It seems like there is no corner of technology considered “safe” from the AI boom. Every single popular piece of technology, from earbuds to laptops, is getting on board with artificial intelligence. Apple’s WWDC 2024 developer’s conference, for example, showed off Siri’s new ChatGPT partnership, part of the company’s new Apple Intellligence service.
Even the best fitness apps aren’t immune. First, running and cycling app Strava made waves with its Athlete Intelligence AI analytics feature, and now it’s AllTrails’ turn. One of the top fitness apps for outdoor adventurers, AllTrails helps walkers find new routes (perfect for getting those 10,000 steps in), aggregating comments, pics and reviews from other walkers on around 400,000 trails.
AllTrails is debuting a redesigned Trail page and a handful of new features, presented to me during an exclusive chat with AllTrails’ Chief Technical Officer James Graham. First up is a redesign of the individual Trail pages, which according to Graham are “the most popular pages on the app.”
Distance and Difficulty levels are now at the top of the page, and you can use the new Photo Tour feature to view key points along the way. AllTrails uses machine learning in conjunction with “geoclustering” to look through photos of each trail taken by its user base, ensuring photos appear at the correct points. You can see exactly where people took certain photographs along the trail. Clever.
On each trail page, you can also find reviews of the trail, and AllTrails has collaborated with OpenAI on what Graham refers to as “ReviewGPT”: A ChatGPT-powered summary of all the reviews on the trail can now be seen at the top of each review section on the Trail Page. For example, opening one page near me shows the following AI-generated summary:
“Recent reviewers of Shiplake Circular… mention that the trail may be muddy in parts, making walking boots necessary. The route offers varied scenery and is recommended by visitors. While some parts are muddy along the river, there is no flooding currently. It’s suggested to walk the route in reverse to tackle the hill first.”
The app also offers a new Collections feature, which allows you to sort local trails by categories such as “Among Trees”, “Epic Views”, “Wildflowers” or “Dog-Friendly”. AllTrails has also introduced comprehensive Parks pages, showing complete guides and information about national parks, including popular trails, fees, parking, times to visit and more.
Although it’s fairly focused on the US at the moment due to its bounty of national parks (and the app’s creators are based in Silicon Valley, of course), there are guides to parks outside of the US. In the UK, there are guides to 10 parks in England, two in Scotland, and three in Wales.
Analysis: Helpful AI features, but how is security?
While it makes a lot of sense to use AI in this context (no more parsing through reviews to find out what gear to bring or whether a trail is dog-friendly), pouring information about our location, health and walking habits into a third-party AI service could be cause for concern for skeptics.
Graham was quick to reassure me that AllTrails’ user data was safe. “We have certain guarantees that OpenAI will not train and leverage any user context we provide in order to train their models or use any user data,” he said.
“We also use RAG – retrieval augmented generation – that allows the LLM to ask for specific data, so it never actually makes it into the LLM. Data separation exists at the architecture level to ensure user privacy is guaranteed.”
Graham later shared the following on AllTrails’ behalf:
“To ensure our user data remains safe and secure, we adhere to the OpenAI data-sharing agreement, which guarantees that data sent to the OpenAI API will not be used to train or improve OpenAI models. Our pre-trained model usage remains within our secure AWS environment and is not shared, sold, or accessed by any third party.”
You might also like:
https://cdn.mos.cms.futurecdn.net/EZe58uTFVARWpqwAikhCuH-1200-80.jpg
Source link
matt.evans@futurenet.com (Matt Evans)
