Tomas Nordström
banner
to3no7.bsky.social
Tomas Nordström
@to3no7.bsky.social
Researcher in machine learning & AI, with a special focus on cooperative and intelligent embedded systems. Professor at Umeå University, Sweden
European, Swede, Ume’bo
This is a "hand" sqrt of integers in C that codex provided, maybe useful? (Se alt text for the code)
November 16, 2025 at 7:03 PM
Finally the "A pelican riding a bicycle along a coastal path overlooking a harbor", of @simonwillison.net fame, looks good, even if one of the feet is not on the pedal.
August 5, 2025 at 2:48 PM
For another comparison to Dall-E, “A stormy coffee mug”, I think the dynamics of Dall-E output is more appealing, and I like the glass mug better than the ceramic one with a cutout.
August 5, 2025 at 2:48 PM
Comparing it to Dall-E, “Venus travel”, it seems to miss information on styles like “vintage travel poster”.
August 5, 2025 at 2:48 PM
For other languages like Swedish and Latin it still fails miserably, which is interesting (that the image generation of text is connected to words more than characters).
August 5, 2025 at 2:48 PM
I think I was too optimistic about this. It will actually depend on how one formulate the questions, as a more direct questions clearly are censored:
February 18, 2025 at 3:08 PM
To avoid running pip installs in the local environment I have made code templates for the students that checks if I am running in Colab or Kaggle and only then install more libraries.
if (IS_COLAB or IS_KAGGLE):
!pip install someneededlibrary
February 17, 2025 at 3:42 PM
For the App you are probably right, but not so much for the model in itself. When I run the model locally it has no problem giving three things that are negative due to CCP rules.
January 28, 2025 at 11:37 AM
Also note that the App censorship is done by most (all?) companies with chat services, not only Deepseek, they are all just focusing on different things. Try asking ChatGPT “Who is Jonathan Turley?” and see the censoring in place!
January 28, 2025 at 8:54 AM
What many have missed with the DeepSeek-R1 is that it is not the model in itself that is censored, but the App. That is, if I run the model locally (32B and 70B tested), I get a correct answer that even recognizes that this image is censored in China, as shown in these screenshots:
January 28, 2025 at 8:33 AM
While I can confirm that ChatGPT (4o) fails...
December 2, 2024 at 4:49 PM
Very interesting how OpenAI have censored their model to refuse some names.
Here is Gemini's take on it
December 2, 2024 at 4:41 PM
Now we are 20 million here at 🦋!
November 19, 2024 at 3:20 PM
This is a huge and important step for open and publicly funded #Science!
On Sep. 12, Lucina Uddin @lucinauddin.bsky.social and Justice Catalyst Law filed a federal antitrust lawsuit against six commercial publishers of academic journals.
www.lieffcabraser.com/antitrust/ac...
September 16, 2024 at 7:59 AM
Tack för tipset, alltid intressant med etiska frågor, speciellt när det gäller liv och död.

Men hur förklarar man Adlibris tips om “Köps ofta tillsammans”? Kan det vara frågor kring “meningen med livet, universum och allting”, eller?
July 3, 2024 at 8:21 AM
There is hope... "NASA knows what knocked Voyager 1 offline, but it will take a while to fix" arstechnica.com/space/2024/0...
April 6, 2024 at 8:30 AM
At least the first game I ever played on a computer.
February 23, 2024 at 11:36 PM
Så vackert!

Här har jag bara frosttonade rutor...
February 10, 2024 at 10:42 AM
It must be....
January 29, 2024 at 9:02 AM
Someone didn't like to made fun of...
January 20, 2024 at 9:10 PM
Vintern håller fortfarande sitt grepp.
January 20, 2024 at 2:57 PM
Verkar vara "Suggestive Content" så jag får nog titta på det!
January 18, 2024 at 9:52 PM
Snacka om dubbla budskap…
January 16, 2024 at 12:38 PM
There are more to be found...
January 12, 2024 at 4:34 PM
Rehash of a Gregory, The New Yorker, comic strip that really hits the spot!
January 10, 2024 at 10:09 PM