ML engineer at @StabilityAI. Formerly postdoc in philosophy of mathematics. https://teapawt.net/t.co/IIlJuT7k4G
London, England
Joined May 2018
- Tweets87
- Following675
- Followers139
- Likes615
Kate Hodesdon retweeted
My take: the founder cult in SV is totally out of control. It formalizes a de facto aristocracy as the preferred form of governance in tech. It leads to outsize discrepancies in risk taking and outcomes for most tech workers. It enables some of the most toxic psychopaths to be venerated as “visionaries”. Most of the worst people I came across in tech were “founders”.
Kate Hodesdon retweeted
I just realized men are finally comprehending the psychological impact of having a baby. No wonder everyone's gone mad
Chatting about Stable Diffusion 3 and Sora's diffusion transformers, and what they might unlock for multimodal models with Helen Byrne:
open.spotify.com/episode/7wg…
Kate Hodesdon retweeted
We have now hit 330 *million* cumulative downloads since our first model was uploaded @huggingface 18 months ago.
Just overtook @Google (321m all time), on our way to 1 billion thanks to the great efforts of the team @StabilityAI
Some announcements to come, big weeks ahead..
Stability AI is the leader in open AI models.
Open models trained on our cluster are the most downloaded globally (below @huggingface) & with what we have coming up by us & ecosystem it’s going to be a fun year ahead
Open ftw
Super proud of team, onwards 🚀 🌟
Kate Hodesdon retweeted
Thank you
Holiday project! I've started a Substack. First order of business: to unconfuse myself about some of x-risk debates around AI
open.substack.com/pub/hodesd…
Holiday project! I've started a Substack. First order of business: to unconfuse myself about some of x-risk debates around AI
open.substack.com/pub/hodesd…
What looks like the absorption of OpenAI into MSFT is a turning point for the industry. Product-wise, it's onwards and upwards: tools at least as fantastic as Copilot and ChatGPT await. But we'll look back on this as the moment when some of the magic around AI was lost.
Geoff Hinton says it is crucial that we come to a consensus about whether LLMs do or do not understand language.
He frames it as a binary issue. But the evidence shows that at most, these models have only a partial grasp of semantics.
Had an insightful conversation with @geoffreyhinton about AI and catastrophic risks. Two thoughts we want to share:
(i) It's important that AI scientists reach consensus on risks-similar to climate scientists, who have rough consensus on climate change-to shape good policy.
(ii) Do AI models understand the world? We think they do. If we list out and develop a shared view on key technical questions like this, it will help move us toward consensus on risks.
I learned a lot speaking with Geoff. Let’s all of us in AI keep having conversations to learn from each other!
Acquiring the ability to use human language isn't like a speaker of, say, English becoming fluent in Mandarin. It's more like Wittgenstein's lion acquiring the ability to speak. Famously: 'if a lion could speak, we could not understand him'.
Kate Hodesdon retweeted
Widespread GPT-4-32k access will be a bigger leap than GPT-3.5 to GPT-4 was.
The 4x larger context window combined with the power of GPT-4 will be absolutely insane.
People will be completely blown away by the complexity of workflows and use-cases that it will unlock.
Kate Hodesdon retweeted
UX x AI is not a nice interface on top of a nice model.
It is about incorporating the UX inside the model at every possible level.
6/6
From magic to infrastructure -- @rinireg compares today's LLM moment with the brief period in which electricity was a marvel, before quickly becoming unremarkable.
More philosophers' essays on LLMs here:
dailynous.com/2023/03/16/phi…