#computational science

[frank-olivier]

Deep Learning, Deconstructed: A Physics-Informed Perspective on AI’s Inner Workings

Dr. Yasaman Bahri’s seminar offers a profound glimpse into the complexities of deep learning, merging empirical successes with theoretical foundations. Dr. Bahri’s distinct background, weaving together statistical physics, machine learning, and condensed matter physics, uniquely positions her to dissect the intricacies of deep neural networks. Her journey from a physics-centric PhD at UC Berkeley, influenced by computer science seminars, exemplifies the burgeoning synergy between physics and machine learning, underscoring the value of interdisciplinary approaches in elucidating deep learning’s mysteries.

At the heart of Dr. Bahri’s research lies the intriguing equivalence between neural networks and Gaussian processes in the infinite width limit, facilitated by the Central Limit Theorem. This theorem, by implying that the distribution of outputs from a neural network will approach a Gaussian distribution as the width of the network increases, provides a probabilistic framework for understanding neural network behavior. The derivation of Gaussian processes from various neural network architectures not only yields state-of-the-art kernels but also sheds light on the dynamics of optimization, enabling more precise predictions of model performance.

The discussion on scaling laws is multifaceted, encompassing empirical observations, theoretical underpinnings, and the intricate dance between model size, computational resources, and the volume of training data. While model quality often improves monotonically with these factors, reaching a point of diminishing returns, understanding these dynamics is crucial for efficient model design. Interestingly, the strategic selection of data emerges as a critical factor in surpassing the limitations imposed by power-law scaling, though this approach also presents challenges, including the risk of introducing biases and the need for domain-specific strategies.

As the field of deep learning continues to evolve, Dr. Bahri’s work serves as a beacon, illuminating the path forward. The imperative for interdisciplinary collaboration, combining the rigor of physics with the adaptability of machine learning, cannot be overstated. Moreover, the pursuit of personalized scaling laws, tailored to the unique characteristics of each problem domain, promises to revolutionize model efficiency. As researchers and practitioners navigate this complex landscape, they are left to ponder: What unforeseen synergies await discovery at the intersection of physics and deep learning, and how might these transform the future of artificial intelligence?

Yasaman Bahri: A First-Principle Approach to Understanding Deep Learning (DDPS Webinar, Lawrence Livermore National Laboratory, November 2024)

Sunday, November 24, 2024

[prokopetz]

Wrong: Ada Lovelace invented computer science and immediately tried to use it to cheat at gambling because she was Lord Byron's daughter.

Right: Ada Lovelace invented computer science and immediately tried to use it to cheat at gambling because that was the closest you could get in 1850 to being a Super Mario 64 speedrunner.

[feyosha]

Computers are very simple you see we take the hearts of dead stars and we flatten them into crystal chips and then we etch tiny pathways using concentrated light into the dead star crystal chips and if we etch the pathways just so we can trick the crystals into doing our thinking for us hope this clears things up.

[noosphe-re]

"There was an exchange on Twitter a while back where someone said, ‘What is artificial intelligence?' And someone else said, 'A poor choice of words in 1954'," he says. "And, you know, they’re right. I think that if we had chosen a different phrase for it, back in the '50s, we might have avoided a lot of the confusion that we're having now." So if he had to invent a term, what would it be? His answer is instant: applied statistics. "It's genuinely amazing that...these sorts of things can be extracted from a statistical analysis of a large body of text," he says. But, in his view, that doesn't make the tools intelligent. Applied statistics is a far more precise descriptor, "but no one wants to use that term, because it's not as sexy".

'The machines we have now are not conscious', Lunch with the FT, Ted Chiang, by Madhumita Murgia, 3 June/4 June 2023

[warakami-vaporwave]

The 84 Series

[quasi-normalcy]

Hey, so--we cooled your boyfriend down to a hundredth of a kelvin above absolute zero. Yeah, it was so cold that all of the chemical reactions in his body ceased. Sorry. We, uh, yeah, we used him as a dielectric material in a tiny qubit. And then we quantum-entangled him with another qubit, just to see if we could. Sorry. Yeah, anyway, we thawed him out after two weeks and apparently he's doing fine now. Didn't really teach us anything about how quantum processes work in biological systems, but it sure was, uh, cool. If you'll pardon the pun.

[captaindrewboy]

I hate when I tell people I'm into Comp Sci and programming and they go "ooh there's a lot of money there". I'm not DOING it for the MONEY I'm DOING it because I want to FUCK the COMPUTER!!!!

[microwavedmetal]

This is as unhinged as this laptop soon will be

[c-u-c-koo-4-40k]

What should you bring to school?

For the people of Gaza, the memories of days spent in school are distant. But education is not Only a Right but a Duty.

Meet Anas and Ahmad, they are related to Osama Basil, who has his own gofundme.

While managing his own gfm, Osama is also trying to help Anas and Ahmed fund their education somewhere safe, away from the Turmoil in Gaza.

They are Bright Wonderful students who want to pursue careers in computer business and medicine!

Please give some time to this Verified campaign!

Current Funding

€4,870 / €29,000

~16% Funded

Tag List for Reach DM me for Removal

And to others putting me in mentions I am So Sorry I'm not getting to many of them, I'm trying to balance between DM campaigns, ask campaigns and the mentions I get each day.

@a-shade-of-blue @sunnylittledragon

@selflovejolteon @virovac

@frustrated-froglet @qattdraws @heydreamchild

@amvs @boosting-donations

@sweet-honey-bunnies

@dyspunktional-leviathan

@nobelgasxenon

@zone0neko

@comrademango

[jcmarchi]

David Darmofal SM ’91, PhD ’93 named vice chancellor for undergraduate and graduate education

New Post has been published on https://thedigitalinsider.com/david-darmofal-sm-91-phd-93-named-vice-chancellor-for-undergraduate-and-graduate-education/

David Darmofal SM ’91, PhD ’93 named vice chancellor for undergraduate and graduate education

David L. Darmofal SM ’91, PhD ’93 will serve as MIT’s next vice chancellor for undergraduate and graduate education, effective Feb. 17. Chancellor Melissa Nobles announced Darmofal’s appointment today in a letter to the MIT community.

Darmofal succeeds Ian A. Waitz, who stepped down in May to become MIT’s vice president for research, and Daniel E. Hastings, who has been serving in an interim capacity.

A creative innovator in research-based teaching and learning, Darmofal is the Jerome C. Hunsaker Professor of Aeronautics and Astronautics. Since 2017, he and his wife Claudia have served as heads of house at The Warehouse, an MIT graduate residence.

“Dave knows the ins and outs of education and student life at MIT in a way that few do,” Nobles says. “He’s a head of house, an alum, and the parent of a graduate. Dave will bring decades of first-hand experience to the role.”

“An MIT education is incredibly special, combining passionate students, staff, and faculty striving to use knowledge and discovery to drive positive change for the world,” says Darmofal. “I am grateful for this opportunity to play a part in supporting MIT’s academic mission.”

Darmofal’s leadership experience includes service from 2008 to 2011 as associate and interim department head in the Department of Aeronautics and Astronautics, overseeing undergraduate and graduate programs. He was the AeroAstro director of digital education from 2020 to 2022, including leading the department’s response to remote learning during the Covid-19 pandemic. He currently serves as director of the MIT Aerospace Computational Science and Engineering Laboratory and is a member of the Center for Computational Science and Engineering (CCSE) in the MIT Stephen A. Schwarzman College of Computing.

As an MIT faculty member and administrator, Darmofal has been involved in designing more flexible degree programs, developing open digital-learning opportunities, creating first-year advising seminars, and enhancing professional and personal development opportunities for students. He also contributed his expertise in engineering pedagogy to the development of the Schwarzman College of Computing’s Common Ground efforts, to address the need for computing education across many disciplines.

“MIT students, staff, and faculty share a common bond as problem solvers. Talk to any of us about an MIT education, and you will get an earful on not only what we need to do better, but also how we can actually do it. The Office of the Vice Chancellor can help bring our community of problem solvers together to enable improvements in our academics,” says Darmofal.

Overseeing the academic arm of the Chancellor’s Office, the vice chancellor’s portfolio is extensive. Darmofal will lead professionals across more than a dozen units, covering areas such as recruitment and admissions, financial aid, student systems, advising, professional and career development, pedagogy, experiential learning, and support for MIT’s more than 100 graduate programs. He will also work collaboratively with many of MIT’s student organizations and groups, including with the leaders of the Undergraduate Association and the Graduate Student Council, and administer the relationship with the graduate student union.

“Dave will be a critical part of my office’s efforts to strengthen and expand critical connections across all areas of student life and learning,” Nobles says. She credits the search advisory group, co-chaired by professors Laurie Boyer and Will Tisdale, in setting the right tenor for such an important role and leading a thorough, inclusive process.

Darmofal’s research is focused on computational methods for partial differential equations, especially fluid dynamics. He earned his SM and PhD degrees in aeronautics and astronautics in 1991 and 1993, respectively, from MIT, and his BS in aerospace engineering in 1989 from the University of Michigan. Prior to joining MIT in 1998, he was an assistant professor in the Department of Aerospace Engineering at Texas A&M University from 1995 to 1998. Currently, he is the chair of AeroAstro’s Undergraduate Committee and the graduate officer for the CCSE PhD program.

“I want to echo something that Dan Hastings said recently,” Darmofal says. “We have a lot to be proud of when it comes to an MIT education. It’s more accessible than it has ever been. It’s innovative, with unmatched learning opportunities here and around the world. It’s home to academic research labs that attract the most talented scholars, creators, experimenters, and engineers. And ultimately, it prepares graduates who do good.”

[mushrooms-and-millipedes]

YO NEW HELL ANT JUST DROPPED

IN THE SOUTHERN HEMISPHERE NO LESS

This is from a study just published a few days ago by Lepeco et al. (2025).

Hell ants (subfamily: Haidomyrmecinae) are super cool ants that possessed vertically closing jaws.

Below is a model of the ant's body/anatomy generated via micro-computed tomography

Vulcanidris cratensis is a new species discovered in Brazil. It dates back to the lower Cretaceous, and at the time of writing, is now the oldest known ant to science!

[scipunk]

Alien (1979)

[pyroclastic727]

meta put out a video saying that they want to phase out mid-level software developers. and okay, besides the obligatory "haha programmers are coding their jobs out of existence"... i think it's really worrying that we're phasing out mid-level? like, forget about entry level. only high-level programmers, the ones smart enough to control the ai and do what it can't, can get a job. but, if you get rid of the path to learning how to be a high-level programmer (that path being years of experience), how will anyone reach that level of expertise? do you get your first programming job at 40, after studying intensely in your parents' basement for years? or does technology become a thing that only the ai can control?

[389]

[12neonlit-stage]

"Scientists today have been able to experimentally prove that quantum entanglement is in fact linked to the creation of wormholes."

tiny :3