>> alora, alora. the thing is i wanted to take a picture all together.huh? >> and so i -- it -- i mean -- there was -- an ice moment. you know when everything freezes? >> jimmy: uh-huh? >> everybody was frozen. me, i was behind the table with this cake, waiting for people -- "hey, guys, let's have a picture all together." [ laughter ] and no one -- >> jimmy: no one would get in the photo with this. [ laughter ] so then there's this. [ applause ] a little distressed there. well, you're learning. you know, you're learning our ways here. >> no, it's unbelievable. and then -- so jon that is like my angel, jon, he was the only brave, brave one. and he came next to me and he said, "sabrina, no one will ever make a picture with you with this cake." [ laughter ] "this is a racist cake." [ laughter ] a racist? why racist? >> jimmy: yeah. >> my gosh. >> jimmy: you know, you get a pass, don't worry about it. [ laughter ] if kanye had this cake at his birthday, we'd have a problem. [ laughter and applause ] it's so great to meet you. sabrina impacciatore, everybody.

we love all baby alora appearances. annie, what can we do to stop it?bye, dr. patel, come later. what can be done to stop it? annie: it's a great question and it is evolving as the technology evolves, which is extremely fast , making it difficult to come up with a solution to this. i think there needs to be a technology solution in tandem to a policy solution. there is a need for better deepfake detection, automating the techniques that are used at the pixel level, instead of just initial cursory view. or from the policy standpoint, there really is i think a need for a new classification around what defines a deepfake as well because part of the problem we are facing is certain mediums, different types of social media varying in definition and therefore there is a huge difference between what's allowed and what's not allowed to be posted and what is being shared with the public. some standardization could also help. it's a technology and policy solution in tandem, which can be difficult as the technology advances with ai, as we have seen in the past week with chat gb t, but hopef