@kellogh@hachyderm.io
@kellogh@hachyderm.io avatar

kellogh

@kellogh@hachyderm.io

I'm a software engineer and sometimes manager. Currently #Raleigh but also #Seattle. Building ML platform for a healthcare startup. Previously, built an IoT platform for one of "those" companies.

Open source: dura, fossil, Jump-Location, Moq.AutoMock, others

Do I have other interests? No, but I do have kids and they have interests. I think that counts for something. I can braid hair and hunt unicorns!

I put the #rust in frustrate

He/Him

#metal #science #python

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

kellogh, to python
@kellogh@hachyderm.io avatar

goodness! no GIL, incremental garbage collection, and a JIT? all in one release? #python 3.13 is starting to feel like a real language! https://docs.python.org/3.13/whatsnew/3.13.html

adron, to random
@adron@metalhead.club avatar

Talked to some Americans yesterday that are still of the mindset that the reason every nation on Russia's border is because of "western" influence.

Yeah, no shit, why would any nation want to be like Russia, that'd be batshit insane. But oddly, they didn't mean it that way.

But also oddly, they were able to admit there's zero advantage to being aligned with Russia vs. the western world.

The juxtaposition of that thought just boggles my mind. I don't get it, and reasoning is atrocious.

kellogh,
@kellogh@hachyderm.io avatar

@adron right? the countries that border us want to be exactly like us! πŸ˜‚

Cmastication, to random
@Cmastication@mastodon.social avatar

Microsoft does so much so well these days. But this Microsoft Recall product is like watching a grown man carefully placing his hand in a door jamb and then repeatedly slamming the door on it. Obviously stupid from conception.

From: @SecurityWriter
https://infosec.exchange/@SecurityWriter/112558224281615019

kellogh,
@kellogh@hachyderm.io avatar

@Cmastication @SecurityWriter I figure Apple wouldn't build a product like Recall because their entire business is built on consumer trust, whereas Microsoft sells to enterprises and it's actually very appealing to be able to spy on employees

kellogh,
@kellogh@hachyderm.io avatar

@williamgunn @Cmastication @SecurityWriter yeah, thus far they weren't subject to an ad business, so they've never had tension between privacy and business, but AI does change this.

i've heard news that they're partnering with OpenAI/Anthropic. So maybe they deal with it by not dealing with it. But I also see a lot of academic papers from them training new (small) models. So idk, but i do know they're not going to directly endanger consumer trust, that seems clear

kellogh, to OpenAI
@kellogh@hachyderm.io avatar

anyone else seeing an outage? even perplexity is out for me

kellogh,
@kellogh@hachyderm.io avatar
kellogh,
@kellogh@hachyderm.io avatar

, , and are all out. is this the uprising?

kellogh,
@kellogh@hachyderm.io avatar

@mercutio also llm works fine too (which also makes it easy to switch to another model completely in case the API stops working)

kellogh,
@kellogh@hachyderm.io avatar

exactly what you'd expect an AI to say

kellogh,
@kellogh@hachyderm.io avatar
carnage4life, to random
@carnage4life@mas.to avatar

GPUs are a scarce resource and if you run multiple companies that are all using AI then you’ll have to play favorites.

Matt Levine had a newsletter on the topic of Musk’s investors now having to guess which company he’ll focus on. Looks like Tesla lost this round.

kellogh,
@kellogh@hachyderm.io avatar

@tob @carnage4life i’m no expert, but i think there needs to be a more clear handoff of assets. especially since tesla is public, and not primarily owned by Musk, siphoning resources from Tesla is effectively stealing from investors

moorejh, to LLMs
@moorejh@mastodon.online avatar

Our KRAGEN paper is out! This method combines LLMs & RAG with Graph of Thoughts for asking complex questions of a knowledge graph or any vector DB https://academic.oup.com/bioinformatics/advance-article/doi/10.1093/bioinformatics/btae353/7687047 #llms #artificialintelligence #bioinformatics #datascience

kellogh,
@kellogh@hachyderm.io avatar

@moorejh that’s super cool. so it both uses an KG to ground it, but also outputs a graph for interpretability? how slow/expensive is it?

norootcause, to random
@norootcause@hachyderm.io avatar

I’ve heard some people suggest that NVIDIA is the only one making money off of AI, but it’s just not true.

https://www.techtimes.com/articles/305003/20240526/ai-now-generating-millions-revenue-small-caribbean-island.htm

kellogh,
@kellogh@hachyderm.io avatar

@norootcause squatting doesn’t happen only at the gym

jimfl, to random
@jimfl@hachyderm.io avatar
kellogh,
@kellogh@hachyderm.io avatar

@jimfl what’s the household protocol?

kellogh, to random
@kellogh@hachyderm.io avatar

β€œwhat tech stack do you prefer on side projects?”

i don’t even understand that question anymore. i used to care a lot, but now i’m like, β€œidk what’s the side project?”

definitely the path of least resistance mostly, but sometimes you just want to learn a new tech stack, so that

lzg, to random
@lzg@mastodon.social avatar

there's a layer of corporate people who are interested in virtual presence in meetings, because for them meetings are not about information or problem solving, but about signaling and self-presentation. being seen at the meeting is much more important than the content. also, seeing their underlings at a meeting gives them some sense of power as well. their problem to solve is "how can I be more visible at all times"

kellogh,
@kellogh@hachyderm.io avatar

@lzg it’s threads like this that make me wonder if i’m autistic

kellogh, to random
@kellogh@hachyderm.io avatar

it occurs to me that transformers are a lot like microprocessors, in that there’s a high level management decision up front to decide how big you want to make it (transistor count vs parameter count) that dictates the capability of the end product. where those transistors/parameters are allocated is a lower level series of decisions that impacts efficiency, but overall capability is decided mostly by the total budgeted size

Cmastication, to random
@Cmastication@mastodon.social avatar

I just spent a couple of days with https://www.perplexity.ai and it amazes me that this startup has done what Google, with unlimited funds, has failed to do. Perplexity is good. Really good. And I'm trying to figure out how to jettison Google and default everything to Perplexity. I'm pleased to spend a few bucks a month to make search useful. But even the free version is better than Google.

kellogh,
@kellogh@hachyderm.io avatar

@Cmastication how is the paid version compared to free?

kellogh,
@kellogh@hachyderm.io avatar

@Cmastication ah, i was hoping i could use you as a probe, oh well

Lobrien, to ML

This is a very nice video on understanding attention in transformers https://www.3blue1brown.com/lessons/attention

kellogh,
@kellogh@hachyderm.io avatar

@Lobrien oh wow, i never could crack it before, i think i vaguely get it now

adron, to random
@adron@metalhead.club avatar

So far for 2024, what's your #AOTY(s) at this point in time?

kellogh,
@kellogh@hachyderm.io avatar

@adron knocked loose!

dev, to random
@dev@discuss.systems avatar

Me, to myself, at the estate sale: No, the place does NOT need an oversized harp

kellogh,
@kellogh@hachyderm.io avatar
norootcause, to random
@norootcause@hachyderm.io avatar

Humans performing cognitive tasks exhibit concurrency but not parallelism. We can multi-task, but we can only focus our locus of attention on one thing at a time.

kellogh,
@kellogh@hachyderm.io avatar

@norootcause i know i’m like that, but i wonder if other people truly multitask

Cmastication, to random
@Cmastication@mastodon.social avatar

LLMs as β€œrubber ducks that babble back” is an underrated use case IMO. iterating with a servant on an idea can sometimes produce interesting ideas.

From: @kellogh
https://hachyderm.io/@kellogh/112547406651437045

kellogh,
@kellogh@hachyderm.io avatar

@Cmastication i wonder how this will affect imposter syndrome. we used to have to ask dumb questions in public, now we can ask a computer, in secret. younger engineers don’t get to see as much of the process

kellogh, to random
@kellogh@hachyderm.io avatar

my 3yo exhibits this thing i call stochastic walking, where her exact walking path is wholly unpredictable, except that (1) it’s always in my way and (2) if i step into her probabilistic path cloud, her actual path will definitely converge with mine with absolute certainty

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