michaelgiba

5 ms·

1.
by michaelgiba 9mo ago
> You surely aren't implying that the model is sentient or has any "desire" to give an answer, right? The model is a probabilistic machine that was trained to generate completions and then fine tuned to generate chat style
2.
by michaelgiba 9mo ago
It’s not surprising that there could be a very slight quality drop off for making the model return its answer in a constrained way. You’re essentially forcing the model to express the actual answer it wants to express in a constrained langu
3.
by michaelgiba 10mo ago
73% of startups are just writing computer programs
4.
by michaelgiba 11mo ago
Interesting idea. Although I wouldn't consider `but restrict the data set to publications from <= year 1600` "easy". If you did have access to a high-quality pretraining dataset and you could explore training up to 1600, t
5.
by michaelgiba 11mo ago
Crazier ideas would be: - extend the concept to also have some sort of “agent mode” where the llamafiles can launch with their own minimal file system or isolated context - detailed profiling of main supported models to ensure deterministi
6.
by michaelgiba 11mo ago
I’m glad to see llamafile being resurrected. A few things I hope for: 1. Curate a continuously extended inventory of prebuilt llamafiles for models as they are released 2. Create both flexible builds (with dynamic backend loading for cpu an
7.
by michaelgiba 11mo ago
They stopped publishing images, not like they changed anything significant about the product itself. Frankly the whole thing is not newsworthy
8.
by michaelgiba 1y ago
For anyone curious here is an interactive write up about this http://michaelgiba.com/grammar-based/index.html
9.
Fictitious Telephone Numbers(en.wikipedia.org)
1 points by michaelgiba 1y ago|1 comments
10.
Oh Snap - An Interactive Article on Quantization Error(michaelgiba.com)
2 points by michaelgiba 1y ago|0 comments
11.
by michaelgiba 1y ago
This is much more thorough, but here is an interactive post covering the related topic constrained sampling I put together a few weeks back: http://michaelgiba.com/grammar-based/index.html
12.
An Interactive Overview of Grammar-Based Sampling for LLMs(michaelgiba.com)
3 points by michaelgiba 1y ago|0 comments
13.
Show HN: Traitorous Models- Reality Show with Open Source LLMs(github.com)
1 points by michaelgiba 1y ago|0 comments
14.
by michaelgiba 1y ago
I was inspired by your project to start making similar multi-agent reality simulations. I’m starting with the reality game “The Traitors” because it has interesting dynamics. https://github.com/michaelgiba/survivor (el
15.
Show HN: Tiny Python+Preact tool for debugging agents(github.com)
1 points by michaelgiba 1y ago|0 comments
16.
by michaelgiba 1y ago
Nice, I’m particularly excited for the tiny models.
17.
by michaelgiba 1y ago
I like the idea but I would hesitate to upload my API keys. why not make the prompting/orchestration pieces open source? A user could run locally to generate a result and the app could focus on displaying the results in a fun way
18.
by michaelgiba 2y ago
A different, darker way to interpret this is computers cannot be held accountable today If systems (presumably AI-based) were conscious or self-aware they would very much be incentivized not to make mistakes. (Not advocating for this)
19.
by michaelgiba 2y ago
Gemini has had this for a month or two, also named "Deep Research" https://blog.google/products/gemini/google-gemini-deep-resea... Meta question: what's with all of the naming overlap in the AI worl
20.
Show HN: Convert a link to a late night show(github.com)
2 points by michaelgiba 2y ago|0 comments
21.
eBay Blames Sun for Outages (1999)(wired.com)
1 points by michaelgiba 2y ago|0 comments
22.
by michaelgiba 2y ago
This is art
23.
by michaelgiba 2y ago
it’s pretty impressive that PyTorch is only 7% slower than this given it can be used so generally
24.
by michaelgiba 2y ago
I am not sure either. Although maybe it just comes down to how the purchased compute is “delivered”
25.
by michaelgiba 2y ago
That would certainly be an obstacle. Hypothecally it could be beneficial to other smaller providers. I think it all comes down to how the purchased compute would end up being used by purchasers
26.
by michaelgiba 2y ago
Sort of, gpulist.ai as well
27.
by michaelgiba 2y ago
There are definitely thresholds you reach that make it difficult to generalize. large cluster setups can differ significantly and actual usability of the clusters makes a big difference too. However I guess there are some differences like t
28.
by michaelgiba 2y ago
I actually went down nearly this exact same rabbit hole recently. Specifically I was curious why, given all of the demand for GPUs, compute not been commoditized and resold through public markets. This also led me to the Enron's ideas
29.
by michaelgiba 2y ago
Ok I updated the main view to show latest pricing by defaultinstead of just the dropdown. Thanks!
30.
by michaelgiba 2y ago
Yeah that is all it is currently, just a drop down that once you pick a GPU configuration it shows the pricing from a few providers. I'm still trying to determine which data would be most useful to collect/display. As for the tech

More ›