Close Ad

The Spoon

Daily news and analysis about the food tech revolution

bots

From Food Bot Discovery to Suggestions, Facebook Messenger Gets Smarter

From guidelines to making the perfect steak to numerous other conversational cooking applications, food-focused bots on Facebook Messenger have been advancing rapidly. Now, along with numerous other advancements, Facebook has launched new features that make discovering and trying new food bots much easier. At its F8 conference in April, the company announced new “Discover” features that would demystify finding an interacting with the best bots, and these features officially launched this week for users in the United States, along with new chat extensions that allow users to conversationally interact with sites such as OpenTable.

Messenger has more than a billion downloads on Google Play alone, and Facebook confirms that more than 1.2 billion users use Messenger each month. Its popularity is giving rise to an enthusiastic community of bot builders. The company has been adding chatbot  features within the nucleus of Messenger through its “M” AI assistant. We’ve written before about cooking with the Joule sous vide machine using Messenger, and new bots are taking conversational cooking in many directions.

So how do you discover the best bots and food oriented businesses and sites that you can interact with? On the lower right side of your Messenger home screen, touching a Discover button brings up categories that you can investigate, including one for Food & Drink. This lets you investigate organizations working with Messenger ranging from The Food Network to restaurants.

You can also investigate a “Featured” section to identify interesting new bots. A video from Facebook shows these features in action.

Bot developers and businesses can also now interact directly with users within Messenger chats. For example, you can reserve tables at restaurants conversationally via OpenTable. These features are enabled through Chat Extensions, which were also first announced at the F8 conference.

“Our goal with Discover is to ensure that experiences in Messenger are compelling, high quality and easy to find. This latest update makes it even more intuitive for people to find what they care about most,” writes Yingming Chen, a Facebook Messenger Engineer.

For developers and businesses interested in getting their bots or conversational cooking experiences added to the Discover section, submission instructions are found here.

Finally, Facebook has also made improvements to M, its AI-fueled personal assistant. As seen below, whether you are having a conversation with a person or a bot, M now intelligently suggests a “Save it for Later” option. If M senses that you are trading URLs for say, vegan recipes, it will automatically suggest that you save the links for later.  In addition to links M will suggest that you may want to save Facebook posts, videos and more.

CNET Founder’s Next Act Is AI Powered Publishing. His First Product? A Kitchen Assistant.

Update 7/7/17: The company contacted us upon publication of this post to emphasize the Tasted app/skill is still in development and not ready for consumer use. 

The cofounder of one of the Internet’s longest standing and most storied tech media brands – CNET – is onto his next act: creating a diversified media brand for the artificial intelligence age.

Shelby Bonnie, who cofounded CNET back in 1993 and later became its chairman and CEO, is the CEO of a new publishing startup called Pylon AI, a company which describes itself as a “conversational engagement platform company.”

What does that mean? From the looks of it, Pylon aims to create diversified lifestyle content that is delivered to consumers through AI centric conversation platforms such as Alexa or Google Home or bots such as Facebook Messenger or Slack.

In a way, the company that Bonnie and other CNET alumni Mike Tatum and Cliff Lyon are creating is reminiscent of Bonnie’s last company, Whiskey Media. Only this time, instead of a collection of different web-based lifestyle media brands, Pylon AI is using a combination of apps and AI platforms like Alexa and Cortana as the content publishing system.

One of Whiskey Media’s most popular brands was tech lifestyle-focused Tested, so now it’s not all that surprising that Pylon AI’s first consumer lifestyle brand is called – you guessed it –Tasted. As the name suggests, Tasted is all about food and comes in the form of voice-assistant apps such as the Tasted Alexa skill, a companion web or iOS app.

What’s intriguing for the smart kitchen crowd is Tasted is essentially a guided cooking system, using a combination of voice assistant, web apps and mobile apps like its newly launched iOS app to help guide the consumer through the creation of a meal.

Tasted uses Alexa and visual guidance to help users to cook

Another interesting aspect of Tasted is it employs the talents of well-known cooking personalities such as Catherine McCord, the creator of Weelicious, and Regan Cafiso, a former editor for Food Network and Martha Stewart. This idea of using popular cooking personalities is a standard playbook option to create buzz for a new platform, but what’s more intriguing is Tasted is another example of the nascent trend of established cooking talents such as Heston Blumenthal and Beth Moncel are embracing AI-centric cooking platforms to reach consumers.

A Pylon AI spokesperson told me that they are still operating in stealth mode, so the company isn’t talking about their forward-looking strategy, but my guess is that we’ll soon see other brands like Tasted in other lifestyle verticals.

For Bonnie, Pylon represents an intriguing new direction for a long-time media innovator. After creating one of the world’s most iconic tech media brands in CNET and a diversified web media brand in Whiskey, he is now looking to AI-powered conversation assistants like Alexa and Facebook Messenger as the next frontier to reach consumers.

Want to understand how AI will impact cooking and the food ecosystem? Come to the Smart Kitchen Summit. Use the discount code SPOON to get 25% off of tickets. 

Behind The Bot: Meet Sure, A Chatbot That Recommends Instagram Food Hotspots

While some people get downright grumpy when it comes to seeing food pics posted in their Facebook and Instagram feeds, I’m one of those that actually enjoys them. In fact, when I see someone showing off a tasty platter from a local restaurant on Instagram, I’ll often make a mental note to check that place out if it’s in my town or somewhere I plan on visiting soon.

If you use Instagram food posts as a restaurant discovery tool like me, I have good news: there’s now a bot that looks for Instagram hotspots and surfaces them in the form of restaurant recommendations. The chatbot is called Sure, and it’s a Facebook Messenger chatbot that curates the most Instagrammed food and drink spots in your neighborhood.

I interviewed the Juraj Pal, the CEO of Sure, to hear about how the idea for the bot came together.

Where did you get the idea for your bot?

Our motivation was simple. We weren’t satisfied with the existing restaurant discovery and travel apps and we quickly started believing that we can build a much better product for the next generation, already spending more time in messaging apps.

Having grown up with internet, we have soon learnt that virtually anything was accessible on our fingertips. It opened up a whole world of opportunities for us. But at the same, it made us feel overwhelmed with choices and options.

At first we actually started with a spreadsheet full of restaurants, cafes and bars that we curated ourselves. And to validate the idea, we launched a simple SMS bot where users would text our number and we would reply manually to each message, recommending a spot from our spreadsheet.

By tying visual social content to specific food locations, are you tapping into how you think this is how Millennials and others choose food?

We knew that others tried to solve this problem and the space is crowded with big players like Yelp or Foursquare. But we truly believe that for the new generation, they just don’t get it.

We quickly learned that millennials trust their friends and influencers more than reviews from strangers on Yelp. Rather than providing endless results like Google, we turned to Instagram as our primary source of all recommendations.

Why did you choose to use Facebook Messenger vs other platforms?

Other than having 1.2B monthly active users, Facebook Messenger is inherently social which makes it easy for people to share with their friends.

The social aspect has however been important also from another angle. When it comes to choosing a place to eat out, majority of the people ask their friends or influencers who they can relate to. By being on a platform where our users naturally chat with their friends, chatbot has the potential to blur the lines between tech algorithms and word of mouth recommendations.

What is unique about developing for a chatbot vs. other AI platforms?

We’ve seen different roles evolve as we were building the chatbot. We for example spend much more time on copywriting and building the bot’s persona and empathy than designing flows.

Building chatbots also costs less and happens much faster. And this in turn allows us to ship our product faster and iterate based on feedback we get. And what I love the most about this experimentation is that we’re focused on value delivery, rather than building potentially useless product features.

Why a food-focused bot?

We decided to start by answering the ‘Where shall we eat?’ question once and for all. Food is a highly personal choice that represents who we are and ties us with a community. Also, in the digitalised world we live in today, eating out is one of the few experiences that we cannot replicate online.

And since people are used to asking their friends for food recommendations on Messenger, we though we could be that concierge friend for everything when you’re out an about.

What have you learned since people have started using your bot?

A lot! There are literally 1000 ways how a person can ask for a restaurant recommendation and trying to support that with natural language processing is hard. Based on this we decided to switch to more pre-defined text and using more GUI elements.

Also, people love to test the boundaries of a bot and eventually they want to break it. Hence it’s equally important to educate our users how to talk to a bot, as it is building a responsive bot.

Finally, the speed in which we were able to capture learnings and improve the experience based on real usage was incredible. As opposed to cross-platform app development, we can instantly ship updates to all of our users without any disturbance.

Tell us a little about yourself – is this your first bot? 

I’ve founded and sold a startup in the food tech space in the past but this is my first chatbot.

What do you have in store for Sure?

After we launched on Product Hunt and expanded Sure to 22 cities around the world 2 months ago, we joined the Just Eat food tech accelerator in London.

We’re currently working on bringing the Sure experience into group chats and making it even more contextual as we grow to become the ultimate concierge for everything when you’re out and about – from choosing a restaurant to ordering an Uber to get home from a bar.

Tell us about your recently launched Sure extension for Messenger

We know that the way we discover restaurants is often by asking one of our friends. But many times, choosing a place to go with your friends can turn into a frustrating argument. This was the main motivation for launching our latest feature, the Messenger Chat Extension. This allows all our user to take the Sure bot with them to any group chat on Messenger and instantly share our recommendations with their friends. With this, we’re hoping to put an end to the “Where should we eat?” ordeal.

I Made Steak With Facebook Messenger. Here’s How It Went

We know that over half of Echos end up in the kitchen, making Amazon’s voice assistant Alexa a good option for those looking for a new-fangled way to help make food.

But what about Facebook Messenger? While we don’t have exact numbers on how many use Facebook’s communication app while in the kitchen, with over a billion downloads of the app in Google Play Store alone, my guess would be a lot.

Still, that doesn’t mean we think of Messenger as an interface to, well, our steak, but that’s exactly what ChefSteps thought when they announced they’d created a Facebook Messenger bot for the Joule.

I’d used Alexa in the past to cook with my Joule, and it worked well for things like starting a cook and checking water temperature, but I wanted to see how cooking with Facebook would go and to see if a bot of the non-voice variety was useful when preparing the nightly meal.

Here’s how it went.

First I went to the ChefSteps support page for using Messenger and tried to talk with my Joule, which I had inserted into the water with a nice ribeye. I was told I would first need to log into my ChefSteps account. Fair enough.

Once logged in, the ChefSteps Chatbot, which we’ll call Joule-bot for this post, reminded me of what I’d named it and gave a few clues about what it could do.

I decided to jump right and tell Joule-bot what I wanted to cook a steak. I got a pop-up message telling me a little about sous vide complete with a visual guide to doneness (a big focus for ChefSteps overall with their guided cooking approach for the Joule app).

As you can see above, the tone of the bot is casual but also informative. I like the ability to choose the length of cook with their visual doneness guide. This is an advantage over cooking with Alexa which (obviously) can’t show you how what a cook will look like as a voice bot.

Once I chose medium rare (you didn’t think I wanted a Trump Cook did you?), Joule-bot asked me a few more questions to understand how to go about cooking my ribeye.

 

Once it knew I was cooking fresh and how thick the steak was, it was able to set the temperature. As you can see, I had already started the Joule (with Alexa – meaning I technically had a battle of the bots over my evening meal), so it told me, in essence, my water was running a bit hot. The Joule, like other sous vide circulators, can adjust down as it lets the water cook and will then hold the temperature, which is what happened for my cook.

You can also see that Joule-bot told me that that it is still young and hasn’t fully matured, meaning it wouldn’t be able to send me notifications in Messenger about when things were done. This is where Joule’s native app has an advantage over the Joule-bot.

I decided I wasn’t done with Joule-bot, since I wanted to see if it could help me out with my ribeye prep and post-cook. I decided to ask it a few more questions and see how it responded.

When I asked it how to prepare steak, wondering if I could surface some of the same types of information that Joule app does with its cooking guides. While it didn’t give me the same, concise cooking guide I get within the Joule, it did give me a link which provides access to much of the same information on the ChefSteps website.

My next message confused Joule-bot a bit, mostly because I think of my language choice. I was trying to get Joule-bot to tell me something it had already done (2 hours of cook time) with a specific question about that. Instead, it guessed that I was trying to see when my Joule would ship by surfacing an FAQ question.

While the logic wasn’t perfect, I think the response was fine. Since Joule-bot lets the user give feedback, this will help refine the bot’s logic over time. It also gave me lots of options of what to do next, with links to the ChefSteps community forum, recipes and also the option to file a support ticket.

Conclusion

Overall, cooking with Facebook Messenger was an interesting – but for now limited – experience. Joule-bot allowed me to set temperature based on visual guidance, told me in a conversational voice when my meal would be done, and directed me to the information rich ChefSteps website when it didn’t have the answers.

What it didn’t do was provide notifications, a big difference which gives the Joule app an advantage for now.  Joule-bot also didn’t have the richness of information provided by the in-app cooking guides (though, as mentioned, it did send me links to the ChefSteps website).

Compared to Alexa, Joule-bot has an advantage in the type of the information it can provide, such as visual guides around doneness.  However, Alexa commands are just a little easier (what’s easier than talking?) and I could see how Alexa would be preferred over Joule-bot when I’m preparing food with my hands.

Lastly, it’s important to ask the question: is cooking with Facebook Messenger a good idea?For now, I would say the Joule app is a better experience, but over time a bot could have some advantages. Messenger’s conversation logic is very good, and those used to using chat as a way to interface with people may also find it also a good way to control their things (like the Joule). I also think as many of us tire of apps for every device, Messenger is a logical candidate to become that universal app, especially as bots become better.

Why don’t you subscribe to our free weekly newsletter to get great analysis like this in your inbox?