@swiftcoder@mastodon.social avatar

swiftcoder

@swiftcoder@mastodon.social

Recovering software engineer. Graphics, Procedural generation. Future farmer.

This profile is from a federated server and may be incomplete. Browse more on the original instance.

TomF, to random
@TomF@mastodon.gamedev.place avatar

Here's a picture. Quickly count how many players are on each of the two teams, then answer the poll in the following comment.

swiftcoder,
@swiftcoder@mastodon.social avatar

@TomF You've apparently located the one form of colour-blind that I do not suffer from 😂

seldo, to random

I have good news for you about HTML, LinkedIn.

swiftcoder,
@swiftcoder@mastodon.social avatar

@seldo Having spent a day recently swearing at the HTML engines gmail and outlook use to render emails... we have a long way to go still

kellogh, to random
@kellogh@hachyderm.io avatar

installing git on windows is…not easy. what is this? 20 questions?? you think this is a game???

swiftcoder,
@swiftcoder@mastodon.social avatar

@kellogh I never used a git UI until I started developing regularly on Windows. The hassle of getting a git that works in all the various types of command prompts... GitHub app it is

kellogh, to ai
@kellogh@hachyderm.io avatar

i appreciate skeptics, they have valuable points. but it annoys me when they double down on some point that’s totally bogus

one example: aren’t reliable and therefore we can’t use them for engineering

it’s misguided, but it’s worse than that. we engineer systems out of unreliable components all the time, that’s a core responsibility of engineers

e.g. amazon s3 has 11 9s if durability, from disks that have less than 99%

e.g. bridges & skyscrapers are built from unreliable materials

swiftcoder,
@swiftcoder@mastodon.social avatar

@kellogh I'd be a lot happier if people were applying engineering to the actual LLMs themselves, instead of treating them as magical black boxes and trying to engineer the outputs.

Everytime domain experts start digging into how blackbox ML systems actually solve problems, we end up rediscovering that interpretable models perform just as well as blackboxes...

swiftcoder,
@swiftcoder@mastodon.social avatar

@kellogh Is there soldi ressearch on (1)? All the research I'm aware of (which admittedly, is mostly in graphics domains) says the opposite.

swiftcoder,
@swiftcoder@mastodon.social avatar

@kellogh That's pretty arbitrary, to my mind. Sure, humans evaluate decision trees slowly, but we do most other computation tasks pretty slowly too.

swiftcoder,
@swiftcoder@mastodon.social avatar

@kellogh I'll grant you that LLMs are largely of interest due to scale and generality. However I'm hard put to find specific tasks for which blackbox models consistently beat models designed to reason along the lines a human would (albeit there's probably publication bias here, with folks only publishing interpretable models that outperform their counterparts)

swiftcoder,
@swiftcoder@mastodon.social avatar

@kellogh I guess that’s a definition problem too. I don’t really care what an end user wants in terms of explanation - that’s the PM’s whole job 😂 I most care that it can operate in a fairly deterministic mode and that you can produce reliable mappings between clusters of inputs and clusters of outputs

swiftcoder, to random
@swiftcoder@mastodon.social avatar

I don't know why, but my definition of "simple" always seems to be a little different from AWS' definition of "simple"

The following steps get everything working: - Build a Docker image with the fetch & run script - Create an Amazon ECR repository for the image - Push the built image to ECR - Create a simple job script and upload it to S3 - Create an IAM role to be used by jobs to access S3 - Create a job definition that uses the built image - Submit and run a job that execute the job script from S3

swiftcoder,
@swiftcoder@mastodon.social avatar

@kellogh I’m sure there’s an AWS ChatGPT service somewhere in the works…

kellogh, to random
@kellogh@hachyderm.io avatar

thinking the other day… if i wanted to go back to big tech (unlikely) i’d probably want to go to microsoft, which is one hell of a change from the past. seems like that’s where all the interesting stuff is going down

swiftcoder,
@swiftcoder@mastodon.social avatar

@kellogh how much do you think that has to do with Microsoft predating most of the other BigTechs?

swiftcoder,
@swiftcoder@mastodon.social avatar

@kellogh as in before. Quite a few of the modern BigTech crowd feel like they are products of very specific post-2000’s phenomena (low interest rates, the gold rush in online ad money, etc)

swiftcoder,
@swiftcoder@mastodon.social avatar

@kellogh Yeah, that's what I meant. It's really only Microsoft (and arguably, Apple) that were wildly successful before that time

swiftcoder, to random
@swiftcoder@mastodon.social avatar

Pretty sure it's just a meme. All the languages that routinely use snake case also use the normal pascal case where differentiation is needed (i.e. typenames in Python/Rust)

swiftcoder,
@swiftcoder@mastodon.social avatar

@kellogh haha. Goddamn Twitter clones 😂

  • All
  • Subscribed
  • Moderated
  • Favorites
  • •
  • anitta
  • InstantRegret
  • mdbf
  • ngwrru68w68
  • magazineikmin
  • thenastyranch
  • rosin
  • khanakhh
  • osvaldo12
  • Youngstown
  • slotface
  • Durango
  • kavyap
  • DreamBathrooms
  • JUstTest
  • tacticalgear
  • ethstaker
  • provamag3
  • cisconetworking
  • tester
  • GTA5RPClips
  • cubers
  • everett
  • modclub
  • megavids
  • normalnudes
  • Leos
  • lostlight
  • All magazines
    •