d3m0t3p

I'm not complaining about the indexing (you can start the indexing by starting spotlight on and off). I'm complaining about having to search for "ghos..." instead of "term...". Because I don't don't like to remember the specific app name. But this complaint is the same for everything; when searching for Excel, I would like "Numbers" to also show up in the results. When looking for vscode I would like "Visual Studio code" to also show up, but I need to type either code, visual or studio.

Just try it. I tried and the launch was so smooth that I'm keeping it a few days to test it. My current biggest problem is that I launch my terminal using the spotlight shortcut (⌘-space on mac) and while iterm2 is found when I search for term, ghostty isn't.

He is working on CNN (Convolution Neural Network) while everyone else uses transformers. Furthermore, he wasn't in the OpenAI scandal with Ilya. Third, they don't publish papers at the same rate. Ilya has 585182 citations, while Alex has 286082 (source google scholar).

> I also feel like guarding their consumer product against bad-faith-bad-use is basically pointless. There will always be ways to get bomb-making instructions

With that argument we should not restrict firearms because there will always be a way to get access to them (black market for example)

Even if it’s not a perfect solution, it help steer the problem in the right direction and that should already be enough.

Furthermore, these researches are also a way to better understand LLM inner working and behaviors. Even if it wouldn’t yield results like being able to block bad behaviors, that’s cool and interesting by itself imo.

IF (and that's a big if) that's true then it means they can factorize number into primes with this quantum computer and break encryption.

Increasing the dimension causes a lot more computation, this is one of the main reason. You can see evidence of this in the multi head where the dim is reduced via a linear projection.

h_i = attention(W_i^Q Q^T @ W_i^K K) W_i^v V

h = W_o @ concat(h_1...h_8)

The video is really sad. No music, no sounds. I remember better videos on youtube in 2006.

Really cool, I'm a CS majoring in AI, but I'm also interested in that domain, would you have any recommendation to get started ?

I don’t agree, We can learn to apply thought process.

We learn by heart patterns and recognize them in new situations.

Learning that a -> b doesn’t mean that b-> a.

I would be able to say it for a few cases intuitively but knowing this pattern exists I’m able to apply it to all construct not a few cases I also allow me to formally prove or disprove something I have an intuition for

woaw nice, Do you open source it / make it available in any way ?

by default telegram doesn't encrypt anything. I'm not sure if you where implying that telegram had encryption and therefore hiding illicite activities.

The problems is the huge telegram channels spreading hacks and illicite tricks.

There is no such threats against Signal which is way more privacy oriented. (except for some new laws in UK where Signal said if they are applied, they would withdraw from the UK market:https://fortune.com/2023/07/13/signal-president-mass-surveil...

New Zealand, developing country ?

Yea, because the math have stripped down the whole thing to : I have data I do operation on them. while in reality we deal with multi head attention / grouped query and the positional encoding.

That’s all without taking into account the broadcasting done on the batch dimension

it says that the released the code, but I couldn't find any except some example code, Did they release the training code ?

isn't a pizza for one ?

I don't agree, imo being asked questions still draws creativity from yourself. It also help to clarify your thoughts. It's like pair programming or rubberducking. The goal isn't to wait until someone else tells you what to write.

But I agree that a lot of people are copy pasting from chatGPT until it works

I think you could say that Github copilot is partly in this way ?

Does it need to be sold in the US or simply chaning os settings to use the en_US language is enough to activate this ? I'm in europe but working mostly in english and therfore I set my mac in english. Could I use Apple Intelligence ?

I really liked that paper, easy to understand and explain every terminology they used, even the variance computation is explained! I don't expect that much detail from every paper, but it's a bliss to read.

According to https://arxiv.org/abs/2405.15071 their grokked model outperformed GPT4 and Gemini1.5 on the reasoning task. We can then argue if the task makes sense and the conclusion stands for other use cases but i think grokking can be useful

You might increase the probability of getting a correct answer for your region, but imo you decrease your awareness to allucination. Overall you can still get a wrong answer

I think it's from the AWS team, they made firecracker (micro VM) So it does exist.

Funnily that's what fly does: take your container uncompress it to a full micro VM and run it on their infra

Hey, nice work, how to get updates about the open source release ?

That's funny you mentioned switching to another language, I recently asked chatGPT "translate this: <random german sentence>" And it translated the sentence in french, while I was speaking with it in english"

What GPU do you use ? Aren’t you losing money with that ?

You can load libs at runtime with dl-open, for exemple if you need a feature you can load the corresponding lib. While with dynamic linking you'd load everything when the process is launched slowing the launch

You might want to check pola.rs then, it's backed by the appache arrow memory models and it's written in rust. All the columns have a defined type and you can easily catch a mistake when loading data

Web application firewall

Cause it’s funny ? Enjoy the little things in life

It sounds like it to me. In the sentence "This project implements Reinforcement Learning from Human Feedback (RLHF) training...

the keypoint for me that trigger my "ChatGPT-sense tingling is the abbreviation of (RLHF) when it's kind of common terminology in the LLM space nowadays.

Otherwise, the sentence's "shape"/construction of : Proximal Policy Optimization (PPO) PPO is an optimization algorithm used in reinforcement learning to update policy parameters by optimizing a clipped surrogate objective function. The objective function for PPO is defined as...

Sounds realy like ChatGPT or the begining of an Wikipedia article.

I'd be interested in knowing if indeed these part have been written by ChatGPT, the guy himself, or even another llm.