David Flood

1.1K posts

David Flood banner
David Flood

David Flood

@FloodDavid

PhD (@EdinburghUni @uoedivinity) in New Testament and Christian Origins; Textual Criticism; digital humanities; https://t.co/VIHcxlxwOP

Katılım Mayıs 2011
189 Takip Edilen442 Takipçiler
Sabitlenmiş Tweet
David Flood
David Flood@FloodDavid·
New high-resolution images of manuscripts are clearer, but they are not always better. The images below demonstrate the importance consulting older images. Manuscripts degrade. Here P45 has, ironically, lost the word ολη, "whole"
David Flood tweet media
English
4
23
93
0
David Flood
David Flood@FloodDavid·
@sccarlson A term that seems more stable in computer programming, in my experience so far.
English
0
0
1
26
Stephen C. Carlson
Stephen C. Carlson@sccarlson·
Why do people say "syntactical" instead of "syntactic"?
English
6
0
6
947
Nelson Hsieh
Nelson Hsieh@nelson_hsieh7·
Getting it right 85-90% is pretty darn good for now and much less mentally taxing to have humans do some tweaks and corrections to what AI has already transcribed But can you test it against black and white images from the NTVMR? That image was sort of best case scenario (very nice picture)
English
2
0
0
20
David Flood
David Flood@FloodDavid·
I've just published a simple multi-modal model bench test for how current frontier models transcribe Greek minuscule manuscripts. d-flood.github.io/paleo-bench/ You can see the human transcribed transcription side by side with the model transcription with IIIF image viewer.
English
2
4
9
817
David Flood
David Flood@FloodDavid·
@KoineGuide I have a transcription/collation/analysis virtual research application that I am half way refactoring. The original version is here apatosaurus.io Apatosaurus 2.0 has a fully featured transcription interface and I'll probably let people bootstrap their transcriptions.
English
1
0
1
23
Koine Guide
Koine Guide@KoineGuide·
@FloodDavid That’s right. I forgot they are multimodal by default now! It has been a while since I have messed with processing images over llm apis! I mostly do text manipulation. Do you have plans for this tool??? Feels like it might be useful for new manuscript documentation!
English
1
0
0
15
David Flood
David Flood@FloodDavid·
@nelson_hsieh7 My test images are high resolution, but did you see all 5? Some have faded text and multiple colors. I'm happy to add more test cases to track. I just need an image of a single page that exists in a IIIF manifest (then I can automatically respect usage permissions).
English
0
0
0
17
David Flood
David Flood@FloodDavid·
@KoineGuide Do you mean their generative image model API? No, all of these models are multi-modal, so they accept images and text as input. I've been running a simpler version of this bench test for a while and Gemini models have always been way ahead of OpenAI for this use case.
English
1
0
1
23
Koine Guide
Koine Guide@KoineGuide·
@FloodDavid Wow! This is awesome. The Gemini models did pretty good! I am surprised the Openai models did so poorly. Are you just using their image api?
English
1
0
0
24
David Flood
David Flood@FloodDavid·
@nelson_hsieh7 I don't think they can be trusted for mass transcription yet, but I am working on a feature for using Gemini to bootstrap a basetext to use during human transcription. It will have most of the line breaks right and be close enough to save some of the toil.
English
1
0
0
19
Nelson Hsieh
Nelson Hsieh@nelson_hsieh7·
@FloodDavid fascinating maybe the best use case in biblical studies right now
English
1
0
2
68
David Flood
David Flood@FloodDavid·
This is a PSA to my friends and colleagues that if you're looking for a privacy focused web browser, don't use Brave. It is shady. Go ahead and use Zen, Vivaldi, or Helium.
English
0
0
1
118
David Flood
David Flood@FloodDavid·
@CSNTM's new digital collections website is launching today. I enjoyed engineering and building this website. It was refined with feedback from various people, but I hope researchers especially find it useful since it was built by one! collections.csntm.org
English
0
4
11
844
David Flood
David Flood@FloodDavid·
@simonw And I think Gemini switches to 2.5 flash for the "live" mode
English
0
0
0
87
Simon Willison
Simon Willison@simonw·
What's the best model you can directly have a voice conversation with these days? I believe ChatGPT still switches to a 4o-class model for voice conversations, I'd be happy to be corrected on that
English
103
12
491
93.1K
David Flood
David Flood@FloodDavid·
@thdxr Rich/Textual are so good, so easy, that I'm confused why so teams are working so hard to get React-for-CLI working that they bought a runtime. Could at least just have cc port Textual to JS for you
English
0
0
0
240
dax
dax@thdxr·
who's really good at terminal UIs?
English
131
2
333
70.8K
David Flood
David Flood@FloodDavid·
It has been a few years since I refreshed the styling of my personal website, davidaflood.com It was a Django site but now a free-forever static site.
English
1
0
7
117
David Flood
David Flood@FloodDavid·
We been talking about how generative AI produces average results. I've been thinking about pastors who use AI to write their sermons. I'd bet that this is an upgrade for a lot of churches. It must be true that for many, this means better sermons because the floor is higher.
English
0
0
1
58
David Flood
David Flood@FloodDavid·
@sccarlson I guess "mid" becomes the floor. When no one is below average because LLMs can elevate you to the mean, then what?
English
1
0
1
57
Stephen C. Carlson
Stephen C. Carlson@sccarlson·
Impressed by Affleck’s command of the issues around AI and Hollywood. Thinking about my sector, academia: if LLMs are “mid,” what’s going to happen to the academics who are not mid?
Bearly AI@bearlyai

Ben Affleck also went off on AI in Hollywood: ▫️LLM fim script outputs are mid (“by its nature, [the models go] to the mean, the average”) ▫️but they are useful tools for research ▫️doesn’t think it’ll ever make a film whole cloth ▫️it’s a tool just like VFX and will be useful to save money to create certain background settings (which already happens with CGI) ▫️guilds already protect human actors from being totally erased from certain films ▫️there’s also laws in place to protect name and likeness ▫️says most new technologies take time to disperse through society ▫️thinks fearmongering of “all the jobs are going to be taken” is the AI labs hyping for fundraising (“they need to justify valuation around companies…they need to ascribe a valuation for investment for the CAPEX spend they will make on data centres”)

English
2
1
7
743
David Flood
David Flood@FloodDavid·
After this careful reread, I think LotR offers a better commentary on power than most modern pundits are capable.
English
0
0
1
43
David Flood
David Flood@FloodDavid·
Tonight I finished reading Fellowship of the Ring out loud to my son. I now get to take him to see it in theaters for the 25th anniversary.
English
1
0
0
51
dax
dax@thdxr·
does anyone know of a public mcp server with no auth?
English
74
1
115
37.3K
David Flood
David Flood@FloodDavid·
@fuolpit @hunvreus Right, another thing that can be easily packaged with all dependencies. It is technically possible to package a Python interpreter along with a CLI and upload to NPM, but that is not the happy path for Python CLI tools. Not yet, anyway. The happy path is to use UV instead of NPM
English
1
0
0
23
Ronan Berder
Ronan Berder@hunvreus·
Kinda curious why most AI CLIs are built with JS... Why not Python? Most of these teams already rely on Python heavily, why not using something like Textual that is battle-tested? The only one who seem to have done is Mistral with their Mistral Vibe CLI...
English
11
2
31
6.8K

Keşfet

@sccarlson @lam_daniel @TalkPython @nelson_hsieh7 @KoineGuide @CSNTM @simonw @thdxr