Hacker Newsnew | past | comments | ask | show | jobs | submit | er453r's commentslogin

For my specific niche (medical imaging) all current models still suck. The amount of expert knowledge required to understand the data and display it in the right way - probably never was in the training set.

We have this one performance-critical 3D reconstruction engine part, that just just has to go FAST through billions of voxels. From time to time we try to improve it, by just a bit. I have probably wasted at least 2 full days with various models trying out their suggestions for optimizations and benchmarking on real-world data. NONE produced an improvement. And the suggested changes look promising programming-wise, but all failed with real-world data.

These models just always want to help. Even if there is just no way to go, they will try to suggest something, just for the sake of it. I would just like the model to say "I do not know", or "This is also the best thing that I can come up with"... Niche/expert positions are still safe IMHO.

On the other hand - for writing REST with some simple business logic - it's a real time saver.


Did you feed back the results of the tests / benchmark to the model?

I’m presuming you have a very robust test framework / benchmark setup etc?

I’m presuming you fed the model the baseline results of that setup as a starting point ?


Everything is contained in unit tests, so it is really easy to check a small change and get the results for the same dataset.

I have had put the agents in a small feedback loop - but limited to ~5 iterations, since I do not want to burn through my credits. Honestly, I think putting it in a loop for a case like this does not show the "intelligence" part of AI, it makes it more of a more-sophisticated random walk.


One even-better approach IMHO

Just keep a .gitconfig in your HOME with aliases for your identities. Then just after initializing/cloning the repo do git config-company or git config-personal

    er453r@r7:~$ cat ~/.gitconfig 
    [user]
        useConfigOnly = true
    [alias]
        config-personal = !echo CONFIG-PERSONAL && \
            git config --local user.email 'personal@email.com' && \
            git config --local user.name 'personal' && \
            git config --local core.sshCommand 'ssh -i ~/.ssh/id_rsa_personal'
        config-company = !echo OLD CONFIG-COMPANY && \
            git config --local user.email 'official@comapny.io' && \
            git config --local user.name 'Name Surname' && \
            git config --local core.sshCommand 'ssh -i ~/.ssh/id_rsa_company'


How would you do the initial clone without the correct ssh config to begin with? I think the benefit of the article's method is that any clone from their org will just work.


I have something like the parent suggests and yes, the article's idea is better because you don't need to do anything manual nor remember to run your own command at all.


You are right - that first clone has to be manually preceded by GIT_SSH_COMMAND="ssh -i ~/.ssh/id_rsa_personal" - but after this you just configure the repo and forget about it.

I just like this workflow better since it is totally directory/remote agnostic (compared to the article).

Just use whatever suits you best :)


The same exact config you'd have to replicate for the OP's solution is what you'd have to replicate as well. It also means your config is semi-copied into every repo instead of being centrally managed. This is like the worst of all options, idgi.

I'm sorry, I don't want to be mean but this has got to be the worst way Ive seen someone try to solve this. I want to cry skimming it. Why would anyone do this and think it's simpler? Wew, gotta just leave this one alone.


You don't need to specify your key file like that if you've got your keys loaded into an SSH agent. Or even better, IMO: use 1Password or similar that can both store your keys securely _and_ can be used directly as an SSH agent. That way, your SSH keys never need to be stored on disk in the first place.


When I was a child I had a mild case of synesthesia. I would often see colors or feel textures when hearing sounds. Touching different textures would trigger memories associated with the sound-texture pair... A wild trip for an adult, but at that time I thought everybody perceived the world that way.

I have also had a feeling for Magic. An unmistakable tingling, in my brain, when I learned/discovered something new. It was the best feeling in the world - it kept me exploring and learning the unknown.

All of that started getting weaker somewhere in my teens. Then it completely disappeared when I was around 20. I still hold the memories, the codex of sound/color/textures, but nothing triggers it. Nothing triggers the Magic sense. It all just faded away.

I would do anything to feel it again.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: