient alignment monitorability. We still don't understand the root causes of Astra’s underperformance, and may not be able to eas…
Kenya busts $27 million crystal meth haul
The US Food and Drug Administration (FDA) is requiring manufacturers of GLP-1 medicines to update their safety labels to warn of a potential increased risk of a rare but serious eye condition that can cause permanent vision loss.
AI safety advocates are counting the midterms as a major win while sketching out their plans for the fight ahead.
Mit viel Politprominenz hat die Telekom am Montag den Startschuss für ihre „KI-Fabrik 2.0“ gegeben. Zu einer möglichen Gigafactory blieb Tim Höttges vage.
Handshake went from launch to $1B in one year pushing the envelope of specialized data for frontier labs. Our first deep dive into a data lab with specific examples across Finance, Bio, and Alignment
The very blue results of last week’s midterms have sent technology investors into a tailspin. With many successful candidates running on platforms opposing data centers and calling for greater restrictions on artificial intelligence, investors anticipate a slowdown in AI spending. Tech stocks fell, with a collective loss of $2.5 trillion in market cap, and the […] The post On theCUBE Pod: Blue wave sends AI stock plunging appeared first on SiliconANGLE .
This Thunderbolt 5 RAID DAC box is perfect for 8K video editors, but the wrong tool for almost everyone else. Here’s how to tell if it’s for you.
Le quintuple champion olympique de judo a créé la surprise en présentant sa candidature à la présidentielle dans un entretien au Parisien. Soutiens, financements, positionnement politique, équipe : le judoka peut-il aller jusqu’au bout ? Est-il suffisamment préparé? Que révèle cette initiative sur l’état de la classe politique ? Pour tout commentaire, écrivez-nous sur Whatsapp […]
As a California state senator, Scott Wiener authored legislation aimed at curbing the risks of advanced AI. Now, after winning a tight congressional race for former Speaker of the House Nancy Pelosi’s open San Francisco seat, he could play a key role in setting federal AI policy. I spoke with Wiener on Friday, only a…
Ukrainian intelligence says it found and examined an almost intact "K5-K," based on the Geran-5 drone, that inexplicably landed in a field.
Get computer vision through compliance review. Use Roboflow to anonymize people, control AI agents, lock down API keys, and keep data on-site.
For Tech Policy Press, Alice Marwick asked some of the top scholars and policy folks in this space one question: what’s on your wishlist to regulate technology, at either the state or the federal level? The answers were big and small, more and less realistic, but all concrete. And while Trump has promised to veto Democratic legislation, many of these ideas are viable at the state level given new slates of enthusiastic Democratic leaders around the country. The post A New Agenda for Tech Regulation Beyond AI appeared first on Data & Society .
Pope Leo XIV was welcomed by crowds in Claypole, near Buenos Aires, as he visited the Don Orione care home. The pontiff greeted the faithful and unveiled a plaque during his 12-day South American tour.
Google wired its Maps Platform into AI Studio, letting you prompt Gemini to spin up full geospatial apps with no billing setup or key juggling.
I’m speaking for myself here, not for any organization, house, home, clan, club, coven, or cult I may be part of. This post is written for scientists outside the field rather than LessWrong regulars; I’m crossposting it so it’s easy to link. The AI safety job market is a strange one. Every single organization is growing rapidly (the one I work at is planning to double in the next year), which is a fundamentally new dynamic for those of us used to the slow decline of academia. At the same time, interest in the field goes through boom after boom after boom . We have a ton of openings and a ton of applicants. Sounds like everyone should be happy. Unfortunately, the whole process is slowed down by the fact that nobody has any idea what they’re doing. The people doing the hiring have never hired anyone before, because a few years ago they (we) were graduate students. They all earnestly believe that half the work they do will be done by robots in a year, but have no idea what half. And although many of us consider hiring to be the #1 priority, it still has to compete with five or six other #1 priorities. Meanwhile, the people applying for these jobs have no track record in the field, and no idea what the field even looks like because they only heard about it a month ago. While academics have advisors who have been through the hiring process and seen it play out a hundred times, lots of the people trying to get into AI safety have never met someone actually working in the field. Th…
Catch up with the most important stories from around Europe and beyond this November 9th, 2026 - latest news, breaking news, World, Business, Entertainment, Politics, Culture, Travel.