Write Beneficiary Image with airSlate SignNow

Write beneficiary image

hi this is an interview with people from lyon whose flagship projects are data sets specifically data sets to train models like dali or clip so pictures and text that goes along with the pictures they scrape these from big internet scrapes the first data set had 400 million images and their newest data set has 5 billion images these are unprecedented scales to be open sourced as data sets the creators of dali or clip openai they never disclose their data set they never put it out there in public and lion does so this is a big service to the community and i was super excited to have them on here another thing is just how grassroots this movement is the founder christoph who's also here today is a father and a teacher and does this on the side just as a hobby and sort of wants to demonstrate a little bit how anyone can take part in open source research now multiple times during the interview his kids would actually come in and be like daddy play with us and so on youtube is very strict on this i cannot show the kids even though the kids themselves would have loved to appear in this youtube video so you know kids please um i'm very sorry it took me a zelit when you are inside open invitation often an iron laden feeling channel i thought this was really cool and inspiring in addition to learning what lion is about enjoy the interview let's dive right in hey everyone today i have the team behind lion5b with me christoph schumann roma bomo and kate gordon are here who contributed to this project in various ways which i hope they'll just tell us about in a second this is a giant data set it's over 5 billion image text pairs so not just images but image text pairs and along with that an open clip model open sourced clip model that matches the performance of open ai's clip model which is really cool these these big companies they rarely give out their biggest models if any if you know if at all and if they give out their biggest models they usually don't give the data set behind the model so it's really cool that we have like a large data set uh there has been some controversy around your smaller data set that you released i want to say half a year or a year ago i hope we can get into all of that today but first of all uh thank you very much for being here welcome to the channel welcome nice to be here yeah um just maybe tell me a little bit uh what is lion and what is lion 5b so it all started like 10 months ago i guess on the elute ai server when we talked about how could we eventually replicate dali and where could we get like 200 300...