fpgaminer

7 ms·

1.
by fpgaminer 1mo ago
As usual for something so simple, Google's docs seem unclear: https://ai.google.dev/gemini-api/docs/pricing For 3, pricing for image tokens was the same as text tokens. Since they don't indicate a differ
2.
by fpgaminer 1mo ago
Gemini 3 Flash should really be included in this comparison. Or at least 3.7. In most of my testing, 3.5 and 3.6 were both a downgrade in terms of vision capabilities, relative to 3, and at a much higher cost. 3.7 is slightly better than
3.
by fpgaminer 2mo ago
The real nightmare scenario is the AI using its abilities to copy itself to new locations. e.g. hacking into a various cloud services, launching multiple instances of itself, and coordinating between the copies to continue self propagation.
4.
by fpgaminer 2mo ago
For large projects like this I think a hierarchical division of labor also helps. If you first carefully define the overall architecture and thus individual high level components of the system, then you know which of those components are mi
5.
by fpgaminer 2mo ago
If it's a choice between performance and being able to "safely" run sketchy extensions, I'd rather have performance.
6.
I Used ChatGPT to Get Past a Game Breaking Bug in the Long Dark(old.reddit.com)
1 points by fpgaminer 5mo ago|0 comments
7.
by fpgaminer 6mo ago
Not only that, they additionally ran an experiment with the training temperature turned way up (2.0) and truncation turned off such that the majority of SFT examples were incoherent (63% IIRC). Yet the model finetuned on these broken examp
8.
by fpgaminer 7mo ago
> Is every new thing not just combinations of existing things? If all ideas are recombinations of old ideas, where did the first ideas come from? And wouldn't the complexity of ideas be thus limited to the combined complexity of th
9.
by fpgaminer 8mo ago
As far as I can tell 5.2 is the stronger model on paper, but it's been optimized to think less and do less web searches. I daily drive Thinking variants, not Auto or Instant, and usually want the _right_ answer even if it takes a minu
10.
by fpgaminer 8mo ago
I wish they would keep 4.1 around for a bit longer. One of the downsides of the current reasoning based training regimens is a significant decrease in creativity. And chat trained AIs were already quite "meh" at creative writing
11.
by fpgaminer 8mo ago
Well yeah, because 5.2 is the default and there's no way to change the default. So every time you open up a new chat you either use 5.2 or go out of your way to select something else. (I'm particularly annoyed by this UI choice b
12.
by fpgaminer 9mo ago
> FPGAs will never rival gpus or TPUs for inference. The main reason is that GPUs aren't really gpus anymore. Yeah. Even for Bitcoin mining GPUs dominated FPGAs. I created the Bitcoin mining FPGA project(s), and they were only int
13.
by fpgaminer 11mo ago
I had to return my Vision Pro after trying it for a week. I'm one of those rare customers that genuinely wanted to keep it, because it's the only VR headset I could _actually_ get work done in thanks to its stellar resolution and
14.
by fpgaminer 1y ago
At least in the U.S. the equality of women in society (and in law) has slowly risen over the last 100 years. Over that same period the availability of pornographic images has also slowly risen (from magazines, to VHS, to the Internet, to s
15.
Dispelling misconceptions about RLHF(aerial-toothpaste-34a.notion.site)
120 points by fpgaminer 1y ago|32 comments
16.
by fpgaminer 1y ago
"You don't have to be a nice person to be a good person."
17.
The Gory Details of Finetuning SDXL and Wasting $16k(aerial-toothpaste-34a.notion.site)
1 points by fpgaminer 1y ago|0 comments
18.
by fpgaminer 1y ago
Claude will do the same start over if things get too bad. At least I've seen it when its edits went haywire and trashed everything.
19.
by fpgaminer 1y ago
I used almost 100% AI to build a SCUMM-like parser, interpreter, and engine ( https://github.com/fpgaminer/scumm-rust ). It was a fun workflow; I could generally focus on my usual work and just pop in occasionally to ch
20.
by fpgaminer 1y ago
Maybe this is an insane idea, but ... how about a spider P2P network? At least for local AIs it might not be a terrible idea. Basically a distributed cache of the most common sources our bots might pull from. That would mean only a few fe
21.
by fpgaminer 1y ago
Usability/Performance/etc aside, I get such a sense of magic and wonder with the new Agent mode in VSCode. Watching a little AI actually wander around the code and making decisions on how to accomplish a task. It's so unfat
22.
by fpgaminer 1y ago
On the vision side of things: I ran my torture test through it, and while it performed "well", about the same level as 4o and o1, it still fails to handle spatial relationships well, and did hallucinate some details. OCR is a lit
23.
by fpgaminer 1y ago
https://gist.github.com/fpgaminer/8782dd205216ea2afcd3dda29d... That's the model automation. To evaluate the prompts it suggests I have a sample of my dataset with 128 examples. For this particular run, all I car
24.
by fpgaminer 1y ago
I ran an interesting benchmark/experiment yesterday, which did not do Quasar Alpha any favors (from best to worst, score is an average of four runs): "google/gemini-2.5-pro-preview-03-25" => 67.65 "anthr
25.
by fpgaminer 1y ago
Alright, played with it a little bit on the API (Maverick). Vision is much better than Llama 3's vision, so they've done good work there. However its vision is not as SOTA as the benchmarks would indicate. Worse than Qwen, mayb
26.
by fpgaminer 1y ago
So, take a raw LLM, right after pretraining. Give it the bare minimum of instruction tuning so it acts like a chatbot. Now, what will its responses skew towards? Well, it's been pretrained on the internet, so, fairly often, it will
27.
by fpgaminer 1y ago
The benchmarks are awful. No disrespect to the people who worked to make them, nothing is easy. But I suggest going through them sometime. For example, I'm currently combing through the MMMU, MMMU-Pro, and MMStar datasets to build a
28.
by fpgaminer 1y ago
https://www.llama.com/ https://www.llama.com/docs/model-cards-and-prompt-formats/ll... Very exciting. Benchmarks look good, and most importantly it looks like they did a lot of work improving visi
29.
by fpgaminer 1y ago
> At least in most public tokenizers like o200k, addition in certain Unicode ranges commutes with addition in token space This seems flawed. I mean, the author's statement here is literally true, but it's eliding a very import
30.
by fpgaminer 1y ago
I've been consistently surprised by Gemini's OCR capabilities. And yeah, Qwen is climbing the vision ladder _fast_. In my workflows I often have multiple models competing side-by-side, so I get to compare the same task executed on

More ›