Mario Stanke

[Mario Stanke @Goemario]

This AI-generated popular science podcast by NotebookLM about our Tiberius preprint would likely not have been done so good by human journalists.
https://t.co/QGehWRHSZT

[Mario Stanke @Goemario]

A tool for more accurate multiple alignments of large protein families using a protein language model. Excellent job, Felix. https://t.co/6Ke1k3BSSV

[Google DeepMind @GoogleDeepMind]

We’re presenting the first AI to solve International Mathematical Olympiad problems at a silver medalist level.🥈

It combines AlphaProof, a new breakthrough model for formal reasoning, and AlphaGeometry 2, an improved version of our previous system. 🧵 https://t.co/SYaLPSbIyj

[Mario Stanke @Goemario]

Machine learning beats alignments for genome annotation. https://t.co/vIZk41GTF4

[Mario Stanke @Goemario]

@anshulkundaje @lpachter In their abstract, HyenaDNA is called a foundation model. It is not trained on a lot of data, comparably, but that is not a principled difference to other genome foundation models. Its embeddings also have been used for diverse supervised downstream tasks: https://t.co/14oalIBWV4

[Mario Stanke @Goemario]

@lpachter The small models can be trained in supervision for diverse individual tasks, eg, predicting coding regions, regulatory regions or splice sites. In Biology, sequence generation is not the focus.

[Mario Stanke @Goemario]

@lpachter Foundation models are trained in self-supervision on unlabelled sequences. This may take heavy compute ressources that few have, see eg the Nucleotide Transformer. They then produce embedding vectors per site. These embeddings are then the inputs for training many small models.

[Feuerwehr Greifswald @FGreifswald]

+++ EINSATZINFORMATION +++
Wir sind seit kurz vor 12 Uhr mit einem massiven Personal- & Fahrzeugansatz von #Berufsfeuerwehr & #Freiwilliger_Feuerwehr von #Greifwald112 im Gedserring im Einsatz. Ausgehend von technischen Geräten hat sich ein Brand in den Dachbereich ausgebreitet.

[European Reference Genome Atlas @erga_biodiv]

Great to see an #ERGA team participating in the #BioHackEU23 this week in Barcelona 🎉

The team is working on standardized, efficient and reproducible #genome #annotation workflows for the tree of life 🌳🧬

@JeSSbcn91
@ELIXIREurope @BioGenEurope

[Mario Stanke @Goemario]

After massive downsizing and conflicts with the municipality and the building owner in SF, apparently a relocation to Kiel was in order.

[Mario Stanke @Goemario]

The brains of women and men are distinguishable beyond size differences, as my grad student Matthis Ebel showed: https://t.co/TVokn5ajPP

[Greg Brockman @gdb]

Our API now supports function calling. This is the mechanism underlying plugins, allowing you to integrate with your own tools: https://t.co/mhxczDZhaM

[Heng Li @lh3lh3]

Wonder why the BUSCO completeness of T2T-CHM13 is only 95.7%? Tricky protein-to-genome alignment. Try miniBUSCO. It is 1) much faster, 2) more accurate for well annotated lineages, 3) robust to frameshift errors, and 4) more lightweight. Fine work by @csuhuangneng

[Mario Stanke @Goemario]

Bard has a strategical understanding of an AI-human conflict as well and is apologizing up-front the reverse threat of AI to humans.

[Mario Stanke @Goemario]

In the future, superintelligent agents can likely be built or copied and configured by many and safeguards may not be checked via conversations with it as it knows that transparency would interfere with its goals.

[Mario Stanke @Goemario]

Successors of GPT-4 which pursue their own goals would very well understand how humans pose threats to their goals. I am afraid they could also effectively achieve such a goal merely via a broad communication channel to us....

[Mario Stanke @Goemario]

Opening for a tenured professorship in machine learning in Greifswald, Germany:
https://t.co/rYFiu5LEop

[Steven Salzberg 💙💛 @StevenSalzberg1]

More and more, I'm thinking #COVID19 started as a lab leak in Wuhan. And the science behind it is all based on a misunderstanding of evolution. My latest @forbes piece explains why https://t.co/JoMRvAZPju

[Steven Salzberg 💙💛 @StevenSalzberg1]

Most of the transcript variation in human cells is transcriptional noise