 |
|
|
|||||
Welcome to AI Decoded, Fast Companys weekly newsletter that breaks down the most important news in the world of AI. You can sign up to receive this newsletter every week here. OpenAI says it will release an open-source modelbut why now? OpenAI CEO Sam Altman said Monday that his company intends to release a powerful new open-weight language model with reasoning in the next few months. That would mark a major shift for a company that has kept its models proprietary and secret since 2019. The announcement wasnt a total surprise: After the groundbreaking Chinese open-source model DeepSeek-R1 showed up in January, Altman said during a Reddit AMA that he realized his company was on the wrong side of history and suggested an OpenAI open-source model was a real possibility. Open models typically come with a permissive license that requires little or no payment to the model developer. Open-weight models can be more cost-effective for corporations trying to leverage AI since they allow businesses to host (and secure) the models themselvesavoiding the often risky prospect of sending proprietary data through an API to a third-party provider and paying fees to do it. More businesses are moving in this directionespecially those holding sensitive user data in regulated industries. The catch: A corporate user doesnt have to pay to use the open model. Some AI labs release open models to gain credibility in the marketpotentially paving the way to eventually sell API access to their more powerful closed models. By releasing open models early on, the French AI company Mistral established itself as a top-tier AI lab and a legitimate alternative to U.S. players. Some AI labs release open-source models, then earn consulting fees by helping large enterprises deploy and optimize the models over time. Metas Llama models are the most widely deployed open modelsthough the company restricts reuse and redistribution and keeps the training data and code secret, meaning they are not by definition open source. Meta had different reasons for giving away its models. Unlike Mistral and others, it makes money by surveilling users and targeting adsnot by renting out AI models. Zuckerberg continues funding Llama research because the models are a disruptive force in the industry and earn Meta the right to be called an AI company. OpenAI now has its own reasons for releasing an open-weight model. Eighteen months ago, OpenAI was the undisputed champion of state-of-the-art AI models. But in the time since, the release of LLMs like Googles formidable Gemini 2.0 and DeepSeeks open-source R1 have cracked the competition wide open. The market has changed, and OpenAI itself has evolved. Like Meta, OpenAI doesnt depend directly and solely on its models for its revenue. Selling access to its models via an API is no longer the companys main source of revenue. Now, most of its revenue, not to mention its staggering $300 billion valuation, comes from selling subscriptions to ChatGPT (most of them to individual consumers). OpenAIs real superpower is being a household-name consumer AI brand. OpenAI will definitely continue pouring massive resources into developing ever-better models, but its main reason for doing so isnt to collect rent from developers for direct access to them, but rather to continue making ChatGPT smarter for consumers. AI video generation is getting scary good AI-video-generation tools are rapidly leaping over the uncanny valley, making it increasingly difficult for everyday internet users to distinguish between real and generated video. This could bode well for smaller companies looking to produce glossy, creative, or ambitious ads at a fraction of the normal cost. But it could spell bad news if bad actors use the technology in phishing scams or to spread disinformation. Its also yet another threat to the film sectors livelihood. The issue is back in the spotlight following several announcements, starting with Runways release of its new Gen-4 video-generation system, which the company says produces production ready video. AI startup Runway says the new system of models understands much of the worlds physics (a claim supported by this video of a man being overtaken by an ocean wave). The company also touts improvements in video consistency and realism, as well as user control during the generation process. Runway posted a demo video of Gen-4s control tools, which makes the production process look pretty easy, even for non-technicals). Some of the samples of finished videos posted on X look somehow more real than real (see Jean Baudrillard, Simulacra and Simulation). Runway faces some stiff competition in the AI video space in the form of perennial contenders including Googles Veo 2 model, OpenAIs Sora, Adobe Firefly, Pika, and Kling. A new math benchmark aims to beat test question contamination People in the AI community have been debating for some time whether our current methods of testing models math skills are broken. The concern is that while existing math benchmarks contain some very hard problems, those problems (and their solutions) tend to get published online pretty quickly. This of course makes the problem-solution sets fair game for AI companies sweeping up training data for their next models. The worry is that, come evaluation time, the models may have already encountered the test problems and answers in their training data. A new benchmark called MathArena was designed to eliminate those issues. MathArena takes its math problems from very recent math competitions and Olympiads, which have obvious incentives to keep their problems secret. The researchers from MathArena also created their own standard method of administering the evaluation, meaning the AI model developers cant give their own models an edge via changes to the evaluation setup. MathArena has just released the results of the most recent benchmark, which includes questions from the 2025 USA Math Olympiad. Heres one of the questions: Let H be the orthocenter of the acute triangle ABC, let F be the foot of the altitude from C to AB, and let P be the reflection of H across BC. Suppose that the circumcircle of triangle AFP intersects line BC at two distinct points, X and Y. Prove that C is the midpoint of XY. Ouch. And to make matters worse, the test requires not only the correct answer but a decription of each reasoning step the model took along the way. The results are, well, ugly. Some of the most powerful and celebrated models in the world took the test, and none scored above 5%. The top score went to DeepSeeks R1 model, which earned a 4.76%. Googles Gemini 2.0 Flash Thinking model scored 4.17%. Anthropics Claude 3.7 Sonnet (Thinking) scored 3.65%. OpenAIs most recent thinking model, o3 mini, scored 2.08%. The results suggest one of several possibilities: Maybe MathArena contains far harder questions than other benchmarks, or LLMs arent great at explaining their reasoning steps, or earlier math benchmark scores are questionable because the LLMs had already seen the answers. Looks like LLMs still have some homework to do. More AI coverage from Fast Company: An AI watchdog accused OpenAI of using copyrighted books without permission Amazon unveils Nova Act, an AI agent that can shop for you What is AI thinking? Anthropic researchers are starting to figure it out How Hebbia is building AI for in-depth research Want exclusive reporting and trend analysis on technology, business innovation, future of work, and design? Sign up for Fast Company Premium.
Category:
E-Commerce
Termination notices sent by billionaire Elon Musk’s cost-cutting team to U.S. Agency for International Development staff were so rife with errors that corrected versions are being issued to avoid affecting pensions and pay, according to five sources familiar with the issue. The Department of Government Efficiency “did this so quickly that they screwed lots of stuff up,” said a U.S. official, who requested anonymity, as did all of those who spoke to Reuters. The State Department, which is assuming some of USAID’s functions under the Trump administration’s plan to cut U.S. foreign aid, did not immediately respond to requests for comment. USAID’s human resources staff, most of whom have been on paid administrative leave and face termination, have been brought back to the office to send out accurate notices, said the U.S. official and a person familiar with the matter. “My letter was completely wrong,” one USAID worker told Reuters on condition of anonymity. “The only thing correct was my name.” It is not the first time that inaccurate termination notices have upended the lives of USAID workers since U.S. President Donald Trump and Musk began in February to dismember America’s main conduit of foreign aid. A first round set April 21 as the final employment day for most personnel and May 30 for those tapped to help shutter the agency. Those dates were reset to July 1 or September 2 in the notices sent to some 3,500 USAID workers last Friday, two sources and workers said. Other errors included inaccurate start dates, lengths of service and salaries, according to the person familiar with the matter, the U.S. official, two former senior USAID officials, a congressional aide and four workers who received notices. Unless fixed, those mistakes could result in reduced or canceled pensions or inaccurate severance pay, the sources said. Several of the sources pointed to the U.S. Office of Personnel Management’s retirement website that says federal workers’ annual pension annuity is based on their lengths of service and three highest average annual salaries. Reuters could not learn how many USAID personnel were issued faulty notices last Friday. SOME STAFF RECEIVED THREE INACCURATE NOTICES Several workers told Reuters that they and other colleagues received a third termination letter on Monday night still containing inaccurate information on promotions, tenure and other data. One worker said the total federal service listed in their notice on Friday was short by three years and by six years in the notice they received on Monday. “I actually have federal service dating to June 2008,” said the worker. “There doesn’t seem to be any logic to the RIF (reduction in force) process.” “We’ve got people who have served for 25 years and their notices are showing they served for only three,” said the U.S. official. “It affects their severance. It affects their future ability to retire.” Trump assigned Musk, a major contributor to his 2024 election campaign whose companies have federal contracts worth billions of dollars, and DOGE to ferret out waste and fraud across the U.S. government. According to its website, the only official window into its operations, DOGE estimates it has saved U.S. taxpayers $140 billion as of April 2 through a series of actions including massive workforce cuts, asset sales, and contract cancellations. Its savings total is unverifiable and its calculations have contained errors and corrections. Musk has said DOGE will correct mistakes when it finds them. Since February, most USAID staff have been put on administrative leave, hundreds of contractors were fired and more than 5,000 programs terminated, disrupting global humanitarian aid operations on which millions depend. Some termination notices sent on Friday to USAID personnel did not account for requests to waive the July 1 termination date, including from overseas staff whose children still would be in school, according to three sources. Others had applied for waivers because they need more time to pack their homes and relocate to the U.S., the sources said. “Some people have the wrong dates. Others have the wrong information,” said the person with knowledge of the matter, adding that people given the wrong termination date “can’t return home” unless their notices are reissued with the correct date. The person said that the error-filled notices were sent under the supervision of USAID acting administrators Jeremy Lewin, a DOGE operative, and Kenneth Jackson, who have been overseeing the agency’s dismantlement. They report to Secretary of State Marco Rubio, who Trump tapped as acting USAID administrator. Jonathan Landay and Patricia Zengerle, Reuters
Category:
E-Commerce
One of the world’s most iconic and controversial maps just got a major redesign. The Metropolitan Transportation Authority (MTA) in New York has unveiled the final version of an updated map of its subway system, marking the first time the map has had a full redesign since 1979. It’s a visually bold, user-centric design that, according to the MTA, will make it easier for people to understand where they’re going and how to use the system. The new maps are expected to be installed in train cars and stations over the next few weeks. The map features bright, color-coded lines for each train line, which criss-cross a stylized map of the city in horizontal, vertical, and diagonal orientations. More abstract than the previous geographically representative map, the new map prioritizes visual clarity and accessible design over pure accuracy. With single-lined black text on a largely white background and black dots representing stations on bright colored route lines, the new map was designed to be easily read by people with varying levels of vision and color perception. Our approach was to make this map inclusive to all, said MTA chief customer officer Shanifah Rieara at a recent press conference unveiling the new design. [Image: MTA] A big part of the inclusivity is managed by simplifying the geography of the map, using abstracted forms to represent the boroughs and straight lines to represent subway routes that are in fact much more sinuous. It’s an approach that was unveiled in the now-famous 1972 subway map designed by Massimo Vignelli and the design firm Unimark International. It was a minimalist design that became a source of controversy, and one literal debate. In 1978, Vignelli was pitted on stage against John Tauranac, then chair of the MTA’s Subway Map committee, who wanted the system to have a more geographically representative map. Tauranac’s approach won out, and the so-called spaghetti version of the map with winding routes and geographically accurate depictions became the map that has been used from 1979 until now. Though the printed map is being put into service as of this week, this design was first piloted back in 2021, and builds on Work & Co’s live, interactive digital map of the system that has a similar Vignelli-inspired aesthetic. When the pilot design was first launched, an MTA official told Fast Company a final version of the map was expected within months. Four years later, the printed maps are finished. Part of the long gestation has to do with the way the MTA vetted the design, conducting rider surveys to learn more about how people use the map, and the ways some maps make using the system more difficult. Based on this feedback, the map’s design evolved. [Photo: Marc A. Hermann/MTA] The biggest changes relate to some of the most challenging parts of riding a complicated, multi-lined subway system: the transfer. Steven Flamm, manager of mapping for MTA’s Creative Services department, says the map’s design was tweaked to improve the way the map visually explains how to transfer train lines, whether on the other side of a platform, through a tunnel, or across a street. You’ll see a different treatment for hubs and complexes that make it more obvious, so people know they can get their trains in that station, says Flamm. The MTA sees the new map as a mix of the Vignelli design’s minimalist simplicity and a more geographically accurate approach from the Tauranac version that helps people to navigate the system more easily. Design-minded riders may see more of the Vignelli in this new map, but that doesn’t mean the Tauranac version in use for the last four decades has disappeared, according to MTA chair and CEO Janno Lieber. The real superfans out there will recognize the colors that were established in the famous Tauranac map, he said.
Category:
E-Commerce
All news |
||||||||||||||||||
|
||||||||||||||||||