I had to download the app and launch it, only to find out that it only serves the SF/Bay Area. I don't live there. And even if I did, it seemed to require a Facebook login. Which I don't have.
This was fun to play with. It suggested an Indian restaurant so I tried saying "I don't like Indian food" and it followed up about whether I don't like it or I'm just not in the mood now. After suggesting another restaurant it correctly understood "Can you suggest something cheaper?" and returned a <$10 option. The "tell me more" option left a bit to be desired -- Luka just said "It's a good place."
Looking forward to when Luka "learns" about Oakland restaurants!
Thanks for your feedback - we're improving our restaurant data now (it's all being collected automatically). Oakland is next to come - we'll keep you updated!
While I kind of participated in the current AI/.ai hype, I'm starting to get really sick of hearing about another new application which "talks to you in natural language" or makes "Websites That Design Themselves".
The truth is, where not there yet and these people try to make one believe we are.
Sure, some may argue that these applications are indeed kind of intelligent and thus can be called A.I., but by this definition even Amazons recommendation algorithm is A.I.
I'm pretty sure that applications which "talk to you in natural language" or make "Websites That Design Themselves" are appealing to some at the moment but I doubt they will be successful in the long run.
How would you call it then? It's an app that mimics a friend that knows a lot about restaurants - I believe that can be called AI. However we used AI in the title just to make it easier to understand what we do. We sometimes just say it's a mobile app that gives you restaurant recommendations in a chat interface.
Things like this would more formally be called 'agents', software agents, or intelligent personal assistants (which are a type of agent).
However, I think AI is perfectly acceptable in common usage nowadays. Some people will just get caught up with the 'formal' definition, but the media and, increasingly, consumers are just using the term "AI" now anyway.
Just curious about your data sources, could you tell us what you are basing your recommendation on?
Are you using review sites (Facebook, Google, Yelp) or is it something completely different?
We use more than 15 different data sources - mainly user reviews (Yelp, 4Square, TripAdvisor), blogs and professional media (SF Gate, Eater etc) and guidebooks (Michelin, Zagat etc). All data is collected automatically and then plugged into our data model. We extract entities and make sentiment analysis to learn what this place is good for exactly and to show relevant quotes. As people share much more in a conversation than they would on sites like Yelp matching our users' preferences with the restaurant data makes our recommendations more relevant and personalised.
Too bad they didn't get the memo about the magic concept. The video is cute, but no way I'm installing this app without being able to try it first somehow.
SMS is going to be our next step - we started with texting in Russia actually. Our own app however let's us get a better UX for the users. Thanks for the tip!
Your next step should be making it available through Slack in some way. Lots of teams would love to be able to figure out where to eat by typing /lunch and then all the team members can chime in with any restrictions.