Latest news with #dataownership
WIRED
09-07-2025
- Business
- WIRED
A New Kind of AI Model Lets Data Owners Take Control
Jul 9, 2025 1:59 PM A novel approach from the Allen Institute for AI enables data to be removed from an artificial intelligence model even after it has already been used for training. Photo-Illustration:A new kind of large language model, developed by researchers at the Allen Institute for AI (Ai2), makes it possible to control how training data is used even after a model has been built. The new model, called FlexOlmo, could challenge the current industry paradigm of big artificial intelligence companies slurping up data from the web, books, and other sources—often with little regard for ownership—and then owning the resulting models entirely. Once data is baked into an AI model today, extracting it from that model is a bit like trying to recover the eggs from a finished cake. 'Conventionally, your data is either in or out,' says Ali Farhadi, CEO of Ai2, based in Seattle, Washington. 'Once I train on that data, you lose control. And you have no way out, unless you force me to go through another multi-million-dollar round of training.' Ai2's avant-garde approach divides up training so that data owners can exert control. Those who want to contribute data to a FlexOlmo model can do so by first copying a publicly shared model known as the 'anchor.' They then train a second model using their own data, combine the result with the anchor model, and contribute the result back to whoever is building the third and final model. Contributing in this way means that the data itself never has to be handed over. And because of how the data owner's model is merged with the final one, it is possible to extract the data later on. A magazine publisher might, for instance, contribute text from its archive of articles to a model but later remove the sub-model trained on that data if there is a legal dispute or if the company objects to how a model is being used. 'The training is completely asynchronous,' says Sewon Min, a research scientist at Ai2 who led the technical work. 'Data owners do not have to coordinate, and the training can be done completely independently.' The FlexOlmo model architecture is what's known as a 'mixture of experts,' a popular design that is normally used to simultaneously combine several sub-models into a bigger, more capable one. A key innovation from Ai2 is a way of merging sub-models that were trained independently. This is achieved using a new scheme for representing the values in a model so that its abilities can be merged with others when the final combined model is run. To test the approach, the FlexOlmo researchers created a dataset they call Flexmix from proprietary sources including books and websites. They used the FlexOlmo design to build a model with 37 billion parameters, about a tenth of the size of the largest open source model from Meta. They then compared their model to several others. They found that it outperformed any individual model on all tasks and also scored 10 percent better at common benchmarks than two other approaches for merging independently trained models. The result is a way to have your cake—and get your eggs back, too. 'You could just opt out of the system without any major damage and inference time,' Farhadi says. 'It's a whole new way of thinking about how to train these models.' Percy Liang, an AI researcher at Stanford, says the Ai2 approach seems like a promising idea. 'Providing more modular control over data—especially without retraining—is a refreshing direction that challenges the status quo of thinking of language models as monolithic black boxes,' he says. 'Openness of the development process—how the model was built, what experiments were run, how decisions were made—is something that's missing.' Farhadi and Min say that the FlexOlmo approach might also make it possible for AI firms to access sensitive private data in a more controlled way, because that data does not need to be disclosed in order to build the final model. However, they warn that it may be possible to reconstruct data from the final model, so a technique like differential privacy, which allows data to be contributed with mathematically guaranteed privacy, might be required to ensure data is kept safe. Ownership of the data used to train large AI models has become a big legal issue in recent years. Some publishers are suing large AI companies while others are cutting deals to grant access to their content. (WIRED parent company Condé Nast has a deal in place with OpenAI.) In June, Meta won a major copyright infringement case when a federal judge ruled that the company did not violate the law by training its open source model on text from books by 13 authors. Min says it may well be possible to build new kinds of open models using the FlexOlmo approach. 'I really think the data is the bottleneck in building the state of the art models,' she says. 'This could be a way to have better shared models where different data owners can codevelop, and they don't have to sacrifice their data privacy or control.'
Skift
09-07-2025
- Business
- Skift
7 Red Flags to Avoid in Event Tech
Event tech vendors often make big promises. Here are seven warning signs that separate reliable partners from problematic providers, and how to spot them before signing contracts. When evaluating event tech vendors, watch for key warning signs. Minor red flags during evaluation can signal major post-implementation problems. Understanding these signs helps planners avoid costly mistakes, implementation failures, and unproductive vendor relationships. This article is based on the Event Tech Almanac 2025. Download your copy of the Event Tech Almanac 2025, the most complete guide to event tech today. 1. Control Over Event Data A key consideration for event planners procuring event tech is the data ownership model. The big question is who owns the event-generated data and how it's handled. Event Tech Evangelist Dahlia El Gazzar outlines key warning signs in vendor presentations. She suggests that "All your data lives with us" implies limited data ownership for clients, and is worth clarifying up front. Most event tech vendors act as data processors, keeping clients as the sole data controllers. This means clients have full ownership and control of attendee data. However, some vendors use a shared data ownership model, requiring attendees to become platform users to access an event, making them also data controllers. The latter option offers convenience for attendees. When they attend multiple events on the same platform, their profile is saved and they can jump straight into a new event with a full profile. The drawback is that organizations with stringent data policies are unlikely to consider event tech platforms that share control of data. 2. 'All-In-One' Platforms The term 'all-in-one' can mean different things to different people. For some vendors, it indicates an event tech platform that can handle in-person, hybrid, and virtual events. For others, it's a platform that offers features that support the whole event journey, all the way from the marketing and registration to post-event surveys. For others, it helps planners source venues, book travel, and manage attendees. While 'all-in-one' alludes to a complete offering and is used in marketing by many vendors, it isn't helpful for planners with different needs. 'We know that the 'all-in-one' term is a myth. A misleading holy grail that gets event profs in trouble and causes disappointment,' said El Gazzar. In the Event Tech Almanac 2025 report, we removed this event tech tool category, referring to many platforms using this terminology as 'event management tools,' if they offer a minimum set of features. But that didn't stop six vendors from using this term in their platform descriptions. The challenge persists, so the term is a red flag and requires clarification. As registration specialist Leanne Velky puts it, "Be wary of anyone who says they can do it all for you. They can't." 3. Tech Under Development The tech world moves fast, and event tech is no exception. Vendors regularly implement new features based on planner feedback. Some even share their development roadmap publicly. As planners explore features and functionality, experts encourage them to watch for phrases like "We'll build that later," that El Gazzar said indicates nonexistent features. Brandon Wernli, CEO of BW Events, warns against vendors doing demos with wireframes without live examples. He said, 'This may suggest they are selling while the build is in flight.' He encourages planners to seek references of satisfied customers to verify claims. 4. Lack of Flexibility Templates are a useful feature of event tech. In the Skift Meetings Event Tech Almanac 2025, 33 of the 40 (83%) event management tools offer them. But planners need adaptable solutions, not rigid systems with narrow constraints. In an ideal world, event tech would adapt to every client's needs. However, platforms must balance flexibility with functionality. Developing it involves making choices about features and customization options. Event tech experts recommend watching out for vendors offering solutions that can only be used as envisioned by developers. Kazia Ekelund, event strategist at Spark Event Collective, recommends avoiding those that restrict clients from deviating from their template or offer limited customization. She calls these vendors "highly proprietary." 5. Integration Challenges Integrating different event tech tools can catch organizations off guard. Solutions marketed as "integrated" may still require significant manual coordination. Learning curves may be steeper than anticipated, causing deployment delays and process changes that disrupt workflows. "Scoping out the true integration needs is a task that is often under-resourced in itself, let alone finding the resources and the budget to actually deliver integrations," said Vanessa Lovatt, Founder of Event Tech World. Doug Muller, vice president of technology and innovation at George P. Johnson, recommends finding "open and collaborative teams willing to show their progress and identify where their solutions end and manual processes begin." 6. Inadequate Support A deal breaker for event tech is insufficient support. Live events require live support, but not all vendors can provide it. Experts recommend verifying direct phone access to support teams, their location, and committed response times. Pay attention to guaranteed response times in service level agreements (SLAs) and watch for vague language that could leave you without help. "Be sure to truly understand a platform or provider's support structure. Ensure you have a provider where you can pick up the phone and talk to someone," said Aaron Dorsey, vice president of product management, information security and privacy at Maritz. 7. Excessive Focus on Flashy Features AI is undoubtedly today's 'shiny object.' However, flashy doesn't guarantee quality or a good fit for events and clients. El Gazzar calls this "Shiny Object Syndrome," and it's rampant among AI-focused startups. But experts warn against choosing technology based on trends rather than needs. Brandt Krueger, senior production manager, EideCom, said, "New and shiny is almost never the answer. I prefer to see old technology used well than new technology used simply because it's new." Of the 85 event tech vendors in the Event Tech Almanac 2025, over three-quarters (76%) use AI. Among event management platforms, that jumps to 82%. It is used across multiple functions, with data analytics and reporting the most common (56%) followed by matchmaking (48%), session or content suggestions (44%), writing assistance (44%), and captioning and transcription (41%). The challenge of separating AI reality from hype illustrates a broader truth in event tech evaluation: Vendors often oversell capabilities while understating limitations. Understanding common red flags helps planners cut through marketing promises to find solutions that truly deliver value. Pricing structures and contract terms can also reveal vendor misalignment. Experts recommend watching for aggressive multi-year commitments, unclear usage limits, or pricing that scales unpredictably with event size. Red flags may reveal everything from platform flaws to simple communication issues. But identifying red flags is only one part of the challenge. Planners should use this information to verify vendor capabilities and start constructive conversations that help drive better results.
