Tiberiu Tesileanu

Research engineer at #Meta Reality Labs.

[Tiberiu Tesileanu @TTesileanu]

@fchollet How exactly is that different from what most people do?

[David Sussillo @SussilloDavid]

If you are lucky enough to be at #Cosyne24 #Cosyne2024, we'll have a poster on the science of these results tonight.

Plus, we'll have a bunch of bands, so you can demo the tech yourself! 🤘🤓🤘

Poster info:
1-039 Scaling neural network training enables neuromotor interface decoders to generalize to new users

[David Sussillo @SussilloDavid]

1/7
For the past decade, our team at Meta Reality Labs (previously CTRL-labs) has been dedicated to developing a neuromotor interface.

Our goal is to address the Human Computer Interaction challenge of providing effortless, intuitive, and efficient input to computers.

[Mike McCabe @mikemccabe210]

Very excited to present Multiple Physics Pretraining!

MPP is a group of techniques for pretraining on a diverse set of time-dependent physics to learn models that can be fine-tuned for multiple problems.

Blog: https://t.co/fJw9xyCOhM
Paper: https://t.co/LF68K5D8rO

[Liam @liamhparker]

Excited to share some of our recent work at
@polymathicAI! With AstroCLIP we are bringing multimodal contrastive pretraining to astrophysics!

[PolymathicAI @PolymathicAI]

Announcing:

*The Polymathic AI Initiative*

We are researching a new class of foundation models for scientific *data*, developing models that share scientific concepts across disciplines.

Site: https://t.co/C2Qsi59HzI
News: https://t.co/JomPjdIJAG
Thread: 👇

[Siavash Golkar @SiavashGolkar]

1/ Did you know LLMs like ChatGPT struggle with number manipulation and data analysis? They make up answers and fail to generalize.

Today I am excited to share xVal, a continuous and efficient number encoding scheme with better generalization performance.
https://t.co/rhrxmrxnG6

[Dmitri "Mitya" @chklovskii]

A biologically plausible network for contrastive PCA - project led by @TTesileanu @SiavashGolkar @dlipshutz - tonight at #cosyne2023 poster I-091

[Dmitri "Mitya" @chklovskii]

Neural whitening via gain control in an overcomplete frame with a possible application to the olfactory bulb. A normative bioplausible network derived in a fantastic collaboration between the Simoncelli and Chklovskii labs led by @lyndoryndo and @dlipshutz https://t.co/JrZucQeqIN

[Randall Munroe @xkcd]

Cold Complaints https://t.co/gsT214CpQ8

[Tiberiu Tesileanu @TTesileanu]

@GaryMarcus So here's an AI that accurately describes mammals and the sailfish (Wiki: "considered by many scientists the fastest fish in the ocean"), but confuses the animal with a mammal. It's about as coherent as my 3-year-old, yes. Are you implying that that makes it a failure?!

[Dmitri "Mitya" @chklovskii]

Biologically plausible learning of commuting transformations from data (cf. Cohen & Welling, 2014). Poster #614 happening now @NeurIPS2022:

[Tiberiu Tesileanu @TTesileanu]

@neurograce To me a figure like that is similar to having a sentence with 17 subclauses without any clear hierarchy -- I doubt any human could parse that without breaking it up into more manageable pieces. Ideally the authors would have broken things down for the readers.

[APS Division of Biological Physics @ApsDbio]

Physics of Neural Systems focus session at the APS March Meeting! Please submit your contribution now https://t.co/e7MQFYT62b [https://t.co/5qHOb5wvn3], sorting categories 04.01.31, 02.01.53, 03.01.39 (image credit: DALL-E 2) (1/4)

[SueYeon Chung @s_y_chung]

First day as an assistant professor at NYU🔥Our lab is now up and running fully & jointly at NYU and Flatiron! Our lab's primary research interests lie at the intersection between theoretical neuroscience and machine learning. (1/)

[Dmitri "Mitya" @chklovskii]

A biologically plausible deep learning network using multicompartment neurons now at #NAISys poster # 122

[Tiberiu Tesileanu @TTesileanu]

Come see our Cosyne poster (1-041) on learning localized motion detectors using biologically plausible neural circuits! With Alexander Genkin and @chklovskii. @CosyneMeeting

[Dmitri "Mitya" @chklovskii]

A postdoc position is available in my group. Background in dynamical systems and data-driven control a plus. If interested, please send me your CV or see me at #cosyne2022 Please retweet

[Tiberiu Tesileanu @TTesileanu]

@KennethHayworth It can be a problem of access, not understanding. Take a microchip with a copy of the human genome but slow access 1bit/second. It has a lifetime of 10 years, and the memory is volatile so you can't just crack the case open and read it. How would you copy what's on that chip?