phonebucket

Easy trick to eliminate difficulties arising from pi: fix them via legislation [0].

[0] https://en.wikipedia.org/wiki/Indiana_Pi_Bill

Is this article actually trying to talk about superintelligence? Or is this article just trying to make GPT look like AGI by taking AGI for granted?

What is the breakthrough? There is no mention of the problem solved, nor the accuracy of the quantum solution, nor the amount of classical computing resources needed for a non-quantum solution.

The engineering probably has some cool novelty: 39 qubit and 10 million layers is a very large circuit. But simulating a large circuit is very different to achieving a quantum computing breakthrough in CFD.

Apart from being fun, this seems like a nifty way to gather a huge amount of data on attempts to prompt engineer LLMs.

I do get where you are coming from. Indeed, it makes little sense to use Julia for lots of machine learning when PyTorch and Jax are just so good. And it sounds like you don't want to use Julia, so who am I to try and convince you? Python/R are capable languages.

But, there are still reasons I reach for Julia.

Interesting packages where I prefer Julia over Python/R: Turing.jl for Bayesian statistics; Agents.jl for agent-based modelling; DifferentialEquations.jl for ODE solving.

I would much rather data-munge tabular data in Julia (DataFrames.jl) than Python, though R is admittedly quite nice on this front.

Personally I reach for Julia when I want to use one of the previous packages, or something which I want to code up from scratch, where base Julia is much preferable to me than numpy.

I'm eager to jump into a Lisp.

Janet seems really tempting for tiny footprint, distributability etc.

But I'm currently leaning towards to Racket just because it would be more or less compatible with a whole host of Scheme books that I'd like to read (The Little Schemer/Typer/Learner, SICP, Functional Differential Geometry).

Does anyone familiar with Janet know if those books can be easily worked through with Janet for a newbie Lisper?

Two big hurdles for ML: a) explainability b) accountability.

ML-enhanced development neatly circumvents this: explainability and accountability is passed on to the developer. This includes bugs and license infringements.

I have no qualms about ML tools in development. But so long as the buck stops with me, I prefer to write from scratch.

Traditional chess engines excel at tactics. But most chess tactics are brute forceable, and traditional chess engines are normally brute force behemoths.

Large language models have tradtionally been very weak at brute force computing (just look at how bad they are at multiplicaion of large numbers). If it can somehow excel at something which typically involves computing power like chess tactics deep into a game well after a novel position has been reached, then put me down as impressed.

There is much more to generative models than building out language models and image models.

Generative models are about characterising probability distributions. If you ever predict more than just the average of something using data, then you are doing generative modelling.

The difference between generative modelling and predictive modelling is similar to the difference between stochastic modelling and deterministic modelling in the traditional applied mathematical sciences. Both have their place. Neither is overrated.

Grab the best tool for the job.

Are experienced devs really ditching Python for production systems? I wasn't under this impression.

You can specify which config to use while launching NeoVim with the -u flag.

You can argue that you read the questions first too, so you should give yourself the answer to querstion 1 as correct as well and have 99/100 overall

A rubric should be specified outside of the context of a specific question if it applies all questions.

I think this test is pretty ambiguous.

If you answer all the questions after reading all of them, you could only get only question 100 wrong and get the other 99 correct, so reach a score of 99/100.

If you answer only questions 1 and 2, you get the answer to 100 right too, so you could score 3/100.

In the latter case, you could score 3/3 too, depending on how the examiner decides to mark it.

So which is it?

You can fully embed Neovim into VSCode: https://github.com/vscode-neovim/vscode-neovim, which seems like it might work for you?

I started down the same path, but ended up somewhere else.

I decided to allow myself as many extra features as I wanted with one caveat: all plugins had to be pure vimscript.

Reproducing an exact dev environment is just a git clone away, because there are no external dependencies outside of vim itself.

This is not very restrictive: you have fuzzy search with Ctrl-P, get full-featured language server support with vim-lsp etc.

This is possible because Vim8+ (released 6 years ago) has a native plugin manager which is very easy to manage via git submodules.

I used to use a Vimium-FF (a Firefox version). I love the better browsing ergonomics.

But, after some thought, I realise that allowing such extension simply requires trusting a third party app access to so much of my personal information. They (understandably) need the ability to read every page I browse. Vimium-FF even requires clipboard access.

I do believe the authors have the best intentions. But the amount of trust I need to have in order for a third party app to have so much access to some of my most confidential information (online banking, emails etc.) is very very high.

While I agree that academic fraud is a blight upon academia, this comment generalises over many thousands of academics.

_nobody_ cares about any of this. Universities don't care, journals don't care, peers don't care, nobody within the academic system cares.

Please be mindful about generalising over such a large number of people. Maybe those who care have little power to do anything about it.

Still find it nuts you need to operate the system at 0 kelvin. I assume if we had super efficient cooling equipment (not LN2) we could also do superconductors on a small scale?

Neutral atom quantum computers (the type outlined in the article) work at room temperature.

Superconducting qubit based quantum computers do indeed require cooling to these levels, though.

Apologies for making a more meta-comment. But I think the title of the post title should be the the paper title "Attention Is All You Need" as per the HN guidelines.

"...please use the original title, unless it is misleading or linkbait; don't editorialize." [0]

[0] https://news.ycombinator.com/newsguidelines.html

remember 10 random words

Alex Graves (now at DeepMind) put a similar demo on his University of Toronto page a long time back: https://www.cs.toronto.edu/~graves/handwriting.html

By migrating from one to the other, aren't you exposing yourself to the exact same risk?

My main gripe with LastPass is that they did not encrypt everything. Vast amounts of important information (email addresses, billing addresses, telephone numbers, IP addresses, website URLS [0]) were not encrypted on user's local machines with the master password, and subsequently have fallen into the hands of a malicious actor.

I would feel much better about LastPass if the security genuinely was safeguarded by a strong master password. But they've demonstrated that it's not.

Other password managers, as far as I can see, provide much greater protection in terms of encrypting everything. That's why I'd feel better about using them.

[0] https://blog.lastpass.com/2022/12/notice-of-recent-security-...

A lot depends on what you're interested in.

Some papers that are runnable on a laptop CPU (so long as you stick to small image sizes/tasks):

1) Generative Adversarial Networks (https://arxiv.org/abs/1406.2661). Good practice to have a custom training loops, different optimisers and networks etc.

2) Neural Style Transfer (https://arxiv.org/abs/1508.06576). Nice to be able to manipulate pretrained networks and intercept intermediate layers.

3) Deep Image Prior (https://arxiv.org/abs/1711.10925). Nice low-data exercise in building out an autoencoder.

4) Physics Informed Neural Networks (https://arxiv.org/abs/1711.10561). If you're interested scientific applications, this might be fun. It's good exercise in calculating higher order derivatives of neural networks and using these in loss functions.

5) Vanilla Policy Gradient (https://arxiv.org/abs/1604.06778) is the easiest reinforcement learning algorithm to implement and can be used as a black-box optimiser in a lot of settings.

6) Deep Q Learning (https://arxiv.org/abs/1312.5602) is also not too hard to implement and was the first time I had heard about DeepMind, as well as being a foundational deep reinforcement learning paper .

Open AI gym (https://github.com/openai/gym) would help get started with the latter two.

A lot of recommendations depend on what you're trying to learn.

But I've enjoyed the following texts to a larger extent than others:

- Algebra: Chapter 0 (Aluffi)

- Real Mathematical Analysis (Pugh)

- Mathematics and its History (Stillwell)

- An Introduction to Manifolds (Tu)

- Gauge Fields, Knots and Gravity (Baez)

- A First Look at Rigorous Probability Theory (Rosenthal)

- All of Statistics (Wasserman)

There are some authors I trust and am happy to buy so long as the topic vaguely interests me: VI Arnold, Tristan Needham and John Stillwell.

I really like the list put out by @enriquto in a separate comment, but I've avoided duplicating those recommendations in the list above.

One example off the top of my head: what if ChatGPT was used to automate large scale personalised scamming?

But in the context of a strong master password, the additional benefit of the secret key is of neglible benefit, while the hassle and dangers of having to synchronise the secret key remain.

I'd rather use an extremely high entropy master password by itself.

Of course it can be misused, like any tool, but no amount of safety features in ChatGPT will make users of it more or less careful in their use of it. If someone using ChatGPT cares nothing for using it safely then it will likely end poorly, just like it will end poorly if I use a hammer without any care for using it safely.

I'm not worried about hammers when I have them. I'm worried about hammers when someone who wants to hurt me has them.

Same with ChatGPT.

It's more than a large piece of key loops that are compiled. Almost everything is compiled. Python hardly does anything beyond orchestration in modern scientific pipelines. Even loops written in Python can be compiled away.

It's almost like you are coding in a separate language altogether sometimes. E.g. you write something in Cython that gets compiled to C/C++. Same holds for things like numba/LLVM or Jax/XLA. PyTorch is going bigger on compilation in PyTorch 2.0.

Python can be seen as a common interface to write code that can be JIT compiled to multiple backends. It's quite convenient to be able to interact with all these backends, and bring the results back to a universally common interface (NumPy).

From this perspective, scientific Python itself is somewhat evolving by itself into something like Julia, which is JIT compiled to LLVM. Except Python's momentum has generated a huge number of backends for users to choose from.

I'm all for a new language dethroning Python in scientific computing. But I don't think it will be for reasons of speed. I think it'll likely be things like typing (e.g. from Dex lang) or library modularity (e.g. Julia).

I'm pretty disappointed with how this has been handled. Mainly in two ways:

- The scale of the breach (leaking of PII and encrypted vaults, that URLs are not encrypted) has only come almost 4 months after the initial investigation. When was such a breach suspected in the first place? I would expect extremely timely reponses for this stuff.

- They suggest no recommended actions? Surely rotating credentials should be advised? Feels like they are putting business before security.

It's not just the leak that makes me lose complete faith in Lastpass. It's how they've handled the leak.