ai.bythebay.io: Lukas Biewald Interview

[Music] I guess it seems like we keep redefining I guess it seems like we keep redefining AI to be things that computers can't do AI to be things that computers can't do that seems like the general definition that seems like the general definition of AI guess to me artificial of AI guess to me artificial intelligence is about making computers intelligence is about making computers think in ways that humans can think think in ways that humans can think making them as smart as as humans are making them as smart as as humans are but specific to our work it's less but specific to our work it's less around making machines think like humans around making machines think like humans and more around helping computers take and more around helping computers take care of business processes that are care of business processes that are important to our customers so simple important to our customers so simple things like back office processing or things like back office processing or identifying what's in an image that's identifying what's in an image that's the kind of AI that crowd Florida's well the kind of AI that crowd Florida's well I think I think AI is amazing at doing I think I think AI is amazing at doing things where it can find lots and lots things where it can find lots and lots of training data so in cases where it's of training data so in cases where it's maybe backing up so AI is is fantastic maybe backing up so AI is is fantastic when it can look at lots and lots of when it can look at lots and lots of examples because it's nowhere near as examples because it's nowhere near as smart as a human but it can have way smart as a human but it can have way more patience than a human can write so more patience than a human can write so it can look at so many more things in it can look at so many more things in human can so cases where there's lots of human can so cases where there's lots of training data it's gonna be really training data it's gonna be really successful like for example Facebook's successful like for example Facebook's face recognition is fantastic right face recognition is fantastic right because you know they get billions of because you know they get billions of selfies uploaded every day right or selfies uploaded every day right or Google Translate has is it's quite quite Google Translate has is it's quite quite good not because the AI deeply good not because the AI deeply understands the text but because Google understands the text but because Google has so many examples of documents in has so many examples of documents in language 1 and language to that that language 1 and language to that that they're able to see so many edge cases they're able to see so many edge cases and situations to make the translation and situations to make the translation work well I think where it's trouble work well I think where it's trouble struggles is cases where it's really struggles is cases where it's really hard to get training data right so it's hard to get training data right so it's it's phenomenally hard for robots to it's phenomenally hard for robots to grab things right I mean amazing how grab things right I mean amazing how what a challenge that is and I think what a challenge that is and I think that's because there's not a lot of easy that's because there's not a lot of easy ways that we know now to generate ways that we know now to generate training data around grasping training data around grasping [Music] sure I mean so so my company CrowdFlower sure I mean so so my company CrowdFlower is key focus is something I've always is key focus is something I've always cared a ton about which is making AI cared a ton about which is making AI actually useful for for real companies actually useful for for real companies in the real world and I think when you in the real world and I think when you look at when you look at the opportunity look at when you look at the opportunity of AI I think every company can benefit of AI I think every company can benefit from it there's there's so many places from it there's there's so many places where we have humans doing things that where we have humans doing things that you know almost like a simple pattern you know almost like a simple pattern matching algorithm would actually make matching algorithm would actually make things much more effective right and so things much more effective right and so you know if you look at why companies you know if you look at why companies struggle with it I actually think the struggle with it I actually think the algorithm is typically the least of algorithm is typically the least of people's problems right like everybody people's problems right like everybody kind of focuses on like oh is it you kind of focuses on like oh is it you know deep learning or SVM's or logistic know deep learning or SVM's or logistic regression and how do you do it but you regression and how do you do it but you know I think that the part that I'm know I think that the part that I'm really interested in because I think really interested in because I think it's the most difficult pieces you know it's the most difficult pieces you know how do you actually train these how do you actually train these algorithms how do you do feature algorithms how do you do feature selection and how do you actually deploy selection and how do you actually deploy them in the real world so that they them in the real world so that they actually get used actually get used [Music] [Music] well one thing I really admire about the well one thing I really admire about the the the by the bay conference is the the the by the bay conference is the authenticity of people that it brings authenticity of people that it brings out so it seems like that it's really by out so it seems like that it's really by the bay does an excellent job of getting the bay does an excellent job of getting getting speakers that AI professionals getting speakers that AI professionals actually really expect it really respect actually really expect it really respect right so you know I think a lot of right so you know I think a lot of conferences they kind of just look for conferences they kind of just look for the most famous people to a general the most famous people to a general audience but you can tell that AI by the audience but you can tell that AI by the bay is is made by people that really bay is is made by people that really understand artificial intelligence and understand artificial intelligence and are deep inside it and so when you look are deep inside it and so when you look at the speaker list it's people that I'm at the speaker list it's people that I'm personally super excited to hear talk personally super excited to hear talk [Music]