French startup Rounded is developing an orchestration platform that allows companies to create their own AI voice agents. Initially focused on web3, the company shifted its attention to AI-powered voice interactions in mid-2023. Its first product, Donna, was designed for anesthetists, helping private hospitals handle large volumes of routine patient calls. The AI agent has already managed hundreds of thousands of conversations, improving in speed and accuracy over time.
After refining its technology, Rounded expanded its focus to offer a platform where businesses can build their own AI voice agents. Users can integrate various AI models, such as speech-to-text and text-to-speech engines, selecting components from providers like Azure, GPT-4o mini, and ElevenLabs. The platform also helps define prompts and parameters to optimise each agent’s performance for specific use cases.
The startup has secured €600,000 in funding from UC Berkeley’s SkyDeck accelerator and business investors. With growing interest in AI-powered customer interactions, Rounded is poised to attract further investment as it expands its product offering.
AI is transforming the way new medicines are developed, with AI-powered drug discovery advancing at an unprecedented pace. Insilico Medicine, a US-based biotech firm, has designed an experimental treatment for idiopathic pulmonary fibrosis (IPF) using AI to identify a potential drug target and generate molecules. The approach significantly reduced the time and resources needed, with the drug discovery process taking 18 months instead of the usual four years.
AI-driven methods are being adopted by both startups and major pharmaceutical companies to accelerate drug development. Insilico Medicine has multiple drug candidates in clinical trials, while Recursion Pharmaceuticals is using AI to analyse vast biological datasets and uncover new treatment possibilities. A molecule designed by Recursion to target lymphoma and solid tumours has already entered human trials, demonstrating the growing potential of AI in medical research.
Despite the progress, experts note that AI-discovered drugs have yet to complete full clinical trials. The technology faces challenges, particularly in data availability and bias, but researchers remain optimistic. As AI continues to refine drug discovery, many believe it will lead to faster, more cost-effective treatments and a higher success rate in bringing new medicines to market.
Google is testing a new feature called “Daily Listen,” which generates personalised AI-powered podcasts based on users’ Discover feeds. The feature, currently rolling out to US users in the Search Labs experiment, provides a five-minute audio summary of topics tailored to individual interests. Each podcast includes links to related stories, allowing listeners to explore subjects in greater depth.
The experience is integrated with Google’s Discover and Search tools, using followed topics to refine content recommendations. Daily Listen functions similarly to NotebookLM’s Audio Overviews, which create AI-generated audio summaries based on shared documents. Users who have access to the feature will see a “Daily Listen” card on their Google app’s home screen, displaying a play button and episode length.
Once launched, the podcast plays alongside a rolling transcript, offering a seamless blend of text and audio. Google aims to enhance how users consume news and stay informed, making the experience more interactive and personalised. The feature reflects the company’s ongoing push into AI-driven content delivery.
Google is merging additional AI teams into its DeepMind division to speed up innovation in AI technologies. Logan Kilpatrick, head of product for Google’s AI Studio, confirmed that both the AI Studio team and the Gemini API developers would now operate under DeepMind.
DeepMind, created in 2023 from the merger of Google Brain and DeepMind, has played a central role in Google’s AI advancements, including the Gemini model series. Kilpatrick stated the restructuring would strengthen collaboration and accelerate progress in making research tools available to developers.
Engineer Jaana Dogan highlighted that the move would make DeepMind’s tools more publicly accessible, with better APIs, open-source contributions, and enhanced developer resources planned. This shift follows earlier integrations of the Gemini chatbot and responsible AI teams into DeepMind as part of Google’s ongoing strategy.
CEO Sundar Pichai previously described Gemini as gaining strong momentum while stressing the need to move faster in 2025 to close competitive gaps. Scaling Gemini for consumers will be a primary focus next year.
Nvidia has voiced strong opposition to a reported plan by the Biden administration to impose new restrictions on the export of AI chips, urging the outgoing president to avoid making a decision that could impact the incoming Trump administration. The company warned that such measures would harm the US economy, hinder innovation, and benefit adversaries like China. Nvidia’s Vice President, Ned Finkle, called the policy a “last-minute” move that would leave a legacy of criticism from both US industry and the global community.
The proposed restrictions, as reported by Bloomberg, aim to limit AI chip exports to certain countries, particularly targeting China to prevent the enhancement of its military capabilities. While some nations would face outright bans, the rules would also cap the computing power that can be exported to others. The Biden administration has yet to confirm the details, and requests for comment from the White House and the Commerce Department went unanswered.
Industry groups, including the Information Technology Industry Council, which represents major tech firms like Amazon, Microsoft, and Meta, have expressed concern about the policy. They argue that it would impose arbitrary limitations on US companies’ global competitiveness and risk ceding market leadership to foreign rivals. Nvidia warned that these restrictions could push international markets toward alternative technologies, undermining the US technology sector.
President-elect Donald Trump, who begins his second term on January 20, previously enacted technology export restrictions to China during his first term, citing national security concerns. Nvidia’s statement reflects apprehension about the continuity of US policy on AI chip exports under the new administration.
Brazil is becoming a major hub for data centres, with billions of dollars in investments expected as AI and digital infrastructure drive demand. Growth limitations in larger markets such as the United States and Europe have led companies to expand in Brazil, taking advantage of the country’s political stability, renewable energy resources, and increasing local demand. Several major players, including Ascenty, Equinix, and ODATA, are making significant investments to establish or expand their facilities.
Grupo FS is entering the sector with a $1.8 billion investment to build three new data centres. Equinix is expanding in São Paulo and Rio de Janeiro, while ODATA continues to grow following its acquisition by Aligned Data Centers. Tecto, backed by Brazilian bank BTG, has secured land in São Paulo for a new hyperscale facility powered entirely by renewable energy. Meanwhile, Elea Data Centers is expanding across Brazil, focusing on sustainable operations in key regions.
Investors see Brazil as an attractive destination due to its growing digital economy and strategic location. Companies are increasing their footprint to meet rising demand for cloud computing and AI-driven technologies. As the industry expands, Brazil is positioning itself as a leading data centre hub in Latin America, benefiting from a mix of international and domestic investment.
Danish pharmaceutical giant Novo Nordisk is strengthening its collaboration with a United States tech firm Valo Health to develop new treatments for obesity, diabetes, and cardiovascular diseases using artificial intelligence and human data. The agreement, originally signed in 2023, has been expanded to cover up to 20 drug candidates, nearly doubling the initial scope of 11 treatments.
The expansion comes as Novo seeks to maintain its competitive edge in the booming obesity drug market, expected to be worth $150 billion in the next decade. A recent clinical trial for its weight-loss drug candidate, CagriSema, delivered underwhelming results, increasing pressure to develop a successor to its best-selling drug, Wegovy. Rival pharmaceutical company Eli Lilly is also pushing forward with its own obesity treatments, intensifying the race for dominance in the sector.
Under the revised deal, Valo Health will receive up to $190 million in near-term payments and milestone payments of around $4.6 billion, significantly increasing its earnings potential compared to the original agreement, which offered up to $2.7 billion. Novo hopes the collaboration will lead to groundbreaking therapies that extend the health benefits of weight-loss drugs beyond obesity treatment.
Halliday, a wearables startup, has launched a pair of smart glasses at CES 2025 that project a tiny digital screen directly into the wearer’s eye. Using a device called the DigiWindow, the glasses display notifications, language translations, and navigation directions in real time without the need for bulky AR lenses.
Priced at $489, the glasses use a small green light to beam an almost 9cm round display into the user’s line of sight. The innovative approach makes US based Halliday’s glasses slimmer, lighter, and more affordable than many augmented reality prototypes. Users can even fit prescription lenses into the frames without affecting the display.
Key features include real-time translation for 40 languages and a proactive AI assistant that offers helpful information during conversations. The device is controlled via a ring worn on the finger, allowing users to navigate its features with thumb gestures. While the AI assistant wasn’t available for testing, the display technology impressed with its functionality.
Halliday’s smart glasses are already available for preorder at a discounted price of $369 via Kickstarter. Shipping is expected to begin in March 2025. The company hopes its sleek design and practical applications will set the glasses apart from other wearables still stuck in prototype stages.
San Francisco-based startup Based Hardware has unveiled Omi, a wearable AI assistant designed to improve productivity. Launched at the Consumer Electronic Show, the device responds to voice commands when worn as a necklace or can attach to the side of the head using medical tape, activating through a unique “brain interface.”
Unlike other AI gadgets that aim to replace smartphones, Omi is meant to complement existing devices. It can answer questions, summarise conversations, and manage tasks like to-do lists and meeting schedules. The startup’s founder, Nik Shevchenko, claims that Omi’s brain interface allows users to interact without saying a wake word by recognising mental focus. However, this feature has yet to be widely tested.
Based Hardware built Omi on an open-source platform to address privacy concerns. Users can store data locally and even develop their own apps for the device. Priced at $89, the consumer version will ship later in 2025, while a developer version is already available.
Omi enters a growing market of AI gadgets that have struggled to meet expectations. Shevchenko hopes Omi’s focus on practical productivity tools will set it apart, but the device’s success will likely depend on whether users embrace its experimental brain interface feature.
Elon Musk’s AI company, xAI, is preparing to launch a controversial feature for its chatbot, Grok, called ‘Unhinged Mode.’ According to a recently updated FAQ on the Grok website, this mode will deliver responses that are intentionally provocative, offensive, and irreverent, mimicking an amateur stand-up comedian pushing boundaries.
Musk first teased the idea of an unfiltered chatbot nearly a year ago, describing Grok as a tool that would answer controversial questions without self-censorship. While Grok has already been known for its edgy responses, it currently avoids politically sensitive topics. The new mode appears to be an effort to deliver on Musk’s vision of an anti-‘woke’ AI assistant, standing apart from more conservative competitors like OpenAI’s ChatGPT.
The move comes amid ongoing debates about political bias in AI systems. Musk has previously claimed that most AI tools lean left due to their reliance on web-based training data. He has vowed to make Grok politically neutral, blaming the internet’s content for any perceived bias in the chatbot’s current outputs. Critics, however, worry that unleashing an unfiltered mode could lead to harmful or offensive outputs, raising questions about the responsibility of AI developers.
As Grok continues to evolve, the AI industry is closely watching how users respond to Musk’s push for a less restrained chatbot. Whether this will prove a success or ignite further controversy remains to be seen.
