LightNews
gdeejay.bsky.social
Giles
@gdeejay.bsky.social
3.8K followers 3.2K following 4.1K posts
Applied economist working in international development policy 🇹🇴🇸🇧🇻🇺🇫🇯🇦🇺🇲🇲 (etc), Mostly memes. Sometimes post about #publicpolicy, research, data science, #PFM and #Rstats Economist | Independent Consultant All views my own
Posts Media Videos Starter Packs
Pinned
gdeejay.bsky.social
Giles @gdeejay.bsky.social · Oct 22
I'm torn whether to do an introduction post as I'm pretty boring.

I'm not a professor of anything or lead Oxford's AI Ethics initiative...

.. I'm just a humble independent consultant working in the fields of Public Policy, Data Analytics and Development policy.

(1/n)
...
2 26
gdeejay.bsky.social
Giles @gdeejay.bsky.social · 6h
Looking forward to giving Lossless Scaling a go on Faster than Light
Lossless Scaling logo
gdeejay.bsky.social
Giles @gdeejay.bsky.social · 12h
I asked #Claude code to show me a seahorse emoji and the best it could do is generate a picture using a python script.

See enclosed for its drawing of a seahorse:
#AI
1 3
gdeejay.bsky.social
Giles @gdeejay.bsky.social · 12h
I'm much less interested in audiophile gear since I did an A/B hearing test.
😬

www.npr.org/sections/the...
How Well Can You Hear Audio Quality?
A hi-def test for your ears (and your audio equipment): Listen to these songs and see if you can tell the difference between an MP3 and an uncompressed audio file.
www.npr.org
1
gdeejay.bsky.social
Giles @gdeejay.bsky.social · 12h
Really leveraging my max plan:
gdeejay.bsky.social
Giles @gdeejay.bsky.social · 13h
Has anyone considered putting the blockchain on the #AI?
2
gdeejay.bsky.social
Giles @gdeejay.bsky.social · 15h
Seriously! 😂
1
gdeejay.bsky.social
Giles @gdeejay.bsky.social · 17h
I used to put a post-it note on my screen with the task on it to avoid getting distracted by rabbit holes 😂
1 1
gdeejay.bsky.social
Giles @gdeejay.bsky.social · 17h
I've wondered about America's reputation as a global innovation hub too:
How much of America's size and wealth attracts talent that breeds innovation vs. its unique cultural institutions.

Presumably a lot is path dependence.
2
gdeejay.bsky.social
Giles @gdeejay.bsky.social · 17h
"...America dominates in part because it is big. As a proportion of population, the number of medal winners is higher in European countries such as Britain, Denmark, Sweden and Switzerland..."

#nobelprize

www.economist.com/graphic-deta...
Plot showing population adjusted Nobel prizes by country
1 2 6
gdeejay.bsky.social
Giles @gdeejay.bsky.social · 1d
Coase equilibrium amirite?!
alanau.bsky.social
Alan Au @alanau.bsky.social · 1d
They should pay you to not go.
1
gdeejay.bsky.social
Giles @gdeejay.bsky.social · 1d
😂

I'll suggest this when I'm negotiating my next consulting contract.
1
gdeejay.bsky.social
Giles @gdeejay.bsky.social · 1d
(One of my colleagues reckons it only happens when I'm there, so I'm the problem 😂.)
1
gdeejay.bsky.social
Giles @gdeejay.bsky.social · 1d
I've decided I need to charge danger pay for these consulting projects.

Every time I do them we discover some epic problem with the budget papers that messes up our carefully laid plans.

😬
1 4
gdeejay.bsky.social
Giles @gdeejay.bsky.social · 1d
(I'd absolutely do this if I could 🧐)
gdeejay.bsky.social
Giles @gdeejay.bsky.social · 1d
Yeah, I worked with a guy that had it on his CV for being part of the UN peacekeepers.

🎗️
1
gdeejay.bsky.social
Giles @gdeejay.bsky.social · 1d
Looking forward to hearing if I get the Nobel Peace prize 🤞🏼
2 6
gdeejay.bsky.social
Giles @gdeejay.bsky.social · 1d
And the fact that if you're working outside your field you'll be less familiar with what to search for and how to evaluate the results.
1
gdeejay.bsky.social
Giles @gdeejay.bsky.social · 1d
I can also relate to the problem of tearing my hair out when companies don't hire experts *and* not having access to the resources I'd like to make the appropriate investments into outside expertise.
1
gdeejay.bsky.social
Giles @gdeejay.bsky.social · 1d
The "looks reasonable" problem is something that concerns me and I warn others about.

LLMs are fundamentally predictive text models, so they always produce something that seems *roughly* correct. Which is a real problem when you're working outside your wheelhouse.
1 1
gdeejay.bsky.social
Giles @gdeejay.bsky.social · 1d
Great post:
"...ChatGPT does something that looks familiar, the search results are consistent with your ideas, so you’re happy with the output. Would an actual statistical expert give you different advice? You’ll never know because ChatGPT is your expert now..."
bharrap.bsky.social
Ben Harrap @bharrap.bsky.social · 1d
I clearly had a lot of thoughts/feelings from reading Terry's blog post, so I wrote one of my own

It ended up being a bit of a stream-of-thought post but I think it captures some of my worries

benharrap.com/post/2025-10...

#statssky #episky #academicsky
2 5
gdeejay.bsky.social
Giles @gdeejay.bsky.social · 1d
Gross.
gdeejay.bsky.social
Giles @gdeejay.bsky.social · 1d
Yeah! This is why I first gave it a go.

Spyder just didn't do it for me.
1 1
gdeejay.bsky.social
Giles @gdeejay.bsky.social · 2d
This is great!

I've found Claude good for creating R code to reproduce a plot I provide it with.

I'll do this when I don't know what a chart is called and can't be bothered to work out the color palette.

😬
1
gdeejay.bsky.social
Giles @gdeejay.bsky.social · 2d
A meme showing a speech bubble from an anthropomorphized paperclip character (resembling Microsoft's Clippy) asking "Hey, it looks like you're waiting for ideal conditions to do that thing you've been wanting to do. Need I remind you that ideal conditions do not exist and will never happen?" with two response options below: "I'm aware" and "Wow, rude." The paperclip character appears in the bottom right corner with googly eyes attached to it.
3 17 64
gdeejay.bsky.social
Giles @gdeejay.bsky.social · 2d
I find it works well as a rolodex.

As a professional platform: 👎🏼
1