Wojciech Jaśkowski

Maxine Learning / Senior Research Scientist @ Snowflake. Opinions are mine own.

[Wojciech Jaśkowski @WJaskowski]

Dear @Google, "Full house"? What is that supposed to mean? Should I kick off one of my children on the street? Not only did I find this offensive but also exclusive! How can I create a supervised gmail account for my fifth child?

[Wojciech Jaśkowski @WJaskowski]

Training this model was a lot of fun. As usual, lots of data + lots of compute = great results. This one is special however due to its small size and 400k context. The power of encode-decoder models has been underappreciated these days.

[Łukasz Borchmann @LukaszBorchmann]

We introduce Arctic-TILT, which achieves accuracy on par with models 1000x its size for #LLM workloads involving answering questions grounded on PDF or scan content. 🧵
https://t.co/pDabKN6OH6

[Wojciech Jaśkowski @WJaskowski]

@BdsLoick @jxmnop Looking forward! Any news on this?

[Santiago @svpino]

Scrum is one of the worst experiences I had to endure in my career.

[Wojciech Jaśkowski @WJaskowski]

@kawecki_maciej Ale po co?! Nie ma dowodów na to, że nauka gry w szachy w jakikolwiek pozytywny sposób oddziaływuje na jakiekolwiek inne umiejętności poza grą w szachy (https://t.co/YOY0G97nxn)

[Wojciech Jaśkowski @WJaskowski]

@kawecki_maciej Można gdzieś zobaczyć nagranie całości?

[Edward Reid @ReidEdwardII]

ONLY IN POLAND!

Most people have no idea about the differences between the countries occupied by the Germans.

The Poles faced a brutal and horrific occupation that differed in many ways from other countries.

This video highlights the differences and the reactions of the Poles.

[Maciej Kawecki - This Is IT @kawecki_maciej]

@elonmusk space travels era with 90’s electronics, really? Check out innovation from @mkratajczak. Saves space in starships. It could be printed in on Mars. Pack the inks instead of gadgets. 2 Poles invented cheapest printed display also to replace printed surfaces with display!

[Wojciech Jaśkowski @WJaskowski]

@vwxyzjn @cleanrl_lib @PyTorch Why Jax has lower variance on HC?

[Wojciech Jaśkowski @WJaskowski]

@AnnaSiarkowska Zdjęcie (podpisane jako ilustracyjne) w takiej formie pochodzi bezpośrednio ze stocka, co można łatwo sprawdzić, tak więc żadnej manipulacji po stronie interii nie ma https://t.co/hiuZnq5fB3

[Wojciech Jaśkowski @WJaskowski]

@kosik_md Mówienie tak szybko (lekarz w szpitalu), żeby pacjent nie zdążył przetworzyć i nie zadał czasem pytań merytorycznych. Taką strategię zaobserwowałem.

[Wojciech Jaśkowski @WJaskowski]

A fascinating debunking of Dunning-Kruger Effecr

[yobibyte @y0b1byte]

Nothing is sacred to these people

https://t.co/c9dSrrYGfc

[Wojciech Jaśkowski @WJaskowski]

@AndreasPSteiner Good but far from perfect...

[Marek Magierowski @mmagierowski]

Words matter, dear @CNN. There was no such thing as „Poland’s concentration camps”. Those were German Nazi camps in German-occupied Poland. You would not refer to Bucha as “Ukraine’s carnage” or to Srebrenica as “Bosnia’s massacre”, would you? @AuschwitzMuseum @HolocaustMuseum

[Wojciech Jaśkowski @WJaskowski]

@fhuszar I don't know about Hungary, but most of the media in Poland is private and independent, e.g., 2 out of 3 major TV stations.

[Wojciech Jaśkowski @WJaskowski]

@hardmaru I recommend Codemonkey.

[Wojciech Jaśkowski @WJaskowski]

simple & solid improvement

[Samarth Sinha @_sam_sinha_]

Deep learning has seen huge gains when you increase the number of layers, but what about Deep RL?

Introducing D2RL! Changing how you parameterize your policy + Q function boosts performance

Co-led with @mangahomanga. @AravSrinivas @animesh_garg

Link: https://t.co/ucud3XHTDE

[zion | w.ai @benzion_b]

2015: the AI revolution is coming
2020:

@geoffreyhinton @SchmidhuberAI

[Wojciech Jaśkowski @WJaskowski]

@rupspace Sociology uses different methods and each of them has its limitations. Asking people is one of them. Do you suggest that people are equally racist as they were 50 years ago but just lie more?

[NNAISENSE @nnaisense]

In a recent technical report, we have developed theoretical connections between energy functions of noisy data learned using neural networks and "randomized smoothing" for training *certifiably* robust classifiers.

Read the report: https://t.co/VqgY6EHjPO (1/2)