Eric DeWitt @[email protected]

Research scientist at Champalimaud Research studying the computational and neural underpinnings of individual and group decision making

[Kording Lab 🦖 @KordingLab]

Announcing a new week-long program for young computational neuroscience/ behavior professors to talk about rigorous science, mentoring, lab management, and networking in a stunning retreat setting. Do great science as a community and have fun doing so.

[Eric DeWitt @[email protected] @NeuroChooser]

@behrenstimb @eLife Couldn’t have happened to a better person! 🙂

[Eric DeWitt @[email protected] @NeuroChooser]

So happy to hear this good news!

[Tim Behrens @behrenstimb]

Super excited to be making this a permanent thing. It has been an exciting ride so far!!!

Thanks so much to Detlef for sharing it with me for the last year and great news that he is staying with @elife as senior editor going forward.

[Tim Behrens @behrenstimb]

This is a truly heroic effort from Kennerley lab to get model based reasoning neurons across pfc and striatum.

https://t.co/t7pn019gvA

[Eric DeWitt @[email protected] @NeuroChooser]

While foundation models have some utility, they lack ingredients critical for scientific progress. They don’t acquire new data or observations and don’t know what they don’t know. They don’t structure their (our) knowledge, to provide conceptual insights—and they hallucinate.

[Sam Rodriques @SGRodriques]

One of the remarkable things for me about NeurIPS this year was how quickly the entire AI for Biology community has gone all-in on biological foundation models. Virtual cell models will enable us to predict how cell states will change in response to chemical perturbations. Protein language models will enable us to identify better enzymes for degrading plastics, and so on. Everyone wants bigger data on more things to throw into bigger models.

These models are going to be awesome, but real biology discoveries look somewhat different. Contrast these dreams of foundation models with the latest table of contents from Science or Nature:

--“A long noncoding eRNA forms R-loops to shape emotional experience–induced behavioral adaptation” — The authors identified a lncRNA in mice that is expressed in response to neuronal activity that modulates the 3D structure of chromatin, thereby activating genes that are involved in neuronal plasticity. The authors further identified that this lncRNA is essential for certain forms of learning.

--“Cancer cells impair monocyte-mediated T cell stimulation to evade immunity” — The authors identified that mouse melanoma cells secrete a lipid metabolite that prevents monocytes from activating CD8+ T cells.

--“Postsynaptic competition between calcineurin and PKA regulates mammalian sleep–wake cycles” — By generating mouse knockout lines, the authors identified phosphatases and kinases that are critical for regulating the sleep-wake cycle, and showed that they act through regulation of proteins at excitatory postsynaptic sites.

I struggle to imagine how any of these discoveries could fall out of a multimodal biology foundation model. This is not intended to be a straw man argument. Surely, a foundation model could potentially identify the lncRNA from the first paper, but I am not sure how such a foundation model would associate it with chromatin remodeling. A multimodal foundation model with enough data could also potentially identify metabolic changes associated with melanoma cells subjected to certain kinds of treatments, but I don’t see how that foundation model could identify the effect of those metabolites in preventing CD8+ T cell activation. Indeed, I do not think that any of the foundation models that are being developed today would be capable of generating rich new biological insights of the kind described in these papers. And yet, these are the kinds of insights that new therapies are made from.

The issue, I think, is that machine learning models work extremely well on structured data, and so all the foundation models that are being built are highly structured. Take a protein sequence as input and produce a protein sequence as output. Take a cell state and a chemical perturbation as input and produce a new cell state as output. Biology, however, is poorly structured. The lncRNA insight is case in point: what structured representation can we use for the action of the lncRNA in modulating chromatin architecture? Protein models cannot represent it; DNA models cannot represent it; virtual cell models cannot represent it. Perhaps a model that incorporates RNA expression and 3D genome state could represent it, but then how would that model represent the lipid modulation of the monocytes? I worry that every discovery may need its own representation space. Indeed, the nature of biology is such that there likely is no representation, short of an atomic-resolution real-space model of the entire organism, that is sufficient to represent the diversity of biological phenomena that are relevant for disease.

Except, of course, for natural language, which is evolved to represent all concepts that humans are capable of contemplating. Indeed, I think natural language has an essential role to play in representing biology, and is ultimately unavoidable, insofar as it is the only medium we know of that is sufficiently structured for machine learning and sufficiently flexible to represent the full diversity of biological concepts. At FutureHouse, we work on language agents, which is one way of combining language and biology, but this is not the only way. Models that combine natural language with protein, DNA, transcriptomics, and so on will also be extremely productive, provided the addition of the structured datatypes does not restrict their ability to represent unstructured concepts. However we do it, I think this essential role of natural language in representing biology is currently largely underappreciated.

The history of biology is built on tools that we have found in nature to study biological phenomena. As all biologists know, trying to engineer things from scratch (almost) never works; what works is finding things in nature and repurposing them. It will be aesthetically pleasing if it turns out that our engineered representations are yet again insufficient for studying biology, and that natural language is simply another such tool that we have found in nature that must be applied instead.

[God bless @God_blessMerika]

To all neuroscientists still here in the darkness, join us where the sky is blue!

[Eric DeWitt @[email protected] @NeuroChooser]

Impact can not really be measured for 10s to 100s of years anyway… and definitely not by a (for profit) company looking to make profit! go eLife!

[Tim Behrens @behrenstimb]

Early days, but it looks like our authors don't think we need an impact factor either! Submissions are strong in numbers and high quality since the Clarivate announcement. We should not let a multibillion dollar company dictate how science is evaluated!

[Joana G Sá @mjoanasa]

Tinha prometido um post longo sobre o orçamento da @fct_pt @ciencia_pt mas saiu longo de mais e virou artigo
Resumido numa linha, não existe financiamento para a Ciência: o dinheiro é pouco, mal gasto e para "inovação"

A cada ano, uma oportunidade perdida
https://t.co/E9lDpS0C0T

[Tom George @TomNotGeorge]

What are the brain’s “real” tuning curves?

Our new preprint "SIMPL: Scalable and hassle-free optimisation of neural representations from behaviour” argues that existing techniques for latent variable discovery are lacking.

We suggest a much simpl-er way to do things.

1/21🧵

[Anna Shpektor @AnnaShpektor]

Excited to share my PhD results showing a hierarchical coordinate system for sequence memory in human EC.
https://t.co/8KicCW0j73
Below is a thread explaining this nonsense. 1/9

[Ella Batty @EllaBatty]

Relatedly, @KempnerInst is now hiring for my previous role! Let me know if you have any questions about it! https://t.co/XYqHFqk4RS

[Alex Wiltschko @awiltschko]

Well, we actually did it. We digitized scent. A fresh summer plum was the first fruit and scent to be fully digitized and reprinted with no human intervention. It smells great.

Holy moly, I’m still processing the magnitude of what we’ve done. And yet, it feels like as we cross this finish line we are instantly at a new starting line. I’ll have more to share about what’s in store that we’re building on top of this.

A huge HUGE congrats to the entire team across scientific, engineering, operational, and creative disciplines. It takes a village named Osmo to do this.

I don’t know if this is embarrassing, but I carry the plum scent with me a lot of places and smell it constantly. It makes me smile.

I’m curious, if y’all want to smell it? If we made a limited release fragrance of the first teleported scent and dedicated the proceeds to science, would you want it?

[Eric DeWitt @[email protected] @NeuroChooser]

@gershbrain Sadly, the most direct examples are probably mostly in the domain of the military—lots of direct cognitive studies on attention, performance, cognitive load for and by the military (sometimes with papers), but definitely directly studied and used…

[Eric DeWitt @[email protected] @NeuroChooser]

@gershbrain I would say probably behavioral economics, behavioral policy, education theory, all modern AI, human computer interface, … only the problem is there is always a big gap between a ‘paper’ and and ‘application’. Ideas propagate out and influence, but there are examples…

[Eric DeWitt @[email protected] @NeuroChooser]

@gershbrain Neural recordings are clearly indirect. Everything other than knowing the whole state of the system moment to moment necessarily involves assumptions, reductions, models, etc. I think the problem is aligning the question to the data—the data must be relevant for the purpose…

[Michael 英泉 Eisen @mbeisen]

It is so depressing that I still routinely hit paywalls when trying to read scientific papers. The scientific leadership of this country and the world have betrayed science and society with their perpetual cowardice on this, and so many other issues.

[Eric DeWitt @[email protected] @NeuroChooser]

And developing open, not-for-profit organizations to support and maintain open science resources beyond the scope of a grant or labs timeframe…

[Dan Levenstein @dlevenstein]

“If you plan on making a tool public… it’s good to think about when you’re going to hire the first software developer to work on this and how you’re going to pay for it.”

So when do we get to start putting this in grants that ask for open science?…

[Eric DeWitt @[email protected] @NeuroChooser]

It’s a deep problem. Almost all problems about an organism and its brain-behavioral relationship require cutting across many, many subdomains and require integration across methods and conceptual approaches…

[Shahab Bakhtiari @ShahabBakht]

Not limited to systems neuroscientists.

Most subdomains are defined by their methods, journals, and conferences so they end up mostly talking among themselves. Even when there’s an external push to engage with other subdomains, they tend to talk past each other.

[Yann LeCun @ylecun]

An article in the Wall Street Journal in which I express my opinion on the limitations of LLMs and on the potential power of new architectures capable of understanding the physical world, have persistent memory, can reason and can plan: four features of intelligent behavior that LLMs are incapable of, but that your cat possesses

https://t.co/xQ4wHKBx9Q

[Eric DeWitt @[email protected] @NeuroChooser]

This is a deeply important look inside the world as inferred by the training process of an LLM. Humans and animals learn rich world models—and humans use language to transmit and induce changes in them. LLMs only learn from language sequences, and build something different…

[𝐒𝐞𝐧𝐝𝐡𝐢𝐥 𝐌𝐮𝐥𝐥𝐚𝐢𝐧𝐚𝐭𝐡𝐚𝐧 @m_sendhil]

Excited to be part of this paper.

We study a question about language models that I find fascinating: have they come to “understand” the world behind the sequences they are trained on?