Matthias Plappert
banner
matthiasplappert.com
Matthias Plappert
@matthiasplappert.com
I sometimes like AI and computers. I always like dogs. Ex OpenAI and @github.com

Website: matthiasplappert.com
Reposted by Matthias Plappert
How sure are we that the problem is only a software problem and not also a hardware limitation?
February 28, 2025 at 4:19 AM
I'm actually very confused by the GPT-4.5 release
February 28, 2025 at 11:26 AM
We still don’t have broadly available self-driving cars but somehow humanoid robots for our homes are imminent. Yeah sure 🙄
February 22, 2025 at 9:47 AM
I’m actually quite bullish on Europe: I think it’s culturally well equipped to deal with the onset of increasing automation due to AI. We already have strong social security systems that can buffer the disruptions and peoples lives are less focused on work.
February 6, 2025 at 3:23 AM
Essays like the one from Dario (darioamodei.com/on-deepseek-...) miss the point on why DeepSeek matters so much to investors. The key issue for OpenAI and Anthropic (and any other VC model company in this space) is that they absolutely require outside investments. (1/n)
Dario Amodei — On DeepSeek and Export Controls
On DeepSeek and Export Controls
darioamodei.com
January 30, 2025 at 8:31 AM
Reposted by Matthias Plappert
DeepSeek R1 appears to be a VERY strong model for coding - examples for both C and Python here: https://simonwillison.net/2025/Jan/27/llamacpp-pr/
ggml : x2 speed for WASM by optimizing SIMD
PR by Xuan-Son Nguyen for `llama.cpp`: > This PR provides a big jump in speed for WASM by leveraging SIMD instructions for `qX_K_q8_K` and `qX_0_q8_0` dot product functions. > > …
simonwillison.net
January 27, 2025 at 6:35 PM
Finally got around to read the DeepSeek-v3 and r1 papers in detail and I’m very impressed. They keep things simple and pragmatic but go deep and optimize when worthwhile.

I think access to compute remains key but they demonstrate how much more can be squeezed out when faced with constraints.
January 26, 2025 at 12:11 PM
It’s almost as if you build a toxic hell-hole of misinformation, conspiracy theories and crypto scams, people don’t want to be there. Who would’ve thought 🤷‍♂️
In email to X staff, Musk says "user growth is stagnant, revenue is unimpressive, and we're barely breaking even"; sources: banks plan to sell their debt in X (Wall Street Journal)

Main Link | Techmeme Permalink
January 24, 2025 at 9:54 PM
Reposted by Matthias Plappert
Ladies and gentlemen: we did it
In email to X staff, Musk says "user growth is stagnant, revenue is unimpressive, and we're barely breaking even"; sources: banks plan to sell their debt in X (Wall Street Journal)

Main Link | Techmeme Permalink
January 24, 2025 at 8:52 PM
Excellent (old but still very relevant) article by @molly.wiki on crypto market caps and the pitfalls with these numbers.

So if your friends or the mainstream media say things like “Trump made $10B in a few days” you can be the person that goes “Well, actually …”
Cryptocurrency 'market caps' and notional value
A naive equation works well when estimating the value of small quantities of popular crypto tokens, but it's also being used to estimate values of illiquid tokens and the market capitalization of all ...
blog.mollywhite.net
January 20, 2025 at 2:20 PM
Deleted (or I guess deactivated 🤷‍♂️) my Twitter / X account yesterday. I no longer wish to be associated with a platform that is this toxic. Quite happy that I finally pulled the plug.
December 22, 2024 at 3:37 PM
I can now solve IMO problems thanks to AI but if I have to do any chore (schedule a meeting, make a reservation at a restaurant, collect and forward invoices to my accountant) it’s completely on me. This obviously sucks since I never want to solve IMO problems in the first place.
December 22, 2024 at 9:31 AM
The year is 2024. The W&B UI is still clunky, buggy, slow and overall terrible. Such a useful service but the frontend web UI is really letting it down massively.
December 17, 2024 at 2:44 PM
Incredible work, super impressed. Also that rate of improvement 🤯
Introducing 🧞Genie 2 🧞 - our most capable large-scale foundation world model, which can generate a diverse array of consistent worlds, playable for up to a minute. We believe Genie 2 could unlock the next wave of capabilities for embodied agents 🧠.
December 5, 2024 at 8:35 AM
I don’t use X anymore because it’s a shitty product, not because of me wanting to be in an echo chamber. Like I can’t even post links to interesting stuff on the internet over there 🤷‍♂️
December 4, 2024 at 8:20 AM
I’m genuinely confused why we’re still arguing about the stochastic parrot claim. It never made much sense to me (or tbh most people who worked on these models) even back then. There’s also a lot of evidence against it.

I think it’s also pretty straightforward to see why it’s unlikely to be true.
November 23, 2024 at 11:50 AM
Reposted by Matthias Plappert
Because some damn fool asked, here's a thread about Baxter.

This was taken on the day I brought home Baxter and threw away all hope of an easy life.

Look at him. Absolutely irresistible.

And that's how they get you, the utter bastards
November 20, 2024 at 2:48 PM
This Lex Fridman interview of Dario, Amanda and Chris is quite good (though unreasonably long): youtu.be/ugvHCXCOmm4?...
Dario Amodei: Anthropic CEO on Claude, AGI & the Future of AI & Humanity | Lex Fridman Podcast #452
YouTube video by Lex Fridman
youtu.be
November 20, 2024 at 7:47 PM
Pretty wild how you only realize how terrible X has gotten once you get out. I can post links again! Nobody has yet tried to run a crypto scam on me! My notifications aren't randomly filled with Elon! What a bliss 🌈
November 19, 2024 at 2:25 PM
Reposted by Matthias Plappert
Have you ever done a dense grid search over neural network hyperparameters? Like a *really dense* grid search? It looks like this (!!). Blueish colors correspond to hyperparameters for which training converges, redish to those for which training diverges.

Even better, a video: vimeo.com/903855670
February 12, 2024 at 5:38 AM
Reposted by Matthias Plappert
Hope I'm the first to post this all time classic on this platform
November 19, 2024 at 4:51 AM
I wish Python had enums with associated values, similar to how enums work in Swift (literally the only feature in Swift I actually like)
November 19, 2024 at 9:35 AM
Early day OpenAI emails that were extracted from the Elon court case files: www.lesswrong.com/posts/5jjk4C...

I find these fascinating (and I was there during that time, but wasn’t fully aware of the drama that was going on back then)
OpenAI Email Archives (from Musk v. Altman) — LessWrong
As part of the court case between Elon Musk and Sam Altman, a substantial number of emails between Elon, Sam Altman, Ilya Sutskever, and Greg Brockma…
www.lesswrong.com
November 18, 2024 at 8:16 AM
Excited about this thing! Let’s go 🚀
November 17, 2024 at 7:27 PM