This is the creepiest usage of technology hidden as some sort of utilitarian project that I know of. Going to be fun to see when they try to go-to-market and they realize that people actually don't want to have their eyes scanned by some sort of "orb".
I think it's quite a stretch to say he was "powerful". Sure he had some influence but he was never on anyone's list of _somebodies_ in Silicon Valley. My personal opinion is that he really really really wants to be seen as a Steve Jobs / Elon Musk type character and he saw OpenAI as a great opportunity since he didn't have any of his own ideas.
Or AI will just get to the point that you will describe the Behavior you want in very simple language, maybe referencing certain known characters and it will do all the hard work for you.
Hey Andre. I've actually gotten a lot of great help talking to your co-founder Andrey on your discord channel and he helped me out a lot with making it work at our scale. Super happy with it and it's working in production with no hiccups.
Actually if you set it up right it doesn't cost you anything more to do it if it's almost fully parallel. It doesn't matter if you paid for one GPU instance for 500 hours or 50 for 10 hours. The cost is about the same. You can also more confidently use spot pricing to reduce the cost.
I do about 100 million embeddings using around 50 GPU instances and feed them into Qdrant. Takes about 12 hours. Very happy with the result and performance as long as you have the option to have a very large memory instance running.
If you don't mind unsolicited advice, I think your service could easily facilitate a paid "newsletter" of recommended new authors based on some sort of user profile.
Very interesting. Do you track any sort of demographic data, such as age, gender and location? Looking at the top authors I would wager millennials are the biggest users?
This happened to me with my iPhone using a QuadLock mount. The camera started just shaking uncontrollably. Luckily I got a new iPhone back from Apple since I told them I had no idea how it happened. Now I'm using a Vibration Dampener https://www.quadlockcase.eu/collections/accessories/products... that seems to work, but I had received no warning from QuadLock that their mount could potentially damage my phone beforehand. Now I see that they have a small warning when purchasing which is a lot better.
I’m actually surprised to hear that the standard approaches used for weather forecasts do not use AI. One would think this would be one of the first areas that would adopt AI given the data available.
Also that these new AI systems are performing at the same level of the old non-AI approaches I feel is a real testament to the developers of those systems. This is not something we see in many other cases.
Just curious, did you ever think about using Flutter for this instead of Swift? To me it seems like a pretty obvious choice given you are a one-man team and want to get it onto multiple platforms.
John Linneman from Digital Foundry made a great point about refresh rates (or framerates in his context). Lower refresh rates generally is perceived as way worse when you often switch around between high and low refresh rates. Case in point if you change from 60hz to 120hz and move your mouse etc. you will definitely feel that its way smoother, and 60hz would feel extremely sluggish. But if you just stuck with the 60hz for a while eventually you will perceive it as smooth again. They use the example of 30fps seemingly feeling very bad when you keep switching between Quality and Performance mode on console games, but 30fps is actually okay for many instances. Of course 60fps will feel smoother no matter what, but the point is is that the difference is heightened when you switch back and forth.
I think the idea is that it should be impossible for the client to know if it is a real or hallucinated player becomes from its point of view it is just a player that nobody else can see.
For a great (and entertaining!) look into how you can use simulations to generate these types of maps you should check out this video by the superb channel AlphaPhoenix: https://www.youtube.com/watch?v=Lq-Y7crQo44
A simple way to monetize tools like this is to create a queue. You specify the amount of requests or concurrent sessions that you feel comfortable with, and then you make it possible to pay to «skip» the queue. This way people that really want to try it out will be able to be patient and wait out the queue, and the impatient rich people can pay for the rest.
What this technology has shown is that is basically superhuman in the _amount_ of «expert level knowledge» it has. Not completely sure if we have said that what makes human special is the amount knowledge an individual can acquire. What we still see lacking, and I would wager will still pertain if we keep just training on the internet, is that it is not good at reasoning to generate _new_ knowledge. It is quite able to compare etc, or give examples but you can clearly see that its just in a way regurgitating some information with a twist.
Hopefully we can find a way to use this technology to enhance the parts that do make us human, whatever your definition of that is.
Which would not be contributing value to the company. I think his point was that a junior should learn that just working blindly like that is not always the right thing either.
Your point is certainly valid, but I have a slightly different perspective on the matter. Google Search Ads hold a certain magic that is unparalleled in the digital marketing world. Unlike video ads, which can be intrusive and irrelevant to a viewer's current needs, Google Search Ads are precisely targeted to solve the exact problem or question a user is currently facing. I've personally experimented with YouTube ads and discovered that while they can be useful for building brand awareness, they don't always lead to direct website traffic. As my company grows, I've realized that spending money on platforms like TikTok or Facebook simply doesn't make sense, as we can't yet promote our products in high enough volumes to truly stick in a user's memory. However, with Google Search Ads, we're able to reach potential customers at the exact moment when our product can be of service, leading to higher click-through rates and ultimately, a higher conversion rate than we could ever achieve with video ads.
Really fun game, was able to get the optimal solution of the first 37 levels, but level 38 really got me stumped! Been trying for maybe 10 minutes and can't even solve it no matter how many moves I do.
What would also be a fun exercise to write like a solver script where you basically try to brute force every puzzle. I'm sure you can get the optimal solution very quickly at least in these 4x4 grid games and only 2 balls.