 |
|
|
|||||
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. An OpenAI browser could directly target Googles ad empire OpenAI is preparing to launch an AI-powered web browser in the coming weeks, according to a bombshell report published by Reuters on Wednesday. The move should be seen as a direct challenge to Googles financial heart: its $200 billion interactive advertising machine. More than 400 million people already use OpenAIs ChatGPT, and many of them routinely rely on it as their interface to the web. Currently, when users click a source link within a chatbot search result, the target website opens in a third-party browser like Chrome. With its own browser, ChatGPT would launch source citations and other links into a platform under its own control. An OpenAI browser would also encourage users to fetch web information using the companys AI models and web index. It could become the ideal platform for AI agents capable of more than search, handling tasks like booking reservations, filling out forms, and interacting directly with websites. Such a browser could eventually pose a serious challenge to Googles core business: selling ads around its search results. Chrome makes it very easy for users to search the web with Google Search, and many of those searches (especially for products and services) offer lucrative ad opportunities. Chrome also supplies Google with mountains of user browsing behavior data, which it leverages to target ads more effectively. Generative AI represents a radically different method of retrieving web content. While Google Search relies on web crawlers and a complex algorithm to return relevant content and links, generative AI models are trained on a compressed version of the internet, allowing them to provide customized packages of information that directly answer the users query. Whether 400 million ChatGPT users will abandon their current browsers for OpenAIs AI-first experience is the billion-dollar question. Googles stock barely moved on news of OpenAIs browser. Either investors dont yet fully grasp the magnitude of the threat, or theyve already factored it into Googles share price. For OpenAI, a browser war with Google would be a hard fight for OpenAI. Over 3 billion people around the world use Chrome. But the timing may be right for OpenAI to make a move. The Department of Justice is demanding that Google divest Chrome, following a federal judges ruling that the browser contributes to an unlawful search monopoly. OpenAI has even testified that it would be interested in buying Chrome if Google is forced to sell. OpenAI may be the only AI company with the momentum to challenge Google right now. Perplexity, which released its own Comet browser on Wednesday, also aims to take on Google Search with AIbut as a startup, it may lack the resources to mount a serious challenge. Perplexity vs. Google is a David-and-Goliath story. But OpenAI vs. Google may prove to be the biggest tech battle of the first half of the 21st centurya clash between an aging heavyweight and a young challenger with a powerful new punching style. AI is spoofing within the highest levels of power Somebody used an AI voice generation tool to clone the voice of Secretary of State Marco Rubio and then used Rubios voice and writing style to send fake messages to several foreign ministers, a U.S. governor, and one member of Congress. The provocateur sent the messages via the encrypted voice and text messaging platform Signal, using an account named Marco.Rubio@state.gov, The Washington Post reported, citing a State Department cable it had obtained. The cable, dated July 3, stated the impostor left voicemails on Signal for at least two targeted individuals and sent text messages inviting others to communicate on the platform. Creating such cloned voice messages is relatively easy. All an impostor needs is a 15-second sample of the targets voice, which is readily available for public figures like Rubio. The sample is then uploaded to an AI voice generation tool such as ElevenLabs or Hume. (The service used in this case is unknown.) After that, the user declares they have permission to use the sampled voice and types in what they want communicated. These tools can even reproduce the emotional tone of the sampled voice (although theres no indication that was done here). The identity of the person who created the recordings remains unknown, and neither the recordings nor their content has been made public. According to the State Department cable, the person appeared to be trying to gain access to the information or accounts of powerful government officials. This kind of incident isnt exactly new. In May, someone breached White House Chief of Staff Susie Wiless phone and impersonated her in calls to senators and business executives. The FBI has since warned of an ongoing malicious text and voice messaging campaign using AI-generated voices to target senior government leaders. We should expect this to happen again and again in the future, because a considerable gap still exists between the capabilities of current AI voice generation tools and the common understanding of their level of sophistication. Ramp data: Did enterprises just hit peak subscription-AI spend? Ramp AI, which tracks enterprise procurement on its platform, says U.S. business spending on AI subscriptions dipped by half a percent last month. This follows steady growth in subscription spending throughout 2023 and 2024, and an even sharper rise during the first five months of 2025. The share of U.S. businesses with paid subscriptions to AI models, platforms, and tools fell from 42.5% in May to 42% in June, the Ramp data shows. The share of U.S. companies subscribing to such products from OpenAI shrank slightly from 34.7% to 34.2%, while 9.9% of companies accessed them from Anthropic, up from 9.3% in May. The spending decreases were driven by companies in the tech and finance sectors. Seventy percent of U.S. tech businesses bought access to AI models, platforms, and tools in June, down from 71% in May. Among finance companies, the percentage declined from 57% to 56%. Slightly more healthcare and manufacturing companies bought access to AI during June, but in lower numbers. Ramp economist Ara Kharazian said on X that the spending dip may be the beginnings of a signal that enterprises are suffering from AI pricing fatigue. He clarifies to Fast Company that his firm doesnt believe tht companies are using less AI, but they may be looking for ways to economize. He says the decrease was driven by companies canceling chat subscriptions or switching to free versions. (Google recently started integrating Gemini Pro for free in all workspace plans, for example.) And the profile of the typical AI subscriber is changing. We are getting out of the early adopter phase, Kharazian says. The metrics are being moved not by the early adopters but the mainstream adopters: companies who dont want or need the most advanced versions but still want to enable employees with a low-cost, still valuable AI tool. Thats the core user now. Ramps AI Index samples more than 30,000 American businesses and billions of dollars in corporate spend using data from Ramps corporate card and bill pay platform. Within its methodology statement, Ramp disclaims that its results may underestimate actual adoption rates due to the prevalence of businesses using free AI tools (which it cant track), or when employees use their personal AI accounts for work tasks. Meanwhile, investors keep pouring big chunks of cash into new AI labs and startups. During the second quarter, 45% of global fundingor $40 billionwent to AI sector companies, reports Crunchbase. Notably, Mira Muratis Thinking Machines Lab and Ilya Sutskevers Safe Superintelligenceboth a year old or lesseach raised $2 billion. Other large investments went to Anduril Industries ($2.5 billion), Grammarly ($1 billion), and Anysphere ($900 million). More AI coverage from Fast Company: Moonvalley releases its power tool for AI filmmaking Racist AI-generated videos are all over TikTok, thanks in part to Googles Veo 3 tool Why the new rulings on AI copyright might actually be good news for publishers These personality types are most likely to cheat using AI Want exclusive reporting and trend analysis on technology, business innovation, future of work, and design? Sign up for Fast Company Premium.
Category:
E-Commerce
Britain’s WPP named board member Cindy Rose as its new chief executive on Thursday, tasking the senior Microsoft executive with leading its recovery a day after a major profit warning showed the scale of the challenge at the ad group. Rose has been on the board since 2019 and will take over from outgoing CEO Mark Read on Sept. 1, the company said, with Read departing four months earlier than expected. Shares in the group, which fell to a 16-year low on Wednesday after it slashed its profit outlook, rose 2% in early deals. Rose has spent the last nine years in senior leadership positions at Microsoft, where she rose to Chief Operating Officer, Global Enterprise after formerly being head of its UK business. Prior to that she worked at Vodafone and Virgin Media. WPP Chairman Philip Jansen said she had supported the digital transformation of large enterprises around the world including embracing artificial intelligence to create new business models and revenue streams. “Her expertise in this landscape will be hugely valuable to WPP as the industry navigates fundamental changes and macroeconomic uncertainty,” he said. Rose will take on a business reeling from the loss of some big accounts, a downturn in client spend, and fewer new business opportunities, which prompted the profit outlook cut that sent WPP shares down as much as 19% on Wednesday. WPP, which lost its crown as the world’s biggest ad group to France’s Publicis last year, is also grappling with the transformational impact of AI, which gives clients the tools to create and manage more of their own marketing campaigns. Rose said: “We have and continue to build market-leading AI capabilities, alongside an unrivalled reputation for creative excellence and a preeminent client list.” Sarah Young, Reuters
Category:
E-Commerce
Minh Pham and JJ Ford have a knack for riding the waves of new tech. The duo joined Uber in its early days, helping to spearhead mobile development. When CEO Travis Kalanick was ousted in 2017, Pham and Ford followed him out. Kalanick soon founded CloudKitchensand Ford was his first call. Fords first call? Pham. After five years at CloudKitchens, they were ready for the next frontier. Over the years, theyd built a battle-tested engineering team that successfully tackled both the mobile surge at Uber and the infrastructure demands at CloudKitchens. They also had a hunch: While many startups were focused on building single-use AI agents, Pham and Ford saw the future in orchestrating many agents to work together. In 2023, they founded the California-based company Invisible. How do you build a system that allows one person to manage thousands of agents? Ford asks, noting that he and Pham used their enterprise experience to develop a solution thats different from any theyve seen in the market previously. Inside Invisible Most AI agents today handle single tasksordering food, booking travelwithout human input. Pham compares them to task assistants. But Invisibles product stacks these agents, enabling them to tackle multiple tasks in parallel. We want to turn AI agents into an elastic, reliable, and scalable workforce for a company, Pham says. Most companies think of using a single agent from Anthropic or OpenAI. We think of it as an orchestration platform for multiple agents, acting like how you have a team of humans doing work. From left: Justin Takamine (CTO), Minh Pham (CEO), Hanna Dang (COO), and JJ Ford (CSO). [Photo: Invisible] Invisibles system is hierarchical. A top-level agent breaks down a task, delegates subtasks to lower-level agents, and coordinates the process. What we found is, when you have a hierarchy of agents like that, the sum is much larger than what a single agent could do, Pham says. Three months ago, Invisible launched A3, short for Action Agent API. While they declined to share revenue, Pham and Ford confirmed a $7 million fundraising round last year. Their customers, Pham says, include large-scale companies like mortgage brokers and solar providers. “Were already running pilots and engaging in active discussions with dozens of Fortune-scale enterprises eager to deploy A3 and streamline their most critical workflows,” Minh says. Stacking a workforce A3 operates on a companys existing standard operating procedures. Uber, for instance, has hundreds of thousands of playbooks. Once A3 is trained on these, a company can scale its agent workforce up or down as needed. Pham cites the example of Know Your Customer checksan expensive compliance requirement in fintech. The work involves compiling detailed reports to verify legitimacy and creditworthiness, often through manual research. We were able to go in and completely automate this process within days of working with the company, Pham says. Invisible is priced like a workforce, too. Unlike competitors that use a subscription model, Invisible charges 10 cents per action. According to Pham, that usually adds up to about $12 per hourroughly the cost of a human workerand could decrease as AI costs drop. We want to think of it as a workforce, and we have to come up with a whole new pricing model to match that, Pham says. Ford adds that per-action pricing makes it easy for companies to easily come online, stand up a workforce, make it elastic, and then pull it down. The implications of an AI agent workforce are far-reachingand potentially disruptive. Still, the founders emphasize the upside. A3, Ford says, allows employees to climb the value chain of what theyre delivering to the company. For better or worse, the agents are coming. And Invisible is betting theyll be most powerful when stacked.
Category:
E-Commerce
All news |
||||||||||||||||||
|
||||||||||||||||||