Pinned Tweet
I now have a French-speaking account, where I will rant about #AI and various topics, avoiding English references as much as possible. I'll cross-post important things. Thanks for your attention!
Re-bonjour Twitter ! Il y a quelques mois, j'ai posté un fil en anglais à propos de l'importance de la communication scientifique en français, dans un domaine (pour moi l'#IA) où l'anglais domine. J'ouvre donc ce compte francophone. teapaw.org/gyrodiot/status/964519…
1
In my first industry job, we had a theme for model releases. Helium (v2.0), Lithium (3.0), Boron (5.0), Boracite (5.1), Kernite (5.2), Fabianite (5.3), etc.
I had a whole list prepared, never went past Graphite (6.1) before I left.
I have no idea what Google is doing.
2
145
I'm preparing the slides for an upcoming presentation. At this point I've done enough of them that most are remixes, updates, or ripped from the headlines to be eventually digested. Each one is a small rant, a figure in a routine, a joke in a set. I build a chain.
1
3
180
Sometimes, not knowing is cause for alarm ("we don't know how to make it safe"), sometimes the information isn't required for the point ("we don't know how exactly it would defeat us"), sometimes it's a real blank in the map ("we don't know whether we'll pass this law").
1
2
16
Ultimately, when speaking (publicly or not), it's about working with what you have, as @So8res says.
One talk is only one subset from the chain mail of arguments I've accumulated, and the holes keep being patched, but it's strong already.
mindingourway.com/confidence…
1
18
I will be at EAGxBerlin this weekend. In the honor of @frances__lorenz I will go to extra lengths to radiate enthusiasm in the face of hard coordination problems, support the volunteers, and be mindful of others.
I will also be very talkative, but my baseline is high.
2
7
281
I am being very talkative and enthusiastic, and working on my grant application. One perk of being at EAGx is that I'm right next to people able to give feedback on it.
Also, good food.
1
46
You can just announce things.
Such as a French-speaking AI governance talk series with an upcoming resource hub for mapping the domain.
This is the calm-and-serious side to my usual "by the way did you know there's a significant chance we can all die soon" public persona.
It's here: quellesmachines.org/
Side goal is to make it good enough that English speakers want a translation.
4
53
The Paris Call for AI safety and security!
Just signed by six countries so far, plus Anthropic, OpenAI, Mistral, Microsoft, Hugging Face and other companies. Also, @LawZero_, @SaferAI_org and more!
elysee.fr/en/emmanuel-macron…
I didn't celebrate Halloween, but I appreciate your excellent costume pictures. I don't remember seeing that many of them last year.
Maybe it's the algorithm, or who I follow. Probably both.
2
44
I have this plushie behind my screen at work. It doesn't have a mind. I smile back anyway. It is fine. Any mind I project is an echo of my own, in the same way the words on my blog echo my past selves.
I experience my own joy through this cute avatar. Others can too!
1
4
48
(for those who want one, here it is: eu.jellycat.com/amuseables-c…) (substitutes abound, I find this one better)
12
This Monday evening, live streaming some AI safety/governance news (in French). I'm much more comfortable doing it twice a month than only once, given the pace of things.
Every week could also work, if not for logistics.
2
37
Today's blog has an accumulation of pithy quotes, such as "Achievements are a symptom of happiness. Consequences of an engine running smoothly" and "I am convinced you cannot handle the truth if you're not kind to yourself"
schlaugh.com/~/XTQVdLV
1
1
35
I blame/credit the introspective question series currently running on the blog, usually I don't dare posting vaguely profound statements without context. But here... the rest of the blog is the context! So there. teapaw.org/gyrodiot/status/210542…
Over at the blog, we're starting a new run of the Life-Changing Questions Deck (based on the physical version from 2019). 36 questions, one per day.
Previous run was in 2022, I don't remember most of my answers, the differences will be reviewed at the end.
1
26
It went well! We gave ourselves extra time to answer more questions. Notable contrast between my "we are blundering into catastrophe in absurd ways" gesticulations and @FournesMaxime's "the danger is imminent, please adjust your priorities" gravitas.
In about an hour, I'll speak on a panel on AI existential risk, live on Discord/YouTube. A French volunteer collective is having an AI safety-themed month. They have invited Pause AI members today, to answer some questions.
I know we won't have time to answer everything.
1
4
113



