maciejgryka

10 ms·

31.
by maciejgryka 2y ago
Isn't "two models slapped together" basically how all of these things work, starting with CLIP? Not sure about GPT4o, obviously, I don't think they released any underlying architecture details?
32.
by maciejgryka 2y ago
I haven’t tried this yet, excited to see how it can do segmentation by outputting series of coordinates! That's something I just assumed transformers will generally be bad at.
33.
by maciejgryka 3y ago
I hear you and am planning to try DSPy because it seems attractive, but I'm also hearing people with a lot of experience being cautions about this https://x.com/HamelHusain/status/1777131374803402769 so I wou
34.
by maciejgryka 3y ago
Nice, looking forward to seeing that! Someone else pointed me towards https://github.com/DAGWorks-Inc/burr/ which also seems related in case you're curious.
35.
by maciejgryka 3y ago
Ha good point, I did say "let's have another shot" when I just meant another try at generating! FWIW "few shot prompting" is how most people refer to this technique, I think (e.g. see https://www.promptin
36.
by maciejgryka 3y ago
I'm a little surprised to hear this, my experience has been a little better. Are you using GPT4? I know 3.5 is significantly more challenged/challenging with things like this. It's still possible to make it do the right thing
37.
by maciejgryka 3y ago
Oh this is fun! So you basically define personalities by picking well-known people that are probably represented in the training data and ask them (their LLM-imagined doppelganger) to vote?
38.
by maciejgryka 3y ago
I think instructor is great! And most of our Python code is typed too :) My point is just that you should care a lot about preserving optionality at the start because you're likely to have to significantly change things as you learn. I
39.
by maciejgryka 3y ago
Yep, it's a DAG, though that only occurred to me after we built this so we didn't model it that way at first. It can be the same LLM with different prompts or totally different models, I think there's no rule and it depends o
40.
by maciejgryka 3y ago
Yeah, the fact that different models might react differently to such tricks makes it hard. We're experimenting with Claude right now and I'm really hoping something like https://github.com/stanfordnlp/dspy ca
41.
by maciejgryka 3y ago
I wouldn't be surprised to see it help, along with the "you'll get $200 if you answer this right" trick and a bunch of others :) They're definitely worth trying.
42.
by maciejgryka 3y ago
Ha, we tried that! Didn't make a noticeable difference in our benchmarks, even though I've heard the same sentiment in a bunch of places. I'm guessing whether this helps or not is task-dependent.
43.
by maciejgryka 3y ago
This is a bunch of lessons we learned as we built our AI-assisted QA. I've seen a bunch of people circle around similar processes, but didn't find a single source explaining it, so thought it might be worth writing down. Super cur
44.
by maciejgryka 3y ago
I haven't read through it yet, but there's FerretUI from Apple (mobile-specific, but I think a lot of learnings are generic) https://arxiv.org/abs/2404.05719
45.
by maciejgryka 3y ago
My experience is that it's pretty good at reading text and pretty bad at understanding layouts. So e.g. asking it to work with tables is asking for trouble.
46.
by maciejgryka 3y ago
Competition is great! But in this case, I don't know, adding JSON mode and function calling was a pretty obvious next steps for the vision model - I bet it'd have happened anyway.
47.
by maciejgryka 3y ago
He, cool to hear you're doing something like this too! We ended up close, but we also need good spatial relationships, which GPT4V isn't great at, so we're using other OCR system and adding the result to the context.
48.
by maciejgryka 3y ago
It's cool to see this update and JSON mode and function calling will both be useful with vision. I wonder, though, if there were any other specific changes to the models since the `preview` versions besides that?
49.
by maciejgryka 3y ago
Congrats on the launch! I haven’t had a chance to try out Active yet, but having had a project with Erik and the team a while back, they’re a great team to work with :)
50.
Show HN: Habanero – terrible hot code deploys with Elixir(maciej.gryka.net)
1 points by maciejgryka 3y ago|0 comments
51.
by maciejgryka 3y ago
Oxide and Friends just did an episode talking about this: https://oxide.computer/podcasts/oxide-and-friends/1643335
52.
by maciejgryka 3y ago
I have similar conclusions so far. We have a custom data set (basically visual Q&A about web apps) and `gpt4` gets roughly 90% correct, while `gpt-4-1106-preview` only 86%. It's a little noisy (I didn't yet check out the new s
53.
by maciejgryka 4y ago
Shameless plug & also a point of support for the article: I used that same stack to build https://regex.help/ (more details here https://maciej.gryka.net/building-regex-help )
54.
The Follower: Using open cameras and AI to find how an Instagram photo is taken(driesdepoorter.be)
499 points by maciejgryka 4y ago|108 comments
55.
Joel Spolsky on Structuring the Web with the Block Protocol(thenewstack.io)
2 points by maciejgryka 5y ago|0 comments
56.
by maciejgryka 5y ago
If I understood what you mean, then yes, I built one https://regex.help/ (powered by https://github.com/pemistahl/grex doing the heavy lifting).
57.
Building Regex.help(maciej.gryka.net)
1 points by maciejgryka 5y ago|0 comments
58.
by maciejgryka 5y ago
Oh thanks for letting us know! Where did you find this link? It should be https://help.rainforestqa.com/
59.
by maciejgryka 5y ago
For sure, that's weird - email support@rainforestqa.com and we'll sort you out.
60.
by maciejgryka 5y ago
Yes, we are dogfooding pretty hard! We release a bunch of times a day and every release goes through a suite of >130 Rainforest tests for regression. We also add tests for every new feature we release, so it keeps growing. It was a cool

More ›