mg

I wonder how the math turns out when we compare the energy use of local vs remote models from first principles.

A server needs energy to build it, house, power and maintain it. It is optimized for throughoutput and can be used 100% of the time. To use the server, additional energy is needed to send packets through the internet.

A local machine needs energy to build and power it. If it lives inside a person's phone or laptop, one could say housing and maintenance is free. It is optimized to have a nice form factor for personal use. It is used maybe 10% of the time or so. No energy for internet packages is needed when using the local machine.

My initial gut feeling is that the server will have way better energy efficiency when it comes to the amount of calculations it can do over its lifetime and how much energy it needs over its lifetime. But I would love to see the actual math.

The way I often visualize the solar system is:

If the sun would be the size of a coin, then earth would be around 2m away from it and so small you could barely see it.

Certainly not in terms of electricity generation:

https://ourworldindata.org/grapher/electricity-generation?ta...

But does Mistral stand a chance to make a state of the art model?

People will always want to use the best model. Like they use the best search engine.

How would Mistral catch up with US companies, who spend tens of billions of dollars per year on improving their models? As far as I know, Mistral raised something like $1B so far.

The article is about email. Europe could probably bring email "in-house". Large players like IONOS and OVH might help do it at scale.

But what about AI? Soon all of our email will be pre-handled by our OpenAI assistant while we will be driven around by Waymo and a good part of our work is done by a Tesla humanoid robot. How can Europe catch up and do that in a sovereign way?

For world-class AI, a country needs:

    1 Photolithography
    2 Wafers
    3 GPUs
    4 Software (SOTA Neural Networks)
    5 Energy

Components 2-5 seem not on the horizon on a world-class level in Europe. So Europe probably won't have the means to do AI "in-house" in the coming decades, right?

I have yet to see a description of consciousness that is testable and is expected to be unreachable by machines.

Every time I have this discussion, it goes something like this:

A friend: ... says something about consciousness ...

Me: What is consciousness?

Friend: The awareness of one's own internal state.

Me: Like when my computer tells me how much free RAM it has?

Friend: No, more complex.

Me: So complex pieces of software are conscious?

Friend: No, they don't have emotions.

Me: What are emotions?

Friend: Being drawn to or away from something.

Me: Like how my computer prefers to have the screen saver on when possible?

Friend: No, more complex.

Me: So a more complex screen saver would have emotions and consciousness?

And on and on it goes...

I see.

That's a pretty big argument to go with an Android phone.

Being able to write your own tools in HTML is so nice.

Traversing local directories is supported via the File System Access API in browsers these days:

https://stackoverflow.com/questions/64283711

It works nicely in Chrome on the desktop and on Android. Not sure how the situation is on iOS.

I wonder if one could build this in HTML, so there is no need to install anything.

On both iOS and Android, HTML can play videos and mp3s while the screen is turned off. So maybe it is possible?

In my experience, showing a project around (friends and social media) is enough for Google to slowly pick it up and, over time, show it to more and more people to whom it is relevant.

For that to work, it is important to think about what people search for. When I started my Product Chart project, which lets users compare products on a two-dimensional interface, I at first called it "The tourist map of flash drives":

https://news.ycombinator.com/item?id=7465980

Of course, nobody searches for that.

Things picked up steam when I renamed it to "Product Chart":

https://news.ycombinator.com/item?id=8907681

Seeing 20 years into the future is quite possible in some aspects.

I remember how me and my friends discovered email in 1999 and were like "Yay, in the future we'll all do this instead of sending letters!". And it took about 20 years until letters were largely replaced by email and the web. And when the first videos appeared on the web, it was quite clear to us that they would replace DVDs.

Similar with the advent of self driving cars and solar energy I think.

I am thinking about the future here.

I don't think there will be much carbon intensive energy creation in a few decades from now. It does not make sense economically.

Building and running a nuclear reactor involves a lot of physical activity. And if the past is an indicator, we always move from physical activity to the flow of electrons.

The discussion about nuclear vs solar remind me of the discussions about spinning HDs versus solid state drives when they were new.

Existing rides will be done more efficiently but since rides are so much cheaper without a driver, much more rides will be done.

A car driving from A to B will cost less than 50% of the current price. Which will unlock a huge amount of new rides.

Most of the sunlight that hits a roof is already turned into heat. Whether you use that for calculations or not does not make a difference.

Not sure about the exact numbers, but I guess that at the moment normal roofs and solar panels absorb very roughly about the same percentage of sunlight.

So if in the future solar panels become more efficient, then yes, the amount of sunlight turned into heat could double.

Maybe that can be offset by covering other parts of earth with reflective materials or finding a way to send the heat back into the universe more effectively.

The brain uses 20% of the human body's energy.

I wouldn't be surprised if mankind will evolve similar to an organism and use 20% of all energy it produces on AI. Which is about 10x of what we use for software at the moment.

But then more AI also means more physical activity. When robots drive cars, we will have more cars driving around. When robots build houses, we will have more houses being built, etc. So energy usage will probably go up exponentially.

At the moment, the sun sends more energy to earth in an hour than humans use in a year. So the sun alone will be able to power this for the foreseeable future.

    Python → Scientific computing and machine learning.
    PHP → Web backend.
    

I wouldn't be surprised if more web backend code is written in Python than in PHP these days.

Not sure how to figure it out. Google trends maybe?

https://trends.google.com/trends/explore?date=all&q=python%2...

What I witness is that many people pick a random model and expect a prompt they wrote in 10 seconds to replace an hour of coding. And then are disappointed and go back to coding manually.

For me, the magic of LLMs is that I already get an hour of coding via 30 minutes of prompting and finetuning the generated code. And I know that the ratio will constantly improve as LLMs become better and as I finetune my prompts to define the code style I prefer. I have been coding for pretty much all my life and I never felt more excited about it than I do now.

It would be cool if people shared their prompts and the resuling commits. If someone here is disappointed by the commits LLMs make, I would love to see the prompt, the commit, and which model made it.

I don't think so.

The price of a bus is roughly 10x the price of a car.

The average number of passengers in a car is about 1.5.

So the average number of passengers in the bus has to be over 15 to be more cost-effective than a car. 15 passengers is far from full. And it is on average, not most of the time.

For the space the passengers take up on the road, it is even more extreme. Take a look at how far apart cars drive on the road. About one or two cars take up the same space as a bus that can carry dozens of passengers.

I would expect the price per seat to be lower in a larger vehicle than in a smaller one.

Also the amount of space per passanger the vehicle takes up on the street is smaller for a larger vehicle.

I wonder when buses in public transport will become autonomous.

It seems like a much easier thing to do. They drive on fixed routes and do not have to handle complicated parking situations.

Here in Germany, Volkswagen is working on autonomous buses with its Moia subsidiary:

https://www.youtube.com/@moia6222/videos

As far as I know, the self-driving software they use is provided by Mobileye.

But I have not yet seen any stats about miles driven, number of engagements etc. So I have no idea how far along they are.

The article complains that most movies follow the same plot:

We meet the protagonist in their ordinary world, then an inciting incident changes everything, they are pulled into a new quest, meet someone who shows them a different way of being, they struggle with a powerful antagonist, and in the end the protagonist either triumphs or fails tragically.

HN to the rescue! What are some movies that do NOT follow this plot?

How is that related to the syntax?

Hyper syntax:

    <tr :for="user of users">
        <td>${ user.name  }</td>
        <td>${ user.email }</td>
        <td>${ user.age   }</td>
    </tr>

EJS syntax:

    <% for (user of users) { %>
        <tr>
            <td><%= user.name  %></td>
            <td><%= user.email %></td>
            <td><%= user.age   %></td>
        </tr>
    <% } %>

I prefer EJS, because all you need to know is "<% begins JS" and "%> ends JS". Everything else is just plain HTML and JS.

The EJS website: https://ejs.co

It highlights the items, and the highlighted items are than offered as a comparison when you hover other items.

Have you seen the "compare" button in the popup when you hover an item? It lets you highlight multiple items and compare them later.

Yes, these are also projects of mine.

It is 20 GB per Dollar.

Yes, product selection is not perfect yet. I originally set out to display the 300 most relevant products in each category. It is probably better to have a larger set of products in each category.

I will tackle that. Not sure yet how hard / easy it will be. Because more than 300 items on the screen initially might make them too small. And adding more as one uses the filters might be confusing.

All categories with screens (laptops, tablets, phones, monitors) have the option to switch the axis to "pixels per inch". Hover one of the axis arrows with the mouse to select it.

Does that help?