 |
|
|
|||||
Camb.ai is on a mission to disrupt the dominance of English in global media. Founded in 2022, the AI-powered platform specializes in real-time translation that retains a speakers emotional resonanceprocessing content up to 20 times faster than traditional dubbing services. Major League Soccer now uses Camb.ais technology for live broadcasts. But the company has also found unexpected demand in markets like video advertising and the localization of interactive smart toys. To power its growth, Camb.ai has raised $15.5 million to date. The platform now supports translations in more than 150 languagesincluding Maleku, spoken by just 500 people. CEO Avneesh Prakash, who previously helped build Indias Aadhaar biometric ID system used by more than a billion people, cofounded the company with his son, Akshat Prakash. The younger Prakash, Camb.ais CTO, is a computer scientist and former AI/ML engineer on Apples Siri team. Avneesh Prakash envisions a future where English is no longer the default language for media productionand where global audiences can access any content, in any language, on demand. Fast Company spoke with Prakash about AIs potential to reshape global media, the complexities of preserving emotional nuance across languages, and why rare languages remain central to Camb.ais mission. This conversation has been edited for length and clarity. What misconceptions do people have about AI voice technology, and how do you address those concerns? People are concerned about inaccuracies when using AI, but they often overlook that even human translators have flaws. When evaluating AI, people try to find that moment of Oh, it went wrong there. Often these are subjective opinions, and such analyses do not use a comparative benchmark of how human translators would do on the same piece. The best approach is to enable human translators with AI like ours so they can be 50 times more productive and help cover a large body of work that today remains locked up in one to two languages, like English. What metrics do you use to measure success beyond traditional business growth indicators? One way we measure success is the number of languages we can translate into. Our mission is to redesign the internet for speakers of every language. Thats why weve also put a lot of effort into our capability to translate rare or endangered languages like Icelandic or Indigenous languages like Maleku. We already support more than 150 languages for speech-to-speech translation and our goal is to grow this number three to four times over the next two years. What advantages do you have in competing against large tech players and giant AI companies, and what are the biggest hurdles for you to overcome? Compared to the household names in AI, we can make models that are hyperfocused on being the best at translation and dubbing. We are also focused on the open-source community and can use its feedback to iterate and develop faster. The biggest challenge we face is one plaguing the entire AI industry: access to the computing power necessary to continue innovating. To remain competitive, we have prioritized building smaller models capable of being run on a users device, rather than the race to the biggest model in Big Tech. Your recent partnership with Legible focuses on books. What other content types present the most compelling opportunities? One unexpected opportunity were capitalizing on is translating advertisements (both picture and video). With traditional translation tools, its very difficult to translate ads in a way that makes sense culturally. A lot of advertising relies on metaphors, analogies, and cultural references. In the past, if you were to translate ads directly, a lot of the context would get lost in translation. Our models can overcome that hurdle. What do you envision happening to translators as AI dubbing technology advances? I envision a future where content creators and translators work alongside AI rather than work against it. AI will be able to provide a first draft translation, but there will always be scenarios (especially in literature and poetry) where a human touch is needed. What’s a common assumption about the future of global content that might be flawed? Most people assume content will continue to be English-first. While a majority of the global content is currently produced in the U.S., in English, with demographic and technology shifts, I see a future where a majority of the worlds media is originally produced in languages other than English. Which unexpected industries or sectors have shown the most interest in your technology? One interesting use case has been in the smart toy industry, where more and more toys are becoming interactive and AI-enabled. Localization in this context has the incredible potential of teaching children their own culture and language; this gets increasingly lost in the modern world. Looking ahead five years, what do you expect to be the most significant change in how we consume cross-language content? We will see all content available in all languages. If you go on Netflix or YouTube right now, youll see some content being translated or captioned into a limited number of languages. In less than five years, I expect we will be able to view that same content in tens or hundreds of languages on demand. How does AI-powered dubbing/live translation fundamentally change the economics of global content distribution compared to traditional methods? With AI translation, markets and audiences that were previously considered financially unviable now become accessible. AI translation rapidly increases the speed at which content can be spread around the world. Weve seen our technology dub content up to 20 times faster than traditional dubbing agencies, so content can be released worldwide simultaneously. Beyond cost savings, what unexpected benefits do you see for AI live translations? Certain cultures have populations greater than that of the U.S. For example, the number of Bengali speakers is larger than the populations of many countries combined. In many such cultures, sports/content/media has the opportunity to reach everybody and unlock a new generation of accessibility and viewership for businesses. Critics argue AI-dubbed content lacks the soul of human performance. How do you address this perception, technically and philosophically? With our models, preserving emotion and soul has been the number-one priority. By training our models on both text and raw audio, the model learns how different words, punctuation, and context relate to various emotions and expressions. For us, translation is a way to share human expression across cultural boundaries, and ensuring that we maintain the emotional meaning of speech is the essence of what we do. Camb.ais mission is to let “every story be told in every language.” How might this reshape cultural power dynamics? Could a Gambian filmmaker compete more effectively against someone benefiting from Hollywoods global influence? Exactly. Thats our vision. As technology like ours becomes more pervasive, I expect to see content that breaks the internet coming from all corners of the globe. Major League Soccer used Camb.ai to live-dub commentary into four languages simultaneously. Is sports broadcasting reaching a linguisic tipping point? Prior to MLS using our technology, there was very little appetite for using AI in a livestream context. This milestone has led to AI being considered a viable alternative for commentary and dubbing, and were now seeing more and more global sports organizations adopt the technology. Whats your mission in competing in this ultra-competitive AI arena? The internet was made for English speakers, and we decided to redesign it for the world. While language is a tool of diversity and hence evolution, it is also a tool of exclusion. I wouldn’t be here if I didn’t have the good fortune of going through an English-language education. I’m grateful for that, but I also see the unfairness of that. We created a company to disrupt that disparity. As noted in the film Ratatouille, Not everyone can be a great artist. But a great artist can come from anywhere. We are trying to create a world where a great artist born anywhere, creating anywhere, is able to take their content to any other part of the world.
Category:
E-Commerce
When Anirudh Rao was 4 years old and living in Nashville, his friends house was destroyed by a tornado. A year lateryes, as a kindergartnerRao started sketching a potential solution for better tornado warnings. Now 12 years old and living in Colorado, Rao is pursuing a more advanced version of his concept: a network of drones that could theoretically sense infrasound, a wave phenomenon emitted before and during tornadoes with frequencies below the threshold of human hearing. On the outskirts of the city, therell be a base station and a network of autonomous drones that fly in all directions, says Rao. He envisions sensors detecting infrasound along with temperature, pressure, and altitude, and sending data back to the base station; if a tornado is detected, that information could go to local authorities to trigger an official tornado warning and push notifications on phones. [Photo: courtesy Anirudh Rao/Young Planet Leaders] Currently tornadoes are detected as they have been for decades: through radar and by storm chasers visually spotting them on the ground. But radar doesnt work perfectly. As storms have started to move east out of the traditional Tornado Alley in the center of the country (a trend that may be happening because of climate change), theyre also moving into hillier topography, where radar is even less reliable. As tornadoes are shifting eastward, theyre also reaching more populated areas, increasing the risk. [Photo: courtesy Anirudh Rao/Young Planet Leaders] Raos instinct to use infrasound is in line with the latest science. The idea isnt new, though until tornado patterns changed, radar had been considered good enough. As tornadoes move into hillier areas and it limits the effectiveness of radar [science is] refocusing on the potential use of infrasound, says Brian Elbing, a mechanical and aerospace engineering professor at Oklahoma State University. Though it still isnt fully understood how tornadoes produce infrasound, you can pick up the signal before the tornado touches the ground, and it lasts the life of the tornado, Elbing says. And it carries information about the strength of the tornado. [Photo: courtesy Anirudh Rao/Young Planet Leaders] Rao’s theory is that rather than building large stationary sensors for infrasound, drones could cover more ground. “My bigger idea was to use the fact that infrasound produced by tornadoes travels hundreds of miles,” he says. “Instead of waiting for it to come closer and then detect it using Doppler [radar], drones can fly outwards in all directions to offer an opportunity to reach out to a potential tornado, thus reducing the detection time and increasing the warning time.” (He calls his concept Revere, named after Paul Revere’s warning during the Revolutionary War.) There are challenges, including the fact that the noise from wind interferes with the sensors that detect infrasound. Rao argues that it’s possible to physically shield the sensor and then filter the signal. Altitude is another challenge, since the pressure would change as the drone flies, but Rao thinks that’s also surmountable. Ebling believes that a stationary network of sensors measuring infrasound is more likely. It would be cheaper to use than radar, and more accurate, so people could feel more confident that a warning wasn’t a false alarm. As the science advances, he says it could be feasible to build commercial networks of sensors as soon as a decade from now. Rao, meanwhile, is continuing to pursue his idea, while also working on an array of other inventions, from a sensor that could measure moisture in wounds to help avoid infection to a biomimetic surface for roads that could help ice melt faster. “I’m really interested in science, and I believe science can solve a lot of problems,” says Rao, who is a National STEM champion and was recently honored by a platform called Young Planet Leaders.
Category:
E-Commerce
Le Creuset turned 100, and to celebrate, it released a new cookware collection in a hue inspired by its original color, Flame. Named Flamme Dorée, French for “golden flame,” the set includes round and oval dutch ovens, a braiser, and a saucepan in a rich orange gradient with a shimmering gold finish. Prices for the Flamme Dorée pieces range from $310 to $860, which is more expensive than its standard linesespecially premium prices for what the company is positioning as a special occasion product. [Photo: Le Creuset] The French cookware brand was first known for its orange enameled cast iron cocottes, and it now sells products in a range of more than 200 colors. Last year it partnered with Pokémon and released its first ever out door cookware line. The original molten orange Flame color is its “signature color,” Le Creuset says, and Flamme Dorée is the modern remake. It recommends pairing the color with a deep rich green called Artichaut and White. “More than just a color, Flamme Dorée is a feeling,” Le Creuset says, like warmth or light. [Photo: Le Creuset] Le Creuset teased the collection’s release with a social media post of vintage print ads for the original orange Flame-colored sets. Few brands have been around long enough to have colors they’ve been associated with for 100 years, so by reimagining the color for a modern take, Le Creuset is tapping into its heritage to make something new. “Fiery, vibrant and globally recognizable, this celebratory hue pays tribute to the past while illuminating the path to the future,” it says.
Category:
E-Commerce
All news |
||||||||||||||||||
|
||||||||||||||||||