Steven Dakin

2.4K posts

Steven Dakin banner
Steven Dakin

Steven Dakin

@StevenDakin

Work on how brains do vision and how we can use that knowledge to develop better vision tests and treatments. (He/him)

Auckland, New Zealand Katılım Haziran 2011
250 Takip Edilen1.1K Takipçiler
Sabitlenmiş Tweet
Steven Dakin
Steven Dakin@StevenDakin·
Lisa Hamm's movies ensure children can name our new eye-chart symbols (designed to better detect kids' vision problems). Thanks @CureKidsNZ
GIF
English
1
10
52
0
Gavin Buckingham
Gavin Buckingham@DrGBuckingham·
PEOPLE, STOP ANNOYING THE ROBOTS
English
10
0
24
4.7K
Steven Dakin retweetledi
Shubham Saboo
Shubham Saboo@Saboo_Shubham_·
BREAKING: The future of object detection is here! Meet YOLO-NAS, open-sourced YOLO-based architecture redefining state-of-the-art in object detection.
English
61
598
3.1K
726.3K
Steven Dakin
Steven Dakin@StevenDakin·
@robsek My guess would be that @geoffreyhinton is concerned by the unexpected set of abilities exhibited by latest chatgpt (e.g. passing “theory of mind” tests) which is shocking given the architecture.
English
0
0
0
39
Steven Dakin retweetledi
Patricia Churchland
Patricia Churchland@patchurchland·
By far the best discussion I have read -- shows how to really test GPT-3,4: On Artifice and Intelligence by Shlomi Sher link.medium.com/Di4ChhP0czb
English
5
30
103
19.7K
Steven Dakin retweetledi
Tivadar Danka
Tivadar Danka@TivadarDanka·
I described some of the most beautiful and famous mathematical theorems to Midjourney. Here is how it imagined them: 1. "The set of real numbers is uncountably infinite."
Tivadar Danka tweet media
English
147
2.2K
10.1K
2.9M
Steven Dakin
Steven Dakin@StevenDakin·
@patrickmineault So GPT4 is less consistent than people are at judging if grandmothers are more similar to wednesdays or to thermostats? Is that task really probing common sense or our ability to answer a stupid question?
English
0
0
0
51
Patrick Mineault
Patrick Mineault@patrickmineault·
🤖🧠 Does GPT-4 have common sense? In this blog post, I explore GPT-4's understanding of everyday objects and its implications for AI alignment. TL;DR: it's not very grounded in reality (thread) 1/ xcorr.net/2023/03/30/doe…
English
3
11
69
16.1K
Steven Dakin retweetledi
Michael Bach
Michael Bach@MichaelBach99·
FrACT (Freiburg Vision Tests) now also as a WebApp. This magically avoids installing, yet you get a full app. (Nearly) all platforms! And remains free. #webApp" target="_blank" rel="nofollow noopener">michaelbach.de/fract/#webApp
Michael Bach tweet media
English
1
13
24
3.8K
Steven Dakin
Steven Dakin@StevenDakin·
@neuralengine Science here has been chronically under-funded compared to the UK/US - this is amazing (and quite unexpected) news.
English
0
0
5
2.1K
Steven Dakin
Steven Dakin@StevenDakin·
I kind of assumed that a big driver of @elonmusk’s recent behaviour was generation of public engagement in issues relating to twitter. Banning journos stifles debate and indicates that this not so. #byebyetwitter @stevendakin@mastodon.nz
English
0
0
1
405
Steven Dakin retweetledi
Anna Wexler
Anna Wexler@anna_wexler·
Some thoughts on yesterday’s @neuralink “show and tell” from an ethics & society perspective 🧵 1/
English
4
29
117
0
Steven Dakin retweetledi
Michael Bethke @Bethkem@mstdn.ca
I made Thanksgiving dinner and now the kids are crying and I'm not allowed to help anymore
Michael Bethke @Bethkem@mstdn.ca tweet media
English
123
660
4.4K
0