bluepoint

Cool insights in Desktop! I really hope that someone comes up with a desktop metaphor for the hundreds of open browser tabs (which I think are the new bookmarks) that hang open in an endless horizontal list. Imagine tabs which are like documents, can be moved around, minimized, and spatially organized in folders and subfolders. Actually, why does this not exist somehow?

I started to avoid LLM news and I feel much more relaxed. Most of it started to feel like bs hype. It will take a year or so until the hype settles and everyone has a reasonably idea of how it can be used reliably in a product.

I wish they used a different color for each orientation to see what pattern the angles form.

When a (perfect) ball is thrown at a (perfect) car, it will bounce off with the same velocity, v, but with its direction reversed. If the car is moving with some velocity -V, when the ball bounces off it will have a velocity -v - 2V, gaining an extra -2V. (This is easier to understand from the car's passenger's point of view, who will see the ball arriving with a relative velocity v + V and bouncing with -v - V, or -v - 2V relative to the ground).

In the ball-car collision the electromagnetic forces are the ones responsible for changing the direction of the ball. But in the binary system, it is gravity. In particular as the shuttle enters orbit around the "incoming" star, the star's gravity will pull it forward mostly when it completed half orbit.

I hope I make sense.

Overall, such a joy to read this paper. With basic physics it makes you dream of sci-fi...

EDIT: typos

So the space character is part of the token?

I just read the paper about LORA. The main idea is that you write the weights of each neural network as

W = W0 + B A

Where W0 is the trained model’s weights, which are kept fixed, and A and B are matrices but with a much much lower rank than the originals (say r = 4).

It has been shown (as mentioned in the lora paper that training for specific tasks results in low rank corrections, so this is what it is all about. I think that doing LoRa can be done locally.

[1] https://github.com/microsoft/LoRA

Sarcastic web development deserves to be a genre.

Can someone kindly explain, what does he mean by the use of embeddings? More specifically how are those embeddings calculated?

It is also really fast. But unfortunately it crashes even if you just sneeze (well, a year or two ago ago when I played with it, sorry if this has changed dramatically).

Can beginners without money with by PhDs, contribute? If so what would be the best way to start?

In human learning at least, you need a good teacher that can give you a self consistent and correct basis and then you build on that. If you learn randomly your understanding will be "blurry" and then you have to spend time to unlearn the bad lessons. I have personal experience with this. I definitely like the message of this result, if it is true.

I am not sure what kind of background that you have, but with a bit of basic linear algebra you might enjoy this classic:

https://jalammar.github.io/illustrated-transformer/

Because most people get paid in pesos and access to USD is controlled by the government. And probably because US banks would not give loans in dollars to argentine businesses.

Maybe you need collaborators.

The negations have multiple roles. An important one is that they add “color saturation” to the words. “It is bad”, “It is good”, “It is not bad”, “I can’t say it’s good”, “I can’t say it’s not bad”, are all different statements. So negations are difficult, maybe even for humans…

Actually, this is a really good point. Asking a question means that you have an internal model or the world and you realize that there are some data that is way off. Is it even possible at all to train an LLM to ask a question? This also gives me an idea for the future where most online text is generated by AI: we will play the game of responding questions with questions and then we know we are human!

This seems like the subject of topic modelling. Is anyone working on that? If so message me.

I got curious and searched for the BBEdit manual. Do you mean chapter 30 (there isn't any) or chapter 8, "Searching with Grep"?

If you are like me and you tried to copy paste the python commands and it did not work, you need to generate an access token. Here is what you should do:

1. Sign up (https://huggingface.co/) to hugging face.

2. Setup access tokens (https://huggingface.co/settings/tokens)

3. Install or Upgrade some dependencies `pip install huggingface_hub transformers accelerate`

4. From the terminal run `jupyter lab`

5. Then, if I did not forget any other dependencies you can just copy paste

```python

from huggingface_hub import login from transformers import HfAgent

login("hf_YOUR_HUGGING_FACE_TOKEN")

agent = HfAgent("https://api-inference.huggingface.co/models/bigcode/starcode...")

agent.run("Is the following `text` (in Spanish) positive or negative?", text="¡Este es un API muy agradable!")

```

I have been cross checking it with stored chatgpt4 answers and the wording is different but the answers seem comparable. It fails in my favorite one:

There are 3 events, A, B and C. Event C is always before B. Given this, enumerate all possible permutations of the events.

Also the weirdest thing is happening: chatgpt4 reports that they are busy and does not let me login (I pay for plus). Is everyone cross checking?

EDIT: chatgpt4 started working again. It actually answers the event order question correctly, while bard fails on it.

EDIT2: "how do I install a PC in my RAM?" The both agree, but ChatGPT4 is the one that gives the perfect (and long response). Bard tells that the RAM loses memory when turned off and as a conclusion, I cannot install a PC.

No matter what it is probably easier to inspect a language model while it works, than the language module of a human while he speaks.

The interesting bit that I got is that he made a speech implying that he had weapons of mass destruction because he wanted to scare off Iran. He actually did not have anything left because of the sanctions, but he did not want to appear weak and deal with an iranian invasion. That speech seemed to have affected the US’s decision to invade.

I don’t think so, unless it is the only official currency.

Gym is indeed boring and so is running. Pick an activity that you actually enjoy doing (or you think it has potential). Preferably one that involves flexibility, coordination and strength and pushes you out of your comfort zone. Dancing, acrobatics, martial arts, yoga.

Are you sure about the credit card? I signed up without one. There is a free option.

> The App Store reviews, while mostly positive, are full of dozens of one-star ratings from people complaining that the app is hitting on them too much, flirting too aggressively, or sending sexual messages that they wish they could turn off.

Despite the AI being horny, it still has mostly positive reviews. I find the slightly critical tone of the article a bit inconsistent with that fact.

Spot on. This certainly applies to much of physics academia at least.

Great list! At the speed of light (300000 km/s) you can go around the earth (40000km) 7.5 times per second. The tcp packet speed is almost as fast as it can possibly be and this is amazing (or not, given that it is light most of the way).

> One of the newly available GTK applications is GNOME Web aka. Epiphany, which is based on a very recent version of WebKitGTK. This provides an unfortunately non-native but largely functional web browser for Haiku for the first time in many years, with “just works” status for major websites like YouTube and others.

I was really waiting for this! I want to install Haiku in an old computer for an old lady friend to use (and for me to see how she will use it). They only problem always was that youtube never really worked. That is the chance!

It is a nice read about Qatar and covers diverse topics. What I found interesting cause I did not know is that there is such a thing as islamic economics, which is trying to avoid the inequality between creditors and debtors and also derivatives and gambling:

> So the obvious way of doing this is simply various types of share participation where rather than making a loan, a bank takes a stake and then the person receiving the money from the bank has the option of buying the bank out progressively like it’s essentially an equity-stake type model.

Would that work in the west?