¶1Hello, hope you are doing well. So today I wanted to talk a bit about my new startup that is called Sik Intelligence. [music] Sik means clear view in the Norwegian. So basically what Sik [music] is is it's trying to be an AI super forecaster company. At least that's our goal.
¶2So if you scroll down here, you can kind of see an example. So for example, Fondan and Putin meet in person before 2028. We kind of have our percentages of that happening and we have the market, right? So, I have a bunch of these fun examples you can go check out on my website. These are just made up, of course.
¶3Trump visits Greenland before the end of 2077. You can see the leaders of North and South Korea meet. You can go check out all these cool predictions I created on my website. Snow falls on Miami Beach before 2030. of wildfire forces evacuations in Paris region before 2030.
¶4[music] So I have a bunch of these if you want to go check them out. But basically what I wanted to do today was just to show you some demos of where we are at the moment and how [music] this product will look in the future when I am ready to launch. We do have a small weight list. I will talk a bit about that later. But before we go through the demos, uh last video I talked about kind of my how how well I've been doing lately by using actually sik on on prediction markets like Cali, Poly Market and especially some events like for example a 30 yield year US Treasury yield.
¶5You can see here we have a position at 500% return. We got in at 13 now it's at 78 58 97. So 65% that we have some history here. Uh here we got in at you can see,000% [music] 1,400% 700% for a $400 profit and basically after I've have tuned sik now for I think we've been doing this for like 6 months uh we have started to get some results lately. So that's why it's very exciting and I kind of wanted to do this video now.
¶6So basically, if you want to get in on early access, you can go to this and put you on the waiting list. There is a few spots open, so if you're fast, you might get in. But now, let me just show you a few demos of how this is work. How is this going to work when it kind of gets launched? So yeah, let's just do it.
¶7Okay, so what you see here is paste in any market. So now this is supporting both. We can do poly market events, we can do cali events. So this is just the best way to do it now. But in the future, I plan to have custom questions too that is not related to any to any prediction market.
¶8Right? So yeah, paste in any market. Let's see what we can find here. So if you go to Kali. So let's just see here.
¶9Uh we can try elections maybe or something like that. Let's do I don't know. Uh let me find something here. I don't know. Let's just do Texas Senate winner.
¶10These are quite close, right? They're like 6040. So yeah, you can see uh let's just grab this link here now. So we can go back here and I can paste this in and you can see we are reading the market. Yeah, we get this up.
¶11Texas Senate winner. This is 97 days out, but I think it's actually that's for resolution, but I think it's a bit um quicker. You can see it's in 34 days, November 3rd. It's just the resolution date you get here. And [snorts] you can see this is the market now.
¶125940. Yeah, that's correct. And now we can just do uh run six forecast about 15 minutes. So, let's just click on that. And you can see now we kind of get into this uh Texas Senate winner.
¶13You can see we have uh reading the market. That's done. And now we're going to launch up to 40 lookups. These are like tool calls we can do on the back end to gather data, research information, everything we need around this topic here, right? So we're going to do up to 40 like tool calls on that and then we're going to launch eight forecasters in parallel.
¶14So they're going to yeah try to look at the data we gave them and try to give us the forecast based on yeah basically all the research I have done on this the last months. Okay. And then we're going to wait six against the market and let's see what the results be. So now this is just going to run, right? So it's going to be about 15 minutes or something like that.
¶15So yeah, let's just let this run and I'm going to come back when we have the the outcome here. for the Friday show. I'm going to ask you girl, I just got to know [music] share a soda at the mall too. [singing] There ain't another girl in town for me. >> Okay, so now you can see we have our results.
¶16So our model put James Tal Tal Rico um 52% market is 59 on James. So if you scroll down here you can see we put Paxton on 47 and Tal Rico at 52. So actually closer than the market thinks. So we got like a plus seven here and a minus 7 here. Right?
¶17So that is our model. We can kind of look at the forecaster agreement. So the agreement between the forecasters were high. So that's pretty good. That's kind of good.
¶18It is outside our deep tested range so far. Uh and here we can kind of read the reasoning behind it if you want to do that. So why we kind of have a summary. Uh yeah, these are the Democratic or Republican nominees. Uh the number three general election has not occurred, right?
¶19Uh yeah, we can read about the summary here. We have some base rate information. These are probably from maybe from polling and yeah we can basically we get a lot of additional context here to why the agents uh selected what they did or the AI I guess and we can read about uncertainties and we can read about what drives the the selection or the the results and of course we get all the sources we pull the data from. So you can also double check all this. We also have some experimental models that I just follow just because this is a part of the research in the company.
¶20So you can see we are a bit difference on our experimental models. They are a bit closer to maybe the market I guess maybe here. Yeah. Uh I'm not going to reveal what those are at the moment but basically this is what my system does now. So basically we can take any event on uh if you go to demo we can take any event here and paste it in and we can get a outcome out.
¶21So let's just find something else. Let's just go to tech and science. Let's do AI. I don't know. Let's just find something here.
¶22Or let's just do poly market this time, right? Let's just let's let's check out poly market on tech. Uh, let's do AI and let's just pick something here. Let's just do this one. So, best Chinese AI company, end of October.
¶23You can see these are pretty close at the moment. You can see we have 35, 30, Xiaomi, Alibaba. Let's just copy that. And we can just paste this in here. Uh, okay.
¶24So, these are this is in our tested range. So, that's good for our prediction. And yeah, let's just run it again and see what kind of outcome we get. Okay, so we got it back. So you can see here we are a bit different.
¶25Here we can see we differ by 11 points. That's quite the big difference actually. So if you look at our prediction is 42% on Alibaba, but the market is more like 34% on Xiaomi. So this means that we could have like a plus seven edge here if you want to see it like that. Minus 11 here.
¶26if you trust siks or sik intelligence uh model right uh here we have low agreement but that could be because of the ladder system everything is not 100% yet so of course this is a demo not a final product and again we can read through everything here what drives it we can take a bit of a closer look here so for example September 11 snapshot placew at rank UB7 right GLM at nine something like that and yeah key uncertainty is and yeah, we have the evidence and everything we pulled to do this. So, you can see we could get some different results than the market here. And I'm still calibrating. I'm still trying to improve the product, but uh lately I've been having more and more better results. So, I think it was time to actually share uh what I've been working on for the last few months.
¶27And yeah, it's just a very interesting market. And I also wanted to mention that this market is kind of getting some some um attention from um the outside. Now for example, you can see AI startup mantic raises 25 million for superhuman forecasting. This is also like an AIdriven um super forecast company or startup. This is a bit older.
¶28This is like 2 three years old I think. And for the first time uh an AI won the metaculous summer cup. This is like the the biggest forecasting event that this was the first time an AI beat humans in the outcome here. And this kind of led to uh all the interest in investing money in this company. So yeah, this is what I've been working on.
¶29I want just wanted to share a bit about it and yeah, it's just been so much fun developing this. It's kind of my first big engineering task as a kind of like a solo um founder and I've enjoyed it really much. I've learned puns about engineering and stuff like that. And of course it's AI native. So everything I'm doing is based on uh is lined up for the company to be working with uh yeah language models and AI, right?
¶30So yeah, if you're interested, you can just go to the website. I'm going to leave a link in the description. Click on get early access and sign up for the wait list. There are, like I said, there are a few spots, so maybe uh you will get in. You just have to wait and see.
¶31But yeah, uh and also go check out all the fun um fun uh like fictionary events I put in here. I think they're really cool to read. So yeah, I hope this interesting and I'm going to keep you up to date. I'm going to do like a small devlog on how the startup is doing and maybe if I have some more interesting results uh like we had on the 30 year yield uh forecast uh I will bring you back for that. So yeah, thank you for listening a bit about my startup and I hope you enjoyed it and yeah, see you again