Nikita says they were "never" deboosted, but Musk said they were going to do that and it was a huge topic...?

https://x.com/nikitabier/status/2041911302541730237?s=20

He says here about an interface change. I've noticed this change. The sites are opening in a kind of sub window with the feedback UI still visible. I found this annoying but now I see the point.

Like... That. Rhetorical ellipsis. Like you see in a 12 year old's fanfic.

I know one of the AIs had a style change. I think Grok. But it started using drama dots so now they are everywhere.

And unlike the em dash, _nobody_ notices. _Nobody_ sees it.

People put out AI text, primarily, to run hustles.

So its writing style is a kind of internet version of "talking like a used car salesman".

With some people that's fine, but anyone with a healthy epistemic immune system is not going to listen to you.

If you want to save a few minutes, you'll just have to accept that.

Everything an AI does is downstream of deliberate, albeit imperfect, training.

You know this, you rig it all up and you let things happen.

I’ve gone over the years from Visual Studio fanboy to writing everything in vi, entirely due to software decay.

Our culture and economy can no longer maintain complex GUIs.

This

When I receive a PR, of course it’s natural an AI is involved.

The mortal sin is the rubber stamp.

If they haven’t read their own PR, I only have so many warnings in me. And yes, it is highly visible.

Intel and Nvidia are both offering both C and Fortran compilers, so I was looking at both. I know C well but I decided to not look at it as a presumed default.

When I used C like this in the past, the intrinsics were very low-level, e.g. wrapping specific Altivec or SSE instructions or whatever. I see see it has OpenMP intrinsics, though, which I’m sure I’ll try later.

If I use a library, I’m breaking up the operations and don’t give the optimizing compiler an opportunity to take operations into account together.

With Fortran, I can give the operations directly to the compiler, tell it the exact chip I’m working with and it deals with it.

It would be fun, when I have some time, to go rewrite it in C and see how it compares.

(I considered JAX, but the code in question was not amenable to a compute graph. Another option was to thread by fork, and use IPC.)

I liked the language itself more than expected. You have something like "generics" with tensors. Suppose you pass a parameter, N, and you also would like to pass a tensor, and you would like to specify the tensor's shape (N, N). You can do this; the parameter type constraints can reference other parameters.

Tensors and various operations are first-class types, so the compiler can optimise operations easily for the system you're building on. In my case, I got 80% improvement from ifx over gfortran.

Invocation from Python was basically the same as a C library. Both Python and Fortran have facilities for C interop, and Numpy can be asked to lay out tensors in a Fortran compatible way.

Part of what eased the port was that Numpy seems to be a kind of "Fortran wrapper". The ergonomics on tensor addressing, slicing and views is identical.

If you can produce any kind of economically productive compute node and add it to (for example) the Starlink network, and launch on a reusable vehicle, you carry on installing them as fast as you can build them.

So, the move is to turn the problem of contested land use into a manufacturing problem.

This is not so easy to pin down on a spreadsheet, and will be decided at the level of the business unit. If SpaceX can put a GPU/TPU on the grid more economically than the other guy, then it doesn't matter if they have ammonia in the pipes instead of water.

Grab your popcorn.

A tax on corporate profits is a tax on cost cutting already.

A tax on “AI” is a way to compartmentalize. But you can’t, and you shouldn’t.

First, you won’t be able to formalize which gains are “AI” and which are not. Is it deep learning? If so, a gunshot detector is taxed and a McDonald’s touch screen is not. Is that what you want?

Second, a host of labor savings that don’t look like “robotics” or “AI” are also not covered. If you increase the MTBF on a traffic light, you cut the labor of light replacement. Is this morally different than a McDonald’s kiosk?

What about the traffic light itself? Shouldn’t that be a cop with a whistle?

We can do this all day.