Tech

Dimitri Masin, CEO of Gradient Labs, argues that companies using AI agents for customer support should only pay when the bot does its job.

"If you look at Salesforce, they price the automation per conversation," he told The Register in a phone interview. "So essentially, if you have a conversation with AI, no matter what it leads to, you pay $2."

According to Salesforce's own researchers, leading LLM-based agents tested on the CRMArena-Pro benchmark successfully complete single-turn (prompt and reply) tasks about 58 percent of the time and only about 35 percent of the time for multi-turn (back-and-forth conversation) requests.

That's both bad for customers and bad for the progress of AI agents overall. Paying regardless of results, said Masin, "doesn't create any incentive for Salesforce to actually make their agent better."

Salesforce did not respond to a request for comment. But the stats it touts in its marketing copy are, unsurprisingly, a lot better.

Europe’s ambition for digital sovereignty is more urgent than ever. But instead of backing truly European, privacy-first services, many European businesses and even EU and local authorities still rely on US tech, and might consider the brand-new and shiny "sovereign cloud" offerings as a useful solution.

But what if the "sovereign cloud" is less secure than US companies want us to believe? Let's take a deep dive into the "Microsoft Sovereign Cloud", the "Sovereign Cloud from Google", and the advertised "European Digital Sovereignty from Amazon Web Services" and whether it's safe to use these as a European authority or business.

Microsoft EVP Yusuf Mehdi said in a blog post last week that Windows powers over a billion active devices globally. This might sound like a healthy number, but according to ZDNET, the Microsoft annual report for 2022 said that more than 1.4 billion devices were running Windows 10 or 11. Given that these documents contain material information and have allegedly been pored over by the tech giant’s lawyers, we can safely assume that Windows’ user base has been quietly shrinking in the past three years, shedding around 400 million users.

This is probably why Microsoft has been aggressively pushing users to upgrade to Windows 11 after the previous version of the OS loses support — so that its users would install the latest version of Windows on their current system (or get a new PC if their system is incapable of running the latest version). Although macOS is a threat to Windows, especially with the launch of Apple Silicon, we cannot say that those 400 million users all went and bought a MacBook. That’s because, as far back as 2023, Mac sales have also been dropping, with Statista reporting the computer line, once holding more than 85% of the company revenue, now making up just 7.7%.

Ever since Microsoft announced that it would end support for Windows 10 in October, the company has been trying hard to convince users to make the switch to Windows 11. First, it warned that unsupported Windows 10 PCs will no longer receive security updates, making them easy targets for hackers. Later, it advised users to trade in their old computers and buy a new one that comes preloaded with all the Windows 11 goodies.

Now, once again, Microsoft’s Executive Vice President and Consumer Chief Marketing Officer, Yusuf Mehdi, has published a fresh blog highlighting all the benefits and advantages of Windows 11, including a statement claiming that Windows 11 PCs are up to 2.3 times faster than Windows 10 PCs. However, what they failed to make clear is that this claim is entirely based on a comparison of new versus old hardware, rather than the software itself.

A sharply argued blog post warns that heavy reliance on Microsoft poses serious strategic risks for organizations – a viewpoint unlikely to win favor with Redmond or its millions of corporate customers.

Czech developer and pen-tester Miloslav Homer has an interesting take on reducing an organization's exposure to security risks. In an article headlined "Microsoft dependency has risks," he extends the now familiar arguments in favor of improving digital sovereignty, and reducing dependence on American cloud services.

The argument is quite long but closely reasoned. We recommend resisting the knee-jerk reaction of "don't be ridiculous" and closing the tab, but reading his article and giving it serious consideration. He backs up his argument with plentiful links and references, and it's gratifying to see several stories from The Register among them, including one from the FOSS desk.

He discusses incidents such as Microsoft allegedly blocking the email account of International Criminal Court Chief Prosecutor Karim Khan, one of several incidents that caused widespread concern. The Windows maker has denied it was responsible for Khan's blocked account. Homer also considers the chances of US President Donald Trump getting a third term, as Franklin Roosevelt did, the lucrative US government contracts with software and services vendors, and such companies' apparent nervousness about upsetting the volatile leader.

Elon Musk's xAI made quite a splash when it built a data center with 200,000 GPUs that consumes approximately 250 MW of power. However, it appears that OpenAI has an even larger data center in Texas, which consumes 300 MW and houses hundreds of thousands of AI GPUs, details of which were not disclosed. Furthermore, the company is expanding the site, and by mid-2026, it aims to reach a gigawatt scale, according to SemiAnalysis. Such gargantuan AI clusters are creating challenges for power companies not only in power generation but also in power grid safety.

OpenAI appears to operate what is described as the world's largest single data center building, with an IT load capacity of around 300 MW and a maximum power capacity of approximately 500 MW. This facility includes 210 air-cooled substations and a massive on-site electrical substation, which further highlights its immense scale. A second identical building is already under construction on the same site as of January 2025. When completed, this expansion will bring the total capacity of the campus to around a gigawatt, a record.

Archive : https://archive.ph/7xCP1

The Redditor was experiencing problems with his Asus TUF Gaming Radeon RX 9070 XT OC Edition graphics card, leading him to return it to Micro Center for a replacement. Although the replacement GPU's packaging is labeled for the Asus TUF Gaming Radeon RX 9070 XT OC Edition, the actual card inside features a shroud marked with both AMD Radeon and GeForce RTX branding. This unusual hardware issue appears to be a manufacturing mistake by Asus, directly stemming from the company's production line.

FFS, let it die!

As artificial intelligence becomes the corporate buzzword du jour, executives are finding more and more ways to shoehorn the trendy tech into their everyday business operations.

That has a lot of workers anxious about automation, income inequality, and increased workloads — something c-suite bigwigs are all too happy to take advantage of.

Though AI — really just a fun name for large language models (LLMs), or predictive chatbots — in its current state isn't likely to bring a labor revolution anytime soon, CEOs find that the threat of AI automation works just as well.

Archive : https://archive.ph/KJ1tN

Customers dismayed by Broadcom's move to selling costly bundles such as VMware Cloud Foundation (VCF) will realize its value if they'd just use more of the components, the company's CTO says.

VMware, now a Broadcom subsidiary, is shifting away from selling perpetual licenses for individual products. It instead offers subscription bundles of software and support, such as its flagship VCF private cloud platform – version 9 of which was released this week.

The largest enterprise users seem content with this. Broadcom chief Hock Tan told investors this month that 87 percent of VMware's top 10,000 customers have signed up for VCF.

However some smaller and middle sized customers reacted negatively to the licensing changes, claiming their costs have increased by eight to 15 times since the Broadcom acquisition, and there are many stories of firms planning to migrate their workloads from VMware to an alternative platform in future because of this.

"A lot of those stories around cost don't play out when we actually get to sit down with the customer and talk to them about their situation, what they need, and what we're going to do with them," said Broadcom's EMEA chief technology officer, Joe Baguley.

"Initially people might go 'all the prices have gone up,' but those 87 percent of people that have renewed with us have renewed because they've chosen VCF as their strategy going forward," he claimed.

The fear that generative AI tools such as ChatGPT would lead to a generation of students cheating and plagiarizing work has come to pass. The situation is so bad that educators are now looking at multipe ways to stop the problem, or at least make the practice much more difficult. Ironically, one of them is to use AI.

The European Commission is reportedly in talks to move its cloud services away from Microsoft Azure, according to Euractiv. Per the news outlet, the information came via three senior sources familiar with the matter. Apparently, France-based OVHcloud is the front-runner in these discussions. While other European cloud service providers like IONOS, Scaleway, and Aruba are also being considered.

The UK’s public broadcaster, BBC, has written a letter to Perplexity, the AI search startup, asking it to stop scraping articles from its websites, delete existing copies of content, and propose some sort of financial compensation if it would like to carry on scraping data. If the demands are not met, BBC may seek an injunction against the startup citing alleged misuse of its intellectual property.

BBC is probably responding in this way because it has seen other news organizations cement deals with firms like OpenAI and Mistral. The income stream allows news organizations to raise more funds and also cover the costs of the extra load on their servers caused by AI scraping.

In a statement to the Financial Times, Perplexity labeled the BBC’s claims as "manipulative and opportunistic". The startup accused the broadcaster of having “a fundamental misunderstanding of technology, the internet and intellectual property law.”

This is not the first time Perplexity has had a run-in with the media. Forbes and Wired accused it of plagiarizing content from their websites and The New York Times sent the company a cease and desist notice to stop using its content for AI purposes.

A woman severely hurt in a bicycle crash with a Waymo robotaxi is suing the company, claiming one of its vehicles pulled over in a no-stopping zone next to a bike lane, and a passenger opened a door into her path — despite the car’s “Safe Exit” system touted by the Mountain View company as protection for passing cyclists.

Waymo in online marketing materials says its robotaxi Safe Exit sensor and warning systems provide departing passengers with “explicit audio and visual alerts that inform them when a cyclist or other road user is approaching as they exit the car.” The company cites San Francisco’s transit agency in noting that collisions between cyclists and vehicle doors — incidents known as “doorings” — are the city’s second most common collisions causing death or injury.

The passengers from the robotaxi whose door hit Hanke said at the scene that no alert had been given before one of them opened the door into the bike lane, Hanke said. The lawsuit alleged “a malfunction, failure to engage, or design flaw” in the alert system.