15yrs back I participated in "Google Ants AI Challenge 2011", an ai programming competition, hosted by the University of Waterloo, and I ranked #127 (#1 in my country). The competition gave me a huge learning oppurtunity where developers across the world came to a forum and discussed various techniques.
Now, I've built a similar platform to bring back the fun of building a small neural network that can play the game well. Neural Network optimization seems to be much more fun.
Plz share your feedback to improve the platform and add more games.
Unfortunately the competition site is mostly down and couldn't find much about the old competition other than the participants blog articles. Do a search on "Google Ants AI Challenge - post mortem", and you get a lot of articles around the game.
Thanks for sharing the research. I tried implementing a per Ant decision making model, but gave up as the training time was much longer compared to the current baseline. I think I should rethink the idea.
Nice. I was 72nd. Working in AI research today and still making ai for games as a hobby (tfmbot.com is an ai i’m working on for my favourite board game terraforming mars).
From competing with the world to building a place for the world to compete—what a full-circle moment. Love the challenge: how much strategy can a tiny neural network learn? Excited to see what people build.
Am wondering why no one is talking about traceability of Photos. Ex: CSAM which Apple was fighting for a long time. I thought this feature was the answer to provide proof of who shot the picture.
According to the description in the article, "an outside observer cannot determine whether any pair of reference images were taken by the same device." and they "avoid even implicit public association between different photos taken by the same sensor"
I dont think this is relavent to that use case. It seems like this proposal would be an optional off by default feature. They also seem to be going to lengths to make it privacy first so the verified photos cannot be linked to a specific photographer.
Nothing stops you not verifying, or simply stripping the verification off.
What you are prevented from doing is adding a verification to a photo outside of the iOS image pipeline, or modifying the photo with the verification still in tact.
1. Can't remember which way you came from just after a few mins of passing the same street.
2. Can't recollect your mother's face by closing your eye, but can recognize when you see.
3. Don't remember faces unless seen them for many times.
4. Can remember a favorite color.
5. Don't have a consistent handwriting.
Recent language Claude uses is very heavily loaded and hurts me when reading. So I had to use this to keep my sanity. This started only with Opus 5 where Anthropic thought they optimized the meaning per token and make it very hard to read. Comments and code also looks like cryptic.
He is not hiding that he was wrong. So am fine with that, as my bet is based on reading many of such predictions and take a conclusion for myself. I take these as are opinions backed with some observations.
Agree with this approach. Am building tools around the same concept
80% solid base and rest of for the custom business logic. https://github.com/GoPlasmatic/Orion
Typing code by hand is dead I agree. But calling it the end of programming was bit weird though.
I use to write a lot of code everyday for more than 2 decades, but now with LLMs, I can get most of my code written by LLMs, and I make sure they write the code that I want to. Same as, when I had to manage a team of juniors.
Had a similar experience with my Rust implementation for JSONLogic expression evaluation engine. As it has a full test suite with 1000s of cases and a benchmarking script, I was able to give some basic hints to try different optimization techniques and the end result was impressive. Reached from 1.6s to 200ms for a full benchmarking test.
https://github.com/GoPlasmatic/datalogic-rs
First 3 versions were hand written and maintained for 3yrs, and now 4th version came out in less than a month's time with impressive performance.
This is really nice. It's _almost_ useful for identifying features from where one stands, if it had another three levels of zoom and another three levels of voxel resolution.
I tried adding more resolutions, but had to balance it against cost of hosting. Each resolution is at least 2x the size of the combined all prev zoom level data. Currently I’ve 9 zoom levels total size 15gb, adding 1 more level takes it to 32gb but looks beautiful. With the current open data, I can generate upto 14 levels
Just in case you are interested, you can recreate this locally using the open source repo. The script downloads the open data and generate the voxels based on configuration. The max resolution is configured which you can increase. Just a few prompts and you can recreate high resolution upto 10m resolution in real world distance. :)
Now, I've built a similar platform to bring back the fun of building a small neural network that can play the game well. Neural Network optimization seems to be much more fun.
Plz share your feedback to improve the platform and add more games.