tmearnest

I was terrified to look up through the comments after reading the article, but HN truly surprised me today.

It’s art. Same as all other high fashion. No one is wearing these nails on the subway.

Once I saw that Waymo rage pass all of those bicyclists, I knew we finally solved self driving cars.

1.6 μm

ChatGPT is amazing for neurodivergent folk. When I was applying for jobs this summer, I just word vomited a stream of consciousness professional biography into a text document. I then used this as context for chatgpt to help me write cover letters and resumes. If you want to be especially clever you can also include the job posting. Just be very careful to change up some of the language otherwise it’ll smell like AI.

No. The problem is in a reduction op of some sort (sum or whatever). Since there no guarantee of the order you receive the terms for the reduction, the nondeterminism enters from order of terms reduced. Since float math isn't associative, there will be slight differences depending on the order and these can amplify quickly over a deep net.

You would have to explicitly order the terms prior to reduction but you don't always have that level of control.

Yup, I don’t believe there are any consequences to anyone away from the hole.

I think the right way to approach this is to consider the unshielded radiation flux over the hole and the time that the hole takes to close. This would give a good back of envelope upper bound of the increase in cancer risk. There’s probably other effects, but all I care about is harm to individuals.

I used to work at a 24 hr end user tech support call center. They didn’t use Macs, but we had a machine for the techs to use to understand what the customer is looking at. I wrote a script to sleep until late at night then start saying weird/creepy stuff to mess with the overnight crew.

This is all correct. These simulations actually can model molecular crowding by using a diffusive propensity proportional to the particles in the adjacent cells but I don’t think it was used here. I developed this methodology in grad school, but it didn’t go any where.

Dna does not fit in a voxel. Transcription is modeled by particles fixed in space which produce transcripts at a constant rate. The mRNA is treated the same as the other discrete particles in the simulation, though they are likely a bit bigger than the voxel size.

There are two main reasons to take advantage of the Gpu in lattice microbes. It can simulate the stochastic chemical reaction and diffusion dynamics in parallel: one thread per voxel. For instance, an E. coli sized cell would have ~40000 voxels. It’s not quite embarrassing parallel, but close. Second, the simulation is totally memory bound so we can take advantage of fast gpu memory. The decision to use CUDA over OpenCL was made in like 2009 or so. Things have changed a lot since then. I don’t think anyone has the time or interest to port it over, unfortunately.