Ata Karagoz retweeted
Lots of folks talking about scaling LLM inference over this last year
Internally, I’ve been developing and using a library that makes this extremely easy, and I decided to open-source it
Meet the decoding library: github.com/benlipkin/decodin…
1/7
Ata Karagoz retweeted
Re-upping:
Postdoc position (two-to-three years) at @sfiscience on cogsci-inspired AI approaches to multimodal reasoning.
Start date flexible: target is Summer / Fall 2025.
PhD in AI, ML, CogSci, or related area (by start date)
Apply by Nov. 22.
santafe.edu/about/jobs/postd…
Ata Karagoz retweeted
I'm extremely excited to launch the relational cognition lab at UC Irvine: relcoglab.org/
We study learning and memory in mind, brains and machines. I am open to collaborations and hiring a lab technician in January 2025 (job ad soon). Reach out to chat!
Why did Factorio have to release an awesome expansion this year...
Factorio, Dwarf Fortress, or Dissertation. Tough choices.
Ata Karagoz retweeted
Now out in @NatureComms (lead: @Xu_Xinming)! How do conversations shape how (and what) people guess about the unknown past and future (e.g., in other people's lives)?
Paper/code/data: nature.com/articles/s41467-0…
Ata Karagoz retweeted
Attending #SfN24?
Postdoc @GrahekIvan will be presenting his talk “Control adjustment costs limit goal flexibility: Empirical evidence and a computational account”. Details below👇:
➡️ Room N228, Saturday October 5th at 1pm
➡️SFN '24 NANO07 - VBDM Across Model Systems Symposium
Ata Karagoz retweeted
I will be recruiting 💫two PhD students💫 for my new lab! (computational learning & memory applied to individual & interactive behavior, rouhanilab.com/).
Feel free to e-mail if interested and catch me at SfN! (poster 06 at 1 PM on Wed 10/09)
Go work with Rob, a fantastic scientist and human being!
Thrilled to announce that I will be joining CiNet (Center for Information and Neural Networks; cinet.jp/english/) in Osaka Japan to start my lab in April/May 2025! We have postdoc positions (1 open NOW) & PhD apps (Osaka U) possible. Details below & pls retweet! Short 🧵 1/
Ata Karagoz retweeted
Excited to share this project! Across 3 experiments, we find consistent evidence that encoding and prediction are coupled-- not competitive. See our preprint or come to my poster at SFN (PSTR305.12) to see some follow-up modeling work!
Does encoding the present compete with predicting the future?
In 3 behavioral studies, we find that encoding and prediction are coupled, not competitive!
Proud of @CraigPoskanzer & @hannahtstoll for leading this work w/ Raheema Javid & Edoardo Spolaore!
osf.io/preprints/psyarxiv/us…
Ata Karagoz retweeted
Does encoding the present compete with predicting the future?
In 3 behavioral studies, we find that encoding and prediction are coupled, not competitive!
Proud of @CraigPoskanzer & @hannahtstoll for leading this work w/ Raheema Javid & Edoardo Spolaore!
osf.io/preprints/psyarxiv/us…
Having just gotten to meet with Melanie at SFI, I think this is a fantastic opportunity! She is excellent, SFI is amazing, go do some brilliant work!
I'm looking for a postdoc (to start Spring or Summer 2025) to work with me & others at the Santa Fe Institute on AI models of visual and multimodal reasoning.
See santafe.edu/about/jobs/postd… for more info & application instructions.
Please RT!
Ata Karagoz retweeted
Social psych friends: do you know of--or have and are willing to share--any dataset(s) of open-ended conversations with survey outcome measures (e.g., enjoyment, connectedness)? Dyads preferred, but >2 speakers also fine. Like the 🔥 CANDOR dataset (even if smaller). Thanks!!
Ata Karagoz retweeted
Check out Lauren's awesome work in @PNASNews on how humans change their behavior when they are training AI. Our university wrote a nice article covering this work, with some awkward quotes from yours truly. artsci.wustl.edu/ampersand/h…
Beyond excited that our paper “The consequences of AI training on human decision making” has been published in @PNASNews (pnas.org/doi/10.1073/pnas.24…) with #CJHo and @wouterkool. We report 5 experiments that show people change their behavior when training AI.
Ata Karagoz retweeted
Proud to have the entire @FINNLab_neuro at #CCN2024! Here's where you can catch us:
First up, Katie O'Nell @katieoneuro has a talk in tomorrow morning's session (Thurs 9:30am) about a new way to quantify what makes conversation good (hint: think pink 🩷🎀💅)
Ata Karagoz retweeted
For folks who aren't able to attend in-person, we are excited to be able to stream all talks this year on our youtube channel:
youtube.com/@cogcompneuro
Papers can be found here:
2024.ccneuro.org/poster-sess…
Please share!
Go work with Dr. Spiers, who has consistently been doing fantastic and inspiring work!
This Post is from a suspended account. Learn more
Ata Karagoz retweeted
Glad that this paper is out @SciReports
where we combined Bayesian nonparametrics & vector representations of context to learn shared & context-specific information; and we used the model to account for human data on schema learning and event cognition!
nature.com/articles/s41598-0…
Ata Karagoz retweeted
Our attention waxes and wanes over time. How does this influence the way our memories are organized?
Pupil size at encoding, an index of attention, predicts the temporal structure of subsequent memory
Super proud of @manasijkumar for this work! 👀
osf.io/preprints/psyarxiv/bz…
ALT Schematic of the experimental design and task measures. Participants encoded object images and make a judgment on each, while pupil diameter was monitored. After a distractor task, they recalled the objects in any order they choice.
ALT Plots showing recall (y-axis) as a function of the distance between objects at encoding (x-axis). Recall was more temporally organized when pupil sizes were larger (vs. smaller) at encoding. This is reflected in greater temporal contiguity and forward asymmetry effects (see paper for detail).
Ata Karagoz retweeted
survey on family of predictive representations defined by the successor representation accepted to neural computation!
some updated signposting to ease reading. enjoy :)
arxiv.org/pdf/2402.06590
