Thibaut Lienart

259 posts

Thibaut Lienart

Thibaut Lienart

@t_lienart

Amz; machine learning, privacy, julialang ftw.

Europe Katılım Aralık 2016
69 Takip Edilen344 Takipçiler
Thibaut Lienart
Thibaut Lienart@t_lienart·
@SeBayesian Just in case: make sure you’re always on the ‘following’ tab and not the crappy ‘for you’ one
English
0
0
1
69
Thibaut Lienart retweetledi
Arvind Narayanan
Arvind Narayanan@random_walker·
In a conversation with @JuliaAngwin about AI hype, I argue that the most worrisome uses of ChatGPT aren't malicious ones but rather everyday people and organizations using it to cut corners due to the pressures they face, like CNET's misguided experiment. themarkup.org/hello-world/20…
English
5
67
219
44.2K
Thibaut Lienart retweetledi
Ferenc Huszár
Ferenc Huszár@fhuszar·
“All our machine learning projects undergo review by the AI ethics team”
English
0
4
27
0
Thibaut Lienart retweetledi
Ferenc Huszár
Ferenc Huszár@fhuszar·
learning theory vs deep learning
English
40
356
2.2K
0
Thibaut Lienart retweetledi
Dino Sejdinovic
Dino Sejdinovic@sejDino·
Sobering read about rash use of machine learning in healthcare: how basic methodological flaws are leading to a credibility crisis and are indicative of systemic challenges facing ML research more broadly. statnews.com/2021/06/02/mac…
English
0
3
7
0
Thibaut Lienart retweetledi
Keno Fischer
Keno Fischer@KenoFischer·
Very happy that so many websites in the #julialang ecosystem have standardized on Franklin, so I know I can always just `Franklin.serve()` them to see a live preview. In the old days, there was always a 20 minute hunt to figure out how to install the ssg-de-jour.
English
0
7
49
0
Thibaut Lienart retweetledi
Arthur Spirling
Arthur Spirling@arthur_spirling·
just trying to keep my teaching materials current
Arthur Spirling tweet media
English
5
42
397
0
Thibaut Lienart
Thibaut Lienart@t_lienart·
@oxinabox_frames @Viral_B_Shah Fair enough :) the parser is only about a third of the iceberg though, live rendering, Julia code block evaluations etc... (but you might not want any of this in which case, indeed, you’d be fine with your own stuff and possibly LiveServer.jl for the local rendering)
English
0
0
2
0
Frames Catherine White
Frames Catherine White@oxinabox_frames·
@t_lienart @Viral_B_Shah Yeah, we talked about this. It's not worth it. I'll have to hook up my own markdown parser (means BinaryBuilding cmark). And effectively bypassing most of Franklin. At which point, why use it? It's good for me to use other things. Stretches the mind etc.
English
1
0
1
0
Frames Catherine White
Frames Catherine White@oxinabox_frames·
Every now and again I think about redoing my website onto a static site generator I don't hate (currently GitHub Pages's ancient version of Jekyll). Key requirement: full compliance with CommonMark spec, including inline HTML (this rules out Franklin.jl) What should I use?
English
2
0
2
0
Thibaut Lienart
Thibaut Lienart@t_lienart·
@Viral_B_Shah @oxinabox_frames What if you could use Franklin with your own md->html pass? (which you could do via the pandoc_jll for instance); I’d be happy to help you set this up if you’d like
English
1
0
0
0
Viral B. Shah
Viral B. Shah@Viral_B_Shah·
@oxinabox_frames My experience with jekyll has not been the greatest - mostly because I don't understand how ruby packages work on os x. Hugo seemed a lot easier to work with, and of course, now, I have worked with @t_lienart to migrate several #julialang sites to Franklin.
English
1
0
2
0
Thibaut Lienart
Thibaut Lienart@t_lienart·
@adeil_araujo You likely didn’t set the prepath correctly (check the docs) please kindly ask questions on the Julia Slack (channel #franklin) thanks!
English
0
0
0
0
Thibaut Lienart retweetledi
Roger Grosse
Roger Grosse@RogerGrosse·
An ML analogy: Occam's Razor is a useful guiding principle, and regularizing the description length can help generalization. But premature minimization of description length leads to posterior collapse.
Roger Grosse@RogerGrosse

Tip for submitting deep learning papers: If your selling point is that your method is simple and easy to implement, you still need to show how it overcomes the obstacles that made other people propose solutions which are complicated and hard to implement.

English
0
2
47
0