A-Train

Amen.

I'm a Data Scientist. For some time, I've been working on a library for feature engineering. • GitHub: https://github.com/feature-express/feature-express • Website: https://feature.express It isn't yet complete, and I wouldn't consider it ready for production use or handling larger datasets. Here are some of its characteristics: • Event-based workflows: Initially, everything is converted to an event format, ingested into an event store, and processed from there. • In-memory: Both the event store and evaluation have been built from scratch. • Written in Rust, but there's a Python package available. • A DSL (Domain Specific Language) for defining aggregations, similar to SQL. Why am I developing this? I've always found it challenging to build models based on time. These models can be surprisingly tricky, and there's a high risk of accidentally using future data, which can lead to data leakage. FeatureExpress is designed to nearly eliminate such mistakes. Moreover, I believe that representing data as events is an intuitive approach.

The architecture is something like an ensemble but there is also this control network which chooses 2 experts to generate text.

Right now I start my prompts to chatgpt-4 with "dont be lazy". For every question I have it answers that it is a complex problem... GPT-3.5 in the API is more consistent that chatgpt-4. Even with some additional prompts it is making so many mistakes that it takes me multiple tries to get the right output with conversation resets from time to time to start from the previous solution.

That feels like a great idea to mix to strong models like these in a sequence.

I'm pretty surprised more people dont use logit biases to call openai with. Checking if something is either a or b means that the tokens for those letters must be 100 weight which means they will be chosen no matter what and no other character is allowed.

You can build procedures for fact checking quite easily. It is like asking gpt to write code and tests.

Ah nostalgia... This was my first PC processor. I remember my CPU came with some sort of a bug that prevented me running windows 3.11 on it.

Not sure it you know Terry Pratchett and his discworld series but this was his exact train of thought.

http://www.telegraph.co.uk/culture/books/authorinterviews/10...

Except Paul McCartney is considered as one of the best song writers of all time. There may be many musicians who can do the same on a technical level but even then he is a significant outlier.

It is expensive but even then Youtube has been profitable for many years now. It represents 10% of Google revenue. 5-10 ads is a cash grab imho.

He liked Top Gun Maverick very much too. The review was fantastic.

Except it didnt. Dall-e and copilot are both microsoft because microsoft bought openai.

Google created imagen which is better than dall-e.

Not sure if it is more relevant when you immediately hear about increased radiation from the power plant.

You realize that Scikit-learn is written mostly in Cython (where high performance is needed)? It is a part of the most influential ML library in existence.

If you believe Rust zero cost abstraction - being safe doesn't mean it has to be compared to safe languages because Rust and C can translate into same assembler code. But in Rust the compiler gives you more guarantees.

I would say that a proper way to rephrase it would be to use idiomatic code vs unidiomatic. If you start using any high-level language in a non intended fashion (writing non idiomatic code) probably most of them would be faster.

To some extent it depends on the settings of the tournament because engines can work in the background and analyse position while it is not their turn. It is called ponder.

Time management is based on some heuristics built into the engine. Some positions that are being evaluated are more dynamic and engines have rules to evaluate positions which are settled down.

As for the hardware if 2 CPU engines are playing they have exactly the same resources. The problem arises when one engine is GPU based and the other is CPU. In this situations balancing the compute power is hard but even then they normally have the same time allocated.

Yes you are right it is not a very good UX. For me the most interesting part is observing the evaluations of the position from the perspective of 2 engines. See the charts that track it move by move. It is often the case that one engine sees something that other doesn't. For example, say, that it is completely busted. I imagine internal dialogue: I'm fine, I'm fine, am I in trouble... oh $^#^ I'm losing now. It was especially true in the early days of neural chess engines which saw ideas which were well beyond the event horizon of traditional engines. Most people who are watching these tournaments on twitch are chess engine developers themselves that's why interfaces like these are fine for them.

Just look at new cases and deaths in South Africa and compare to the previous waves. Seems pretty convincing to me.

How exactly increased elo makes lichess worse? I would only argue that because of the number of players chess.com it seems more stable. I really feel every 50 of elo difference.

Agree. Even yesterday I was praising Blender's node editor and I'm encouraging the developers to mimic its functionalities in a product we are building.

For traffic simulation there is https://github.com/a-b-street/abstreet which is very impressive project.

And I disagree with you. The police officer is not doing anything wrong he is only utilizing a strange loophole. Is an actor wearing a reflective hat is breaking law by preventing paparazzis of using the photos in tabloids? Why should they make it easy to post the video on youtube which is a commercial platform that has nothing to do with any kind of law. The video of the police officer can be manipulated too to show limited context. If he was breaking the law the video can be used in the court no problem. They can also write an article or describe it on twitter.

I think the issue is that people are entitled and intoxicated by their "freedom" and publicity. It is not enough if what people say is consumed it must be mass consumed. The freedom of mass publicity is limited by DMCA takedown.

You should be grateful that someone else did it because it is a bad idea. Everyone of those services is easily replaceable with a library call. For example there is no language in the top 20 Tiobe index that doesn't implement all of this stuff already: image resize, key value store, id generation. And also you should not split your system into microservices like that.

Not necessarily true that MLPs are very compute expensive. It is maximum a couple of layers and if your input is sparse (categorical) you can gain even more. For some problems it can be the fastest and decent non linear model from my experience.

The algorithms to train, initialize the networks, new architectures are far more important than the hardware advances. If people knew how to train NNs 50 years ago we would live in a different world.

What is wrong with event sourcing? We also built our app without any microservices but event based processing is actually quite good pattern. Basically the state of the application can be replayed with the events.

Anyone can draw an offensive cartoon. They are withholding the solution for money. If a startup that is removing backgrounds from photos is worth close to $100M (remove.bg) imagine how much this is worth. This technology is worth billions and could replace illustrators in many cases in web design, content, etc.

I like the analogy to learning a language. The immersion is very important. I did exactly what you said. I didn't read many tutorials I just sat down and rewrote a project in Rust. I don't use lifetimes properly right now (I abuse cloning many times) but I am happy with the progress.

Another thing worth mentioning is that while being a great language Rust is still a tool. Designing the architecture, algorithms, logic of the system, refactoring are much more difficult than the language itself in my opinion.

I see thank you for the clarification.