This week in AI: February 4 - 10

Every weekend, we bring together a selection of the week's most relevant headlines on AI. Each weekly selection covers startups, market trends, regulation, debates, technology, and other trending topics in the industry.

Jua, a startup building the first 'large physics model' of the natural world, raised $16M in a successful seed round

The ambitious Swiss startup, Jua, claims to be operating the first AI "Large Physics Model" predicting weather with extreme [sic] high accuracy, precision, and speed. Given that this is the model's primary focus, Jua advertises itself as providing solutions for several energy trading use cases. This may come across as a risky strategy since neither weather prediction nor the leveraging of AI tools to solve problems related to weather prediction is novel. However, Jua's confidence in its proposal rests on its model being significantly larger than alternatives such as Google DeepMind's GraphCast and that the startup found weather to be a reasonable starting point before taking on a more complicated set of physical questions and challenges. Moreover, one should also consider that, despite being at a very early stage, Jua has earned the trust of enough investors to successfully close a $16 million seed round co-led by 468 Capital and the Green Generation Fund, and counting with the participation of Promus Ventures, Kadmos Capital, Flix Mobility founders, Session.vc, Virtus Resources Partners, Notion.vc and InnoSuisse.

ChatGPT is getting an Apple Vision Pro app

OpenAI has released one of the first apps built for visionOS, the operating system powering the Vision Pro headset, and Apple's first foray into spatial computing. An announcement via X features a quick showcase of the app from a Vision Pro user's point of view. The ChatGPT visionOS app features the same minimalistic chat interface as its iOS, where users can type in their prompts and receive answers from ChatGPT. Although the details are still unclear, there is no news on whether the visionOS ChatGPT app will eventually profit from the operating system's unique features, such as Optic ID and VisionKit. Regardless, it is highly likely that the new app will follow the iOS apps in including Voice mode, which enables users to interact with the chatbot via voice messaging rather than text, as well as the iOS app features that allow users to share images as input. The ChatGPT app has been available at the visionOS store, starting February 2.

Satya Nadella's keynote at Microsoft CEO Connection event highlights the company's commitment to helping India become an AI-first country

During his address, Nadella announced the ADVANTA(I)GE INDIA initiative, a Microsoft-led program providing 2 million Indians with AI skilling opportunities. The keynote also highlighted how Microsoft's AI solutions portfolio is helping people be more productive by enabling them to complete tasks faster and with a higher quality standard. For example, recent research on Copilot for Microsoft 365 early users found a 29% increase in the speed at which repetitive tasks such as searching and summarizing were completed. Of all the organizations profiting from Microsoft's AI offerings, the case of Karya stands out. The self-described "world's foremost ethical data company" is creating high-quality datasets in India's local languages, enabling populations that are not Hindi or English-speaking to improve their lives and boost the creation of technological tools for the better of their communities.

Entrust is in exclusive negotiations to acquire Onfido, allegedly for over $400 million

Sources recently revealed that the early and exclusive discussions revolve around an over $400 million pricetag on the transaction. The acquisition of London-based Onfido, a company specializing in cloud-based AI-powered identity verification (IDV) technology, would add the first AI-based biometric and document IDV tech stack to Entrust's portfolio. Onfido currently services over 1,200 global customers, including leading financial institutions, gambling platforms, and sharing economy services. Since its inception, the company has carried out over 200 million identity checks from over 2,500 document-types and passports from 195 countries. If the transaction succeeds, Onfido would be integrated into Entrust's portfolio, which caters to more than 10,000 customers worldwide, including governments and enterprises requiring strong identities, secure payments, and protected data solutions.

Backed by neuroscientists, Elemind came out of stealth and closed a $12M seed round for its AI-assisted wearable

The research team behind Elemind hails from leading research institutions, such as the Massachusetts Institute of Technology (MIT), Imperial College London, Harvard Medical School, and the University of California at Berkeley. Elemind's neurotechnology wearable reads the user's brainwaves and responds to them with customized stimulation that changes their behavior in a specific and targeted manner. Elemind refers to its approach as 'electric medicine' and describes it as "a drug-free, personalized and adaptive approach that fine-tunes the stimulation based on the body’s response until the desired state is achieved." The technology's effectiveness is supported by five clinical trials and several publications in peer-reviewed scientific journals describing applications such as sleep induction, physical tremor suppression, memory boosting and enhanced sedation. Elemind was first funded by Village Global, but other investing entrepreneurs, funds and business leaders include LDV Partners, MIT’s investment fund, E14 Fund, the founders of Skype, Nest, Opentable, Broadvision, Boston Scientific, and others.

Colossyan announces a $22 million Series A round

The raised funds will be invested into furthering the development of Colossyan's AI video platform. The company has been helping industry leaders such as Novartis, Vodafone, and Paramount create training videos that are less time consuming and more cost effective. Colossyan has stated that its customers are able to cut as much as 90% of their video costs, while saving up to 80% of their time. The funding was provided by a new investor, Lakestar, with contributions from Collosyan's previous investors, Launchub, Day One Capital, and Emerge Education. The company plans to direct the funds towards product, customer success, and community teams growth and development acceleration. Development will prioritize essential features such as Gestures (directed body movements for avatars), a new version of its Editor, and advancing its lip-sync technology. Colossyan is also contemplating interactive quizzes and adaptive learning paths to improve the platforms effectivity.

Ambience Healthcare, the company behind “the leading AI operating system for healthcare”, raised $70M from Kleiner Perkins, OpenAI and others

The recently raised investment brings Ambiance's grand total to $100 million. Founded in 2020, the company aims to reduce clinicians' burnout by assisting with repetitive tasks such as documentation, coding, referrals, and visit summaries. Ambience can generate specialty-agnostic visit notes by synthesizing the information it pick up from listening into conversations between medical professionals and patients; the resulting notes are compatible with the most popular EHR systems. The company differentiates itself from competitors by offering an AI operating system rather than point solutions: in addition to visit notes, the operating system can also assist with medical coding, compliance support, referral letter generation, and after-visit summaries for patients. The company is also planning to add features to help clinicians prepare for their appointments, such as providing them with patients' context beforehand and offering suggestions for the appointment agenda.

Analytical Alley raised €700k for its AI-driven marketing solution

According to Analytical Alley, 40% of advertising results in ad waste. Moreover, 65% of budget is lost in offline channels, with investment mismatch and mis-timed investments contributing to the problem. Thus, Analytical Alley's managed software-as-a-service (mSaaS) aims to help companies reduce their percentage of ad waste by enabling them to embark on smart and calculated marketing decisions. The process begins with a review of the client's past performance to extract insights on the strategies that did or did not work. Then, the AI-assisted platform offers suggestions on how future campaigns may develop. Finally, new marketing efforts are evaluated upon execution, generating further insights for campaign optimization. Boasting a client list that includes PHH Group, Norsk Tipping, Kristiania University, and Coop Bank; Analytical Alley operates in the Baltics and is planning to expand further into the European market this year.

Meta is working on a new standard to label AI-generated images on Facebook, Instagram and Threads

Meta currently labels media generated with its Meta AI features so the public can easily identify this kind of media in Meta's social platforms and elsewhere. However, the company has recognized that it has not always been able to tag AI-generated media created using other companies' tools. To cover this gap, Meta has started working on the capability to identify the the “AI generated” information in the C2PA and IPTC technical standards so it can also label AI-generated content coming from Google, OpenAI, Microsoft, Adobe, Midjourney, and Shutterstock as they implement their own plans to identify AI-generated media using their tools. Regardless, Meta's new approach will still have its limits, as there will still be content that escapes the new identification standards, and there currently exist strategies to strip AI-generated content from their identification watermarks and metadata. As a consequence, the company is still focusing on the identification and removal of harmful content, regardless of whether it is AI-generated or not. Communicating the deployment of these efforts and their limitations continues to be part of Meta's mission to develop AI transparently and accountably.

Greyparrot and Bollegraaf look to transform global waste management with AI

Bollegraaf is the world's leading builder of recycling plants, while Greyparrot is a pioneer in AI waste analytics. The companies are joining forces to transform thousands of recycling and sorting centers into more efficient operations. The inaugural agreement in this partnership consists in Bollegraaf becoming the Greyparrot Analyzer's first worldwide distributor. Bollegraaf is also investing $12.8 million in Greyparrot, representing a non-controlling stake in the business. Additionally, Bollegraaf will also transfer its AI vision business to Greyparrot, and Greyparrot will open its first European office in the Netherlands. An expert estimation states that, by 2040, a sixfold increase in recycling capacity will be necessary to keep up with global waste production. Bollegraaf and Greyparrot's partnership represents both companies' commitment to use the power of AI to contribute to the solution of the upcoming global waste crisis.

FCC declares AI-generated voices 'artificial', making them illegal to use in automated calls

The revision of whether AI-generated voices reading a script during an automated call fell into the rule outlawing robocalls was likely motivated by a high-profile case involving automated called received by the residents of New Hampshire, in which an AI-generated version of President Biden's voice asked people not to waste their vote in the upcoming primary elections. FCC's latest ruling clarifies that AI-voices are covered under the regulations banning unconsented robocalls and prerecorded calls. According to the ruling, since AI-generated voices have already been exploited to dishonestly earn the called party's trust, consent for using technology resembling a live agent. By doing this, consumers will be reminded that they have the right not to receive these kinds of calls, and if they do consent to them, it will represent an additional step that reminds the called party to proceed with caution when handling prerecorded and automated calls.

The EU launched a consultation containing draft election security mitigations along with other content moderation recommendations

With the release of these guidelines, the European Union is trying to reduce the democratic risks stemming from AI-generated content and deepfakes, as well as other previously-known threats such as lack of content moderation, media illiteracy, and lack of transparency in political advertising. The election guidelines are especially aimed at social media and search giants such as Facebook, Google, TikTok and X (Twitter), which are companies singled out under the Digital Services Act (DSA). An essential suggestion concerns the clear and persistent labeling of AI-generated media and deepfakes, especially when the content depicts existing people or places, or depicts events that may be misrepresented or non-existing. More generally, the singled-out companies are expected to deploy "reasonable, proportionate, and effective" mitigation measures that address the creation and dissemination of harmful material.

The Biden-Harris administration announces first-of-its-kind consortium on AI safety

The U.S. AI Safety Institute Consortium (AISIC) will bring together AI creators, users, academics, government and industry researchers, and civil society organizations so they can contribute to the development and deployment of safe and trustworthy AI. The AISIC will be part of the U.S. AI Safety Institute (USAISI) and will prioritize the actions outlined in President Biden’s landmark Executive Order. The AISIC is composed by over 200 companies and organizations that are leading the way in the creation and deployment of AI systems in the U.S. The member companies and organizations are especially well-positioned to understand how AI is shaping and will continue to transform our society, and as a result, their representatives are the best candidates to integrate the consortium's collection of test and evaluation teams. The AISIC will also count with state and local government representation, and will aim to cooperate with organizations from like-minded countries towards the development of safer AI.

Crux announces its successful seed round

The $2.6 million round was led by Emergent Ventures, with participation from Y Combinator and Neon Fund, and it includes other early investors. The startup aims to bridge the gap between users and data-backed intelligence with its enterprise Decision-Making AI Copilot. The personalized AI Copilot will be aware of enterprise information, such as business context, preferences, KPIs, and macro-trends, and will assist decision-making based on data. As a first step, Crux wants to help SaaS teams embed the Copilot in their products for its customers and CS teams. Once it is integrated, customers and teams can profit from the clarification's agent capabilities which can provide increasingly accurate assistance, thus boosting efficiency and productivity.

Daedalus is building AI-powered factories for custom precision parts

Daedalus takes orders for medical devices, aerospace, defense, and semiconductors and manufactures the bespoke precision parts to match the customer's specifications. The company has recently announced a $21 million Series A round of funding led by Nokia-funded NGP Capital with the participation from existing investors Khosla Ventures and Addition, which will allow the company to scale its operations. This Series A round has taken Daedalus total raised funding over the $40 million mark. Daedalus uses the hardware available to all manufacturers but deploys it with a software interface that allows the AI-powered automation of key parts of the manufacturing process. The software then extracts key insights from the human-assisted manufacturing process for a specific parts and uses that information to further refine attempts at manufacturing similar pieces. Prior to Daedalus, CEO Jonas Schneider worked as technical lead at OpenAI, where he helped launch the company's robotics division. His experience working with a robotics unit needing replacement parts at OpenAI motivated the creation of Daedalus in 2019.

Arize AI releases Phoenix 3.0

Phoenix is an MLOps and LLMOps monitor for models and LLM applications that provides LLM traces, evaluation, and embedding, RAG and structure data analysis. Phoenix 3.0 is fully OpenTelemetry compliant and natively renders LLM application data via OpenInference. In practice, this means that Phoenix can be deployed as a collector on Docker and K8s with your choice of instrumentation (LlamaIndex, LangChain, OpenAI, etc). Client-based interaction with remote Phoenix instances is also possible, as is the creation and extension of spans with custom attributes. Finally, Phoenix 3.0 natively supports JavaScript.