For those who’ve ever tried to log your meals consumption with an app, you most likely have realized the next:
- Guide meals logging with an app is a ache.
- Meals logging apps are sometimes inaccurate as a result of they require customers to estimate portion measurement, elements, and so on.
- Meals logging apps have cumbersome onboarding processes, ask for lots of non-public information, and often attempt to upsell customers into premium subscriptions.
For these causes, I’ve not used a food-logging app for quite a lot of weeks, however that will quickly change with GPT Meals Cam. What’s GPT Meals Cam? In just a few phrases, it’s a free food-logging app that allows you to snap every meal or snack together with your smartphone digital camera and makes use of AI to estimate energy. The app, which might be downloaded from the iOS App Retailer, doesn’t ask you to take a survey or require a subscription, and, from what I can inform inside a day of use, it’s actually fairly darn good.
That’s my description, however what does Raj Singh, the longtime entrepreneur who’s the visionary behind the app, should say about it? In keeping with Singh, who posted not too long ago concerning the app on LinkedIn, GPT Meals Cam is totally different from different meals logging apps in three main methods:
Instantaneous Digicam Entry: The app opens straight to the digital camera, permitting customers to rapidly seize their meals with out navigating menus. “I wished it to be quick and low friction,” Singh stated. “In social settings, it’s much less intrusive to rapidly snap a photograph.”
Calorie Ranges As an alternative of Actual Figures: Singh stated that as a result of AI has its personal limitations and portion sizes range, the app gives a calorie vary. “By presenting a spread, it’s largely proper,” Singh stated. “The objective is to construct the behavior of meals logging and turn out to be a extra aware eater.”
Free and Unobtrusive: In contrast to many apps that require subscriptions or bombard customers with advertisements, GPT Meals Cam is solely free and supported by occasional, non-intrusive ads. “Proper now, advertisements are making 4 occasions the income of the AI prices,” Singh stated in a cellphone interview with The Spoon. “This permits us to maintain the app free and doubtlessly develop its options and availability to extra nations.”
After working with a meals coach who inspired him to ship photographs of his meals for suggestions, Singh sought a handy digital answer to proceed the follow. Nevertheless, he discovered current apps missing—both too complicated, pricey, or each.
“They had been designed for the 5% who want precision, however I wished one thing easy, free, and for the opposite 95%,” Singh stated.
In keeping with Singh, GPT Meals Cam leverages Gemini Flash, a quick and cost-effective AI mannequin, to investigate photographs and estimate calorie content material. Customers merely snap a photograph of their meal, and the app processes the picture to offer an approximate calorie vary.
“Plenty of that is immediate engineering,” Singh defined. “We use ‘chain-of-thought’ prompting, the place we break down the AI’s job into particular steps. The immediate instructs the AI to have a look at what’s within the image, think about every ingredient independently, estimate serving sizes based mostly on context—like whether or not it’s in a bowl or on a plate—after which estimate the energy of every merchandise earlier than including them up.”
Singh emphasised that whereas AI isn’t excellent—with about 95% accuracy—it’s adequate for selling aware consuming. “AI has persistently been 95% correct,” he stated. “It’s nice for suggestions and recommendations, however on the subject of important workflows, it’d get issues flawed 5% of the time. For meals logging, the place precision isn’t as important, this degree of accuracy is appropriate.”
The creation of GPT Meals Cam got here after a serendipitous dialog with a good friend. Singh’s good friend, Zvika Ashkenazi, talked about that his son, Ben Ashkenazi, was searching for an unpaid summer season internship and puzzled if Singh might mentor him. Singh quickly started working with Ben, and 6 weeks later, GPT FoodCam was born.
“Ben is graduating from ASU in Pc Science in December,” Singh stated. “He taught himself React, iOS improvement, and extra this summer season with minimal assist from my community. He constructed this end-to-end.”
Whereas GPT Meals Cam emerged simply within the final couple of months after Singh’s epiphany and Ben Ashkenazi’s coding work, Singh has been toying with the concept of a low-friction app to trace meals consumption for a decade. In 2009, he tried to develop an analogous software however quickly realized the expertise wasn’t mature sufficient.
“In 2009, I attempted to create this actual app,” stated Singh, who’s presently the pinnacle of product for Mozilla’s Solo after the browser firm acquired his startup Pulse in 2022. “It wasn’t ok, and so we pivoted right into a recipe firm, which turned Allthecooks.” Allthecooks would go on to turn out to be the primary recipe neighborhood on Android in 2010, with 30 million customers, and would later be acquired by Cookpad.
In contrast to then, “the tech is now right here, making GPT Meals Cam a actuality,” Singh stated. “Developments in AI and picture recognition have lastly caught up with the imaginative and prescient I had over a decade in the past.”
With the expertise to make friction-free meals logging a actuality, Singh informed me he desires to disrupt the meals logging trade by providing a free, low-friction app, however he thinks it may accomplish that with little involvement from him going ahead.
“I construct some issues for enjoyable. On the onset of a brand new venture, I’m like, ‘This isn’t gonna generate income, however the world wants it,’ or, ‘That is gonna be my subsequent enterprise, and I’m leaving the place I’m at.’”
Singh made it clear he’s completely satisfied at Mozilla and, in reality, used the product he conceived of constructing for Mozilla (Solo, an AI web site builder) to create the web site for GPT Meals Cam. From right here, he’ll let Ashkenazi run with the product, even when he periodically suggests some concepts to make it profitable.
“I feel it may be very, very disruptive. Individuals are paying $10 a month for apps they don’t have to. This app can encourage higher habits with out the price and complexity.”
Singh stated he’s additionally contemplating increasing the app’s capabilities and attain. With the advert income already exceeding the AI prices by a four-to-one margin, there’s potential to extend day by day utilization limits (presently, customers are restricted to 6 snaps a day) and make the app accessible in additional nations.
Selfishly, I hope he and Ashkenazi succeed as a result of, from what I’ve seen thus far, I feel the app is, in reality, doubtlessly disruptive, and I hope to maintain on utilizing it. Who is aware of, possibly Ashkenazi (with slightly assist from Singh) can put their app on an analogous journey we noticed with Marco Arment’s Overcast app, which initially was a ardour venture that emerged from Arment’s annoyance with the present state of podcast apps to turn out to be probably the most user-friendly podcast app (and hottest, outdoors of Apple’s podcast app) on the earth.
