mickeybeurskens Profile
mickeybeurskens

@mickeybeurskens

Followers
15
Following
85
Statuses
85

🇳🇱 AI Engineer | 🔥 Writing https://t.co/J3Mp17b5iG

Joined August 2012
Don't wanna be here? Send us removal request.
@mickeybeurskens
mickeybeurskens
17 days
@carlolepelaars @jrosenfeld13 Did not see this, will send you dm
0
0
1
@mickeybeurskens
mickeybeurskens
2 months
RT @tegmark: Long live AI surveillance - embrace the Microsoft telescreen!
0
14
0
@mickeybeurskens
mickeybeurskens
3 months
@carlolepelaars @paulg Gifting books to others is great, too
0
0
1
@mickeybeurskens
mickeybeurskens
4 months
Daily reminder that money might not be valuable in a post AGI world
Tweet media one
0
0
0
@mickeybeurskens
mickeybeurskens
4 months
What is your experience coding with ContinueDev? Looking for local, open source setups and this looks promising.
0
0
0
@mickeybeurskens
mickeybeurskens
5 months
0
0
0
@mickeybeurskens
mickeybeurskens
5 months
@carlolepelaars Seems to me that o1 could be the result mainly of restructuring data. More explicitly using prompt chains and a focus on STEM data. What is your take?
1
0
1
@mickeybeurskens
mickeybeurskens
5 months
@carlolepelaars Seems o1 is closer to understanding its own compute limits 👀
0
0
0
@mickeybeurskens
mickeybeurskens
5 months
RT @Hamptonism: Convolution Neural Network Visualization
0
236
0
@mickeybeurskens
mickeybeurskens
5 months
@carlolepelaars @JeremyNguyenPhD @elder_plinius @bramhooimeijer I was thinking more along the lines of training some kind of model to specifically craft prompts that elicit unwanted behavior. The meta prompting approach is a way to generate hypothesis in my mind, but this dataset consists of those that pass the test.
0
0
1
@mickeybeurskens
mickeybeurskens
5 months
Playing around with Stable Diffusion and the Practical Deep Learning for Coders part 2 lectures by @jeremyphoward. Upping my DnD game might take some more time, but it's a start.
Tweet media one
Tweet media two
0
0
2
@mickeybeurskens
mickeybeurskens
5 months
The data would be quite interesting to see, especially because it would allow some insight into shared vulnerabilities across models. I guess this is a dynamic to take into account for any publicly deployed machine learning model, just hadn't explicitly thought about it.
0
0
0
@mickeybeurskens
mickeybeurskens
5 months
RT @anothercohen: Apple announcing the changes from iPhone 15 to iPhone 16
Tweet media one
0
145
0
@mickeybeurskens
mickeybeurskens
5 months
RT @nntaleb: IN CONSTRUCTION Human beings are only healthy when they are still in construction (intellectually). Aging starts when they bec…
0
1K
0
@mickeybeurskens
mickeybeurskens
5 months
@carlolepelaars Exactly! How much compute per kilogram do _you_ have?
1
0
1
@mickeybeurskens
mickeybeurskens
5 months
@carlolepelaars @jeremyphoward @bramhooimeijer Have you checked the repo? I heard yesterday that both it and the book are great!
1
0
1
@mickeybeurskens
mickeybeurskens
5 months
In case you want to learn about machine learning and don't know about this course by @jeremyphoward, I highly recommend taking a look.
0
0
0