Finbarr
@finbarr.bsky.social
📤 2088
📥 58
📝 89
building the future research at midjourney, deepmind. slinging ai hot takes 🥞at artfintel.com
This is one of my all time favorite papers:
openreview.net/forum?id=ByJ...
It shows that, under fair experimental evaluation, lstms do just as well as a bunch of “improvements”
loading . . .
On the State of the Art of Evaluation in Neural Language Models
Show that LSTMs are as good or better than recent innovations for LM and that model evaluation is often unreliable.
https://openreview.net/forum?id=ByJHuTgA-
about 1 year ago
3
24
4
Really fun conversation with
@natolambert.bsky.social
!
add a skeleton here at some point
about 1 year ago
0
5
0
Apparently there *is* another finbar(r) in Alberta.
about 1 year ago
0
5
0
New homeowner fear unlocked; someone hit and ran my neighbor’s garage
about 1 year ago
1
3
0
there’s a type of “not trying” which means not executing at the level of competence of a $XX billion corporation this is the complaint about eg Google products. They’re good! better than most startups! But not “trillion dollar corporation famed for engineering expertise” good.
about 1 year ago
0
8
0
I watched too many ski movies and now am trying to convince my wife we should move to Alaska
about 1 year ago
2
5
0
building my own mlp implementation from scratch in numpy, including backprop, remains one of the most educational exercises I’ve done
about 1 year ago
2
19
0
Love this. Very clean implementations of various inference optimizations.
add a skeleton here at some point
about 1 year ago
1
6
0
Agreed! Folk knowledge is worth publishing!
add a skeleton here at some point
about 1 year ago
1
10
0
my latest article for Artificial Fintelligence is up. i cover the evolution of Vision Language Models over the past few years, from complex architectures to surprisingly simple & effective ones. (link in next tweet)
about 1 year ago
4
38
4
the tech industry is like California: it gets a few of the most important things right, so it can make a lot of other mistakes and still be wildly successful
add a skeleton here at some point
about 1 year ago
0
3
0
the remarkable success of the Google brain (and OpenAI) resident programs is an indication to me that smart, hardworking people can do more than you expect
about 1 year ago
4
19
1
My favorite thing about Bsky so far is not having the dumb algo requirements. Link in reply and not mentioning substack is stupid!
about 1 year ago
3
11
0
for all the work we researchers do, the best way to improve your model by far is to 1) use better data and 2) use higher quality data
about 1 year ago
1
7
0
If I was doing a phd, this would be one of my top choices for programs.
add a skeleton here at some point
about 1 year ago
0
2
0
active learning is top of my list of "things that seem like they should work but don't" I haven't had much success when I actually implement it
about 1 year ago
1
4
0
Why does DeepSeek have so many GPUs? (Purportedly >10k H100s). Is this useful to their main hedge fund business?
about 1 year ago
1
1
0
Tulu is very exciting!
add a skeleton here at some point
about 1 year ago
0
2
0
Deepseek r1 seems really good
about 1 year ago
1
4
0
If I was in a position to direct significant research resources/headcount I’d be putting a significant effort behind better exploration in RL.
about 1 year ago
1
8
0
honestly if houses in Whistler were less expensive I’d be way less motivated to earn money the prospect of retiring there motivates a non-trivial amount of my life
about 1 year ago
0
2
0
the amount of time I spend thinking about skiing rn is ridiculous
about 1 year ago
0
1
0
ok but seriously at some point we really need to solve exploration in RL
about 1 year ago
2
8
0
reposted by
Finbarr
Venkatesh Rao 🔹
over 2 years ago
galaxy brain take: Google no moats memo is a psyop and a no-downside shot against openAI It’s a little too flattering to the conceits of open source etc
4
23
3
chain of thought reasoning keeps racking up the Ws
over 2 years ago
1
4
0
my thesis is that there’s gonna be one foundation model company for each major cloud provider that they either partner with or acquire, and these are the only companies that make money Azure: OpenAI Amazon: Stability (?) GCP: Anthropic/DeepMind
over 2 years ago
0
1
0
the most useful thing I’ve done in my career as a research engineer is to slowly build a bag of tricks and to try them on all the problems I come across
over 2 years ago
1
0
0
active learning seems like what everyone should be doing Particularly in RL
over 2 years ago
0
0
0
ok emergent properties of LLMs have nothing on emergent properties of babies my son just learned to say dada, very exciting
over 2 years ago
0
0
0
copilot makes it so easy to comment my code & write docstrings my coworkers should be paying github
over 2 years ago
0
1
0
the fact humans haven’t eliminated mosquitoes from the earth is the best argument against AI foomerism
add a skeleton here at some point
over 2 years ago
0
1
0
not that “tweeted” is particularly dignified, but “as I skeeted yesterday” is egregiously bad
over 2 years ago
0
0
0
is anyone actually using cerebras in production
over 2 years ago
0
0
0
in the pnw they use greysky instead
over 2 years ago
0
0
0
skoot skoot mahout
over 2 years ago
0
0
0
there are so many aspects of productionizing ML models we know nothing about
over 2 years ago
0
0
0
it’s crazy how little LLM experience you need to be an “expert”
over 2 years ago
1
0
0
Ok American Jesus ads during the playoffs are a new experience
over 2 years ago
0
0
0
American fast food ads are so much more enticing than Canadian ones
over 2 years ago
1
0
0
how hard could it be to start a semiconductor company
over 2 years ago
0
0
0
I keep thinking about parameter quantization for large models I have periods where I’m convinced it’s the future of inference, and then other periods where m I’m convinced it doesn’t do anything Any anecdata out there? 👀
over 2 years ago
0
0
0
you reached the end!!
feeds!
log in