 I mean, and as I said, so we did try to make it a neural network proper neural network, but that didn't improve either over the simple logistic regression model that we have. So, our conclusions from story one were that the referential attributes of entities can be inferred to some extent from web of words vectors with very simple supervised methods, but the results differ substantially by attributes and attribute groups, and there is the crucial role, the question of, you know, how well can you predict the value based on just the words and the context. And so, this then gives, of course, two possible follow-ups. You can either then do proper entity training where you don't just take, you know, vectors from some free-analyzed corpus, but you essentially learn the word representations as required by the test, and then these word representations probably would try and capture exactly those more specific kind of context patterns that I talked about before with respect to the population. Also, you know, I think imagine quite nicely what they would look like for attributes like, I don't know, like the foundation here or something of a country, then you would want to learn something like this. What's found in X would be a very, very good idea for that. Maybe you don't even have to do it in an active way, but maybe it might actually be sufficient to use richer word-vectors that take more linguistic structure into account. You can tell, okay, this is a verb in the sentence, this is the subject, this is the object, and they contribute in different ways to the representation. Okay, so these are ideas, we haven't done that so far, and I'm just kind of wondering, I've been talking for an hour now and I don't want to bore you to death. I mean, I have slides now on that second study that we did, but I guess I would be talking for another half an hour or so. So I can also just call it a day now, so what do people think? So probably individually and collectively we're in a difficult position to answer that question. I mean, I'll just say as an organizer, we typically do meet for about an hour, although I know that no one was put up to the task of asking you questions every two minutes, but we've successfully accomplished that, which I'm sure Elon dated the time you expected to spend on the first study, but we're informal enough here that I think I can ask, how is everyone's time, and how about this then, is can I find a middle ground? Can you give us a digest of study number two enough for people to appreciate what you've done, even if we skip over a lot of the details in just a few minutes? I can try to do this. I mean, the other alternative is of course that we just, you know, I stop here and I'll be happy if somebody is interested in the additional stuff to just, you know, do it in a smaller room to go through within 14 hours. So yeah, we could take a break and restart in five minutes or something. Yeah, if there are, and I know, again, I don't know anyone scheduled here, but hey, it's an informal group and we're, so thank you for the offer, Sebastian. How do people sound with Adam's suggestion, a few minute break, and then are some people interested in coming back and those of you who maybe have plans or other responsibilities and you have to go, who would be interested in rejoining after a few minutes and hearing. So we have a handful, and that's great, and if it's really a handful, we could even, because I also don't even know about the room reservation. Usually they reserve for an hour and a half, but it might actually be, we might only have about 20 minutes. So our suite, we have an engineering suite, right, if you leave this exit this door and turn right, which is at the end of the hall is the glass doors. And so how about we sort of take either a break or call it a day for those of us to go, but in five minutes we'll reconvene in the suite, and we have those small conference from there which will easily accommodate even half of this group. So thank you, Sebastian. Leave this door, it's right in front of you, the big glass door is on the right. So don't forget any of your stuff. Great plan. I just want to introduce myself. How are you? Good. Thanks for making it happen. No, no, no. I didn't do anything like that. Okay. Great. How are you? Good, good. Yeah, all right. Thank you. Okay. See you in the next session. Okay. No, no, no. Okay. So yeah, and then we went back to that area. I'm asking you because Sunday. This is my dialysis. Well, I'm not asking for a selfish reason. So DAD is organizing. In fact, let me tell you this too, this is what I want to tell you as we move. But this is like all about me and maybe this should be all about your work. Yeah. And, and, and Sebastian, both. I'll start that. I want you to do both in NLP. Yeah. But maybe you also want to break it. I don't see it. So I was going to follow up with you about that. You know about that, right? Yes. Yes. I'll read them later. There's basically the key to the question, you know, which university is the most