search

swkWien / sessions

Session presentations for Coding Dojo Vienna
http://swkwien.github.io/
16 stars 7 forks source link

Zero Overhead Coderetreats 2024 #134

Open caichinger opened 4 months ago

caichinger commented 4 months ago

About

Locations

Appointments

You want to take part and the event is not yet being taken care of? Go ahead and lead the way!

Dates:

Rules:

caichinger commented 4 months ago

@swkWien/organisers see above for this year's Zero Overhead Code Retreat schedule. Hoping to see many of you in course of the coming gatherings!

caichinger commented 3 months ago

ZOCR 2404

Thank you for joining! :clap:

You find today's state in above commit.

I would like that next time we think of our template to remind us of our rules of engagement, check-in etc. I will try to do that.

I am not entirely happy with our "let's try interactive programming session", maybe I offer one in the future where I take a more active role in guiding you through the process (!= navigator role), in case you want to join. I am interesting in better understanding the relationships between IP and TDD and how the can complement each other.

Here the links to the two publications I mentioned:

codecop commented 3 months ago

Regarding Brainf*k language that you rejected. It is not nice to code it for sure but it was not created to be like that. It seems to have been created to allow the smallest compiler possible: "Compilers für den Amiga in lediglich 240 Byte zu schreiben. Brian Raiter konnte dies unterbieten, indem er – unter Verwendung von nur 171 Bytes – einen Brainfuck-Compiler für x86 Linux schrieb. Für MS-DOS gibt es einen Brainfuck-Interpreter von Bertram Felgenhauer mit einer Größe von nur 98 Bytes" So I guess the exercise would be to create the compiler not code in the language ;-) Also it seems to be proven Turing complete.

gregorriegler commented 3 months ago

thank you for posting the links.

Citing The Curse of Recursion: Training on Generated Data Makes Models Forget (Model collapse)

very interresting. makes me think how this effect not only applies on llms but humans as well, like confirmation bias :)

caichinger commented 3 months ago

Regarding Brainf*k language that you rejected. It is not nice to code it for sure but it was not created to be like that. It seems to have been created to allow the smallest compiler possible: "Compilers für den Amiga in lediglich 240 Byte zu schreiben. Brian Raiter konnte dies unterbieten, indem er – unter Verwendung von nur 171 Bytes – einen Brainfuck-Compiler für x86 Linux schrieb. Für MS-DOS gibt es einen Brainfuck-Interpreter von Bertram Felgenhauer mit einer Größe von nur 98 Bytes" So I guess the exercise would be to create the compiler not code in the language ;-) Also it seems to be proven Turing complete.

Interesting, I did not know that. From the perspective of being a very limited language

Brainfuck requires one to break commands into microscopic steps.

could be interesting to try but I am still vary about the syntax. There is a VS Code extension for Brainfuck though.

caichinger commented 3 months ago

thank you for posting the links.

Citing The Curse of Recursion: Training on Generated Data Makes Models Forget (Model collapse)

very interresting. makes me think how this effect not only applies on llms but humans as well, like confirmation bias :)

You are welcome :) I too think that is pretty interesting.

FWIW

[...] this effect [...]

With regards to your comparison, I am not sure though. No being an expert in any of these matters... when we follow this definition of confirmation bias (that's a nice link as well btw):

Confirmation bias occurs when an individual looks for and uses the information to support their own ideas or beliefs. It also means that information not supporting their ideas or beliefs is disregarded.

I would say that an LLM has no beliefs (in particular they may hallucinate) and feeding it with LLM-created content does not imho introduce certain biases but instead feeds a degenerative process that is affecting overall model performance.

However, I too think that if we continuously succumb to our own biases then yes, our model of the world collapses. In that sense I agree with you. :)

gregorriegler commented 3 months ago

thank you for posting the links.

Citing The Curse of Recursion: Training on Generated Data Makes Models Forget (Model collapse)

very interresting. makes me think how this effect not only applies on llms but humans as well, like confirmation bias :)

You are welcome :) I too think that is pretty interesting.

FWIW

[...] this effect [...]

With regards to your comparison, I am not sure though. No being an expert in any of these matters... when we follow this definition of confirmation bias (that's a nice link as well btw):

Confirmation bias occurs when an individual looks for and uses the information to support their own ideas or beliefs. It also means that information not supporting their ideas or beliefs is disregarded.

I would say that an LLM has no beliefs (in particular they may hallucinate) and feeding it with LLM-created content does not imho introduce certain biases but instead feeds a degenerative process that is affecting overall model performance.

However, I too think that if we continuously succumb to our own biases then yes, our model of the world collapses. In that sense I agree with you. :)

how would you define belief?

maosmurf commented 2 months ago

Hi all, as the Dojo on Saturday 18th is coming up - how and where should I send the reminder?

caichinger commented 2 months ago

Hi all, as the Dojo on Saturday 18th is coming up - how and where should I send the reminder?

Thank you for hosting and reminding us! :)

For me it is up to you do the reminding part.

Options include:

caichinger commented 2 months ago

[...] how would you define belief?

@gregorriegler let's continue this conversation elsewhere (e.g. at our next meeting).

redseacomputing commented 2 months ago

Hey there, the events were created on Softwerkskammer and Meetup, but not on coderetreat.org and twitter for now. @caichinger @ReaSu @maosmurf @gregorriegler @goerge @codecop cu

adamzanyline commented 2 months ago

Thank you @redseacomputing for announcing via swk (and meetup)! My first tweet is out: https://x.com/swkWien/status/1788980108826398868

EDIT: adding links

maosmurf commented 2 months ago

Thank you @gregorriegler and @redseacomputing for todays session about approvaltests in Python using Inline Approvals Workflow. We started via cyber-dojo, using the Python 3.12, approvaltests-12.0.0 setup, then quickly moved to local PyCharm (and Code With me feature for faster handoff). One can find today's code at https://github.com/maosmurf/20240518_ant_approval_2

gregorriegler commented 2 months ago

Today I joined the Python Approvals Mob and told them about our CodeRetreat. I explained our problem where we lost the previous results on a failing refactor, and proposed an option that would preserve the previous result. We briefly implemented it. Amazing! https://github.com/approvals/ApprovalTests.Python/releases/tag/v12.2.0

caichinger commented 1 month ago

Who can organize a venue for the next session on Friday, 14.06.2024 09:00-14:00?

At WKO there is still one free room but I have few hours left.

goerge commented 1 month ago

WKO EPU 8 is booked on Code Mentor KG, Görge Albrecht on 2024-06-14 09:00-14:00

HTH

caichinger commented 1 month ago

WKO EPU 8 is booked on Code Mentor KG, Görge Albrecht on 2024-06-14 09:00-14:00

Thank you for taking care of that.

ReaSu commented 1 month ago

As you might have noticed I created the event on softwerkskammer.org and sent out an invite (thanks @goerge for the venue!), however, unfortunately I won't be able to attend.

Looking forward to July! I would really like to give interactive programming another try.

caichinger commented 1 month ago

As you might have noticed I created the event on softwerkskammer.org and sent out an invite (thanks @goerge for the venue!), however, unfortunately I won't be able to attend.

Thank you for taking care of that (even though you do not have time to join yourself).

Looking forward to July! I would really like to give interactive programming another try.

Happy to facilitate that in the future. :)

caichinger commented 1 month ago

I created https://pads.c3w.at/code/#/2/code/edit/IQyRDjgpy-0a2Vj6+BhDKoWR/ in case you want to document something.

See you tomorrow :)

goerge commented 1 month ago

@caichinger So it seems it's only the two of us. Do you still want to make it happen? If not, it's fine for me. In that case, I will cancel the WKO reservation and hopefully get my time budget back.

goerge commented 1 month ago

No worries. At least one "external" participant will join. I"ll be there.

caichinger commented 1 month ago

I"ll be there.

Thank you for the nice and productive session! (And not cancelling it:)

I documented and shared some of our observations. Feel free to make amendments.

ReaSu commented 3 weeks ago

Is anyone up for meeting in July, or are you all on holiday?

caichinger commented 3 weeks ago

Is anyone up for meeting in July, or are you all on holiday?

I would be in. 12, 19, 20 and 21 would work better for me though.

codecop commented 3 weeks ago

12th and 19th would work for me, Saturday is not an option. I did not want to start the discussion about moving dates, but now that Claus has started it ;-)

ReaSu commented 3 weeks ago

12, 19, 20 and 21 would work better for me though.

Out of these, 19th is least convenient for me.

Shall we move it to Friday, 12th?

Please vote:

:tada: : yes, 12th works for me :eyes: : no, 12th doesn't work, but 13th would be fine :slightly_frowning_face: : neither date works for me

codecop commented 3 weeks ago

@ReaSu / @goerge are we going for 12th? I could book WKO venue?

ReaSu commented 2 weeks ago

@codecop Thank you for booking the venue!

I have created this, in case someone wants to add a topic already: https://pads.c3w.at/code/#/2/code/edit/8vXivcYHWzWLZDqwcf3J1UrW/

Will send out an invite/reminder later today.

codecop commented 2 weeks ago

EPU Venue for 12th (this Friday) has been booked. It is "EPU 9", ask for meeting with Peter Kofler. It is available 9-14 as usual on Friday, for max 6 people. Unfortunately without windows.

ReaSu commented 2 weeks ago

Thank you!

without windows

looking at the weather forecast, I hope there will be air conditioning! :-)

gregorriegler commented 2 weeks ago

I hope I can join. Cannot promise though.

rolger commented 2 weeks ago

I hope there is enough space available. I'll join, too cu tomorrow

gregorriegler commented 2 weeks ago

Code from our first sessions: https://github.com/gregorriegler/coderetreat_2024_07_12_async

gregorriegler commented 2 weeks ago

Code from our final session: https://github.com/gregorriegler/coderetreat-2024-07-12-incremental

caichinger commented 2 weeks ago

Thank you all for contributing to this collaborative effort! :)

codecop commented 1 week ago

Flip with notes is in the folder https://github.com/swkWien/sessions/tree/main/2024/cr20240712