pamelafox

6 ms·

31.
by pamelafox 1y ago
I would argue that we're passing on 5% of life to the machines, not 100%. By the time bedtime has rolled around, my kids have been home for 5 hours - we have already spent hours reading, playing, parkour'ing, role-playing, paintin
32.
by pamelafox 1y ago
Gemini wrote that whole story with a short prompt about a "King Dragon that farts". I assure you that our actual improv'd story is far superior in plot points. And yes, I was confused too as to how farting would clear away fo
33.
by pamelafox 1y ago
Lol, yes, the dragon's torso turned into a man. That man does show up earlier in the story - I think perhaps the model so closely associates dragon stories with stories of men, it just desperately wanted to add one in? The text itself
34.
by pamelafox 1y ago
Lol, I just tried to get it to draw the story about King Dragon farting, but it could not come up with a picture of a dragon farting - it turned it into fire coming from its mouth instead! It's too far outside its training data. Link:
35.
by pamelafox 1y ago
I think it'd be amazing if I had the energy to make up improv bedtime stories every night. (We have a "King Dragon" improv series happening lately, which involves a lot of farts) BUT, I don't always have that energy, and
36.
by pamelafox 1y ago
Yep, 20B model, via Ollama: ollama run gpt-oss:20b Screenshot here with Ollama running and asitop in other terminal: https://bsky.app/profile/pamelafox.bsky.social/post/3lvobol3...
37.
by pamelafox 1y ago
I ran it via Ollama, which I assume uses the best way. Screenshot in my post here: https://bsky.app/profile/pamelafox.bsky.social/post/3lvobol3... I'm still wondering why my MPU usage was so low.. maybe
38.
by pamelafox 1y ago
Update: I tried it out. It took about 8 seconds per token, and didn't seem to be using much of my GPU (MPU), but was using a lot of RAM. Not a model that I could use practically on my machine.
39.
by pamelafox 1y ago
Anyone tried running on a Mac M1 with 16GB RAM yet? I've never run higher than an 8GB model, but apparently this one is specifically designed to work well with 16 GB of RAM.
40.
Red-teaming a RAG app: What happens?(blog.pamelafox.org)
1 points by pamelafox 1y ago|0 comments
41.
by pamelafox 1y ago
This is my favorite thing today! I only saw one penis fish (with the penis nestled inside the face, as a facial feature of sorts). That's pretty good for a drawing app on the internet, well done! I've given up on running public ap
42.
by pamelafox 1y ago
I ran an automated red-teaming against a RAG app using llama:3.18B, and it did really well under red-teaming, pretty similar stats to when the app was gpt-4o. I think they must have done a good at the RLHF of that model, based on my experim
43.
by pamelafox 1y ago
That's true, I am continually learning to temper my idealism, particularly when working in developer tools and education.
44.
by pamelafox 1y ago
I generally am impressed by Anthropic's focus on safety, but I was taken aback by this quote from Dario: https://bsky.app/profile/kylierobison.com/post/3lujbtfdzyk2e “Unfortunately, I think ‘no bad perso
45.
by pamelafox 1y ago
Yep, that observation is discussed frequently in the book "Insect Crisis". Highly recommend!
46.
by pamelafox 1y ago
I love native bees, I've been trying to find ways to incorporate native bee facts into my tech talks. The "Insect Crisis" book was a nice overview of issues like overuse of honeybees, plus others. Highly recommend planting na
47.
by pamelafox 1y ago
I was exaggerating slightly - I think it's some combo of the apps I use: Edge, Teams, Discord, VS Code, Docker. When I get the RAM popup once a week, I typically have to close a few of those, whichever is using the most memory accordin
48.
by pamelafox 1y ago
Are they quantized more effectively than the non-reasoning models for some reason?
49.
by pamelafox 1y ago
Alas, my 3 year old Mac has only 16 GB RAM, and can barely run a browser without running out of memory. It's a work-issued Mac, and we only get upgrades every 4/5 years. I must be content with 8B parameters models from Ollama (som
50.
by pamelafox 1y ago
(Disclosure: I work for Microsoft) I run automated red-teaming on my RAG samples through the azure-ai-evaluation SDK, which uses an adversarial LLM (an LLM without the guardrails) plus the pyrit package to come up with horrible questions to
51.
by pamelafox 1y ago
I don’t know specifically how this container was implemented, but Microsoft has a standard way to do isolated Python sandboxes: https://learn.microsoft.com/en-us/azure/container-apps/sessi... Hopefully this f
52.
Automated repository maintenance with the GitHub Copilot coding agent(blog.pamelafox.org)
3 points by pamelafox 1y ago|0 comments
53.
by pamelafox 1y ago
My personal philosophy is that TODOs should only be used while working on a branch. Once its ready for a pull request, any pending TODOs should be abandoned or logged in an issue tracker.
54.
by pamelafox 1y ago
Unhook extension: https://chromewebstore.google.com/detail/unhook-remove-youtu... That's the one I use now, it has more features than the one I made.
55.
by pamelafox 1y ago
I gave a very similar talk back in 2018: "Getting un-hooked from technology: using tech to fight tech". I wrote it up here: https://uxdesign.cc/getting-unhooked-from-technology-86ca8be... I still use a YouTube Unh
56.
by pamelafox 1y ago
Great list! I also subscribe to Gergeley Orosz' "Pragmatic Engineer" which covers many AI topics now, and to Gary Marcus' substack, which tackles topics more from an LLM skeptic perspective. https://newsletter
57.
by pamelafox 1y ago
Their plan is to offer hosted products, as described in their current job openings: https://jobs.ashbyhq.com/astral/a357ab40-9da5-4474-acc7-5888... We'll see if that works out for them, but I also worked with thei
58.
by pamelafox 1y ago
The nice thing about videos is the play/pause/slider UI. Some platforms do add play/pause explicitly to GIFs, using some JS, but as far as I know (and you would know more), that's not built into browsers yet. That's
59.
by pamelafox 1y ago
For Python apps, I've gotten good CI speedups by moving over to the astral.sh toolchain, using uv for the package installation with caching. Once I move to their type-checker instead of mypy, that'll speed the CI up even more. The
60.
by pamelafox 1y ago
Fantastic FAQ, thank you Hamel for writing it up. We had an open space on AI Evals at Pycon this year, and had lots of discussion around similar questions. I only wrote down the questions, however: # Evaluation Metrics & Methodology * W

More ›