Author: TechBrunch
According to The Information, OpenAI has recently been looking into building its own humanoid robots. The report names “two people with direct knowledge” of these conversations. ChatGPT manufacturers have been involved in this field for some time with financial support. The company has previously invested in Figure and 1X, along with “general purpose AI” company Physical Intelligence. In 2021, OpenAI abandoned such ambitions after quietly shutting down its robotics division. Of course, a lot has happened in the last three years, with breakthroughs in both the hardware and the AI systems that power it. Barring some big-ticket startup acquisitions, a…
Is it possible to train an AI using only data generated by another AI? It may sound like a crazy idea. However, this has been around for quite some time and is gaining traction as new real data is becoming increasingly difficult to obtain. Anthropic used some synthetic data to train one of its flagship models, Claude 3.5 Sonnet. Meta used AI-generated data to fine-tune the Llama 3.1 model. And OpenAI is said to be sourcing synthetic training data from o1, its “inference” model, for the upcoming Orion. But why does AI need data in the first place? And what…
Samsung's CES reporter is always a weirdo. Korean electronics giants generally maintain a dry attitude when it comes to home appliances. After all, the company is expected to announce its latest flagship device, the Galaxy S25, towards the end of January. CES 2025 continues the company's television and consumer electronics heritage. There are also plenty of possibilities, including consumer robots that will probably never see the light of day. Samsung has adopted the tagline “AI for All: Everyday, Everywhere” for its presentation, which begins on January 6th at 2pm PT/5pm ET. That doesn't give us much to work on. After…
Contractors working on improving Google's Gemini AI are comparing its answers to the output produced by Anthropic's competing model, Claude, according to internal communications seen by TechCrunch. When asked for comment by TechCrunch, Google declined to say whether it had permission to use Claude in its tests against Gemini. As technology companies compete to build better AI models, the performance of these models is often measured against their competitors, typically by having contractors carefully evaluate their competitors' AI responses. rather than by running proprietary models through industry benchmarks. Contractors working on Gemini, tasked with evaluating the accuracy of the model's…
In an era when raising venture capital has become more difficult, many companies are turning to non-dilutive revenue-based financing (RBF) as an alternative. As a result, many startups sprung up to meet the challenge, especially in the MENA region, which adopted the model like a duck to water. The latest is CredibleX, which raised $55 million in a seed financing round led by Abu Dhabi-based Further Ventures to capitalize on the burgeoning market for revenue financing in the Middle East. CredibleX, which provides working capital loans to small and medium-sized enterprises, operates in a market where MENA competitors such as…
Clop, a prolific ransomware gang, is credited with stealing data from at least 66 companies by exploiting a bug in a widely used corporate file transfer tool made by Cleo Software. The cybercriminal organization on Tuesday listed some of the names of the companies it hacked on a leaked dark web site, which TechCrunch saw, but the companies did not respond to the organization's efforts to reach out to them. The organization said it would soon release the official name of the hacked company, presumably in an attempt to force the hackers to pay a ransom to prevent victims from…
Since 2018, my colleagues and I have published year-end lists highlighting the best cybersecurity stories reported by other news organizations, first on VICE Motherboard and now on TechCrunch. Cybersecurity, surveillance, and privacy are big topics that no single publication can effectively cover. Journalism is an inherently competitive field, but it's also a very collaborative one. That is why it may make sense to refer readers to other publications and their works to learn more about these complex and vast beats. Without further ado, here are our favorite cybersecurity stories of the year written by our friends at Rival Retailer. —…
People with chronic kidney disease or at risk for heart failure are greatly affected by potassium imbalances in the body. These can even be life-threatening. While wearable blood glucose monitors are now commonplace and have revolutionized the lives of people with diabetes, potassium monitoring is still in its infancy because it is difficult to implement. Start-ups are now emerging to address this problem. Proton Intelligence is a Canadian-based startup developing continuous potassium monitoring products. The company has now completed a $6.95 million seed funding round led by SOSV in the Bay Area. The product is scheduled to be launched in…
Lux, the developer of the iPhone camera app Halide, on Monday released a roadmap detailing its next version, Halide Mark III, scheduled for release next year. According to the company, Halide Mark III will come with two new features: color grade and unique HDR (high dynamic range) photography. The company plans to include color grades and filters that will make images look like they were taken with a film camera. The company's video app Kino already has a similar feature. The app will be redesigned to integrate new features into the UI. Notably, Lux released Halide Mark II in 2020.…
Last month, AI founders and investors told TechCrunch that we are now in the “era of the second law of scaling,” and how established techniques for improving AI models are facing diminishing returns. I pointed out whether it shows. One promising new method they proposed to maintain profits is “test-time scaling,” which appears to be behind the performance of OpenAI's o3 models, but which has its own drawbacks. there is. Many in the AI industry took the announcement of OpenAI's o3 model as evidence that progress in scaling AI is not “hitting a wall.” The o3 model performs well in…