Scott Dodelson

"A few days ago, I typed into codex: “i wrote a paper 25 years ago with a model for evolving dark energy. Can you write code that chooses parameters in that model so that it gets distances consistent with the latest DESI results?” In about 5 minutes, it produced ..."

... the introduction of a cosmological constant was not just the proposal that a specific model is correct. It was the fundamental belief that we are on to something, that our audacity to answer the most profound questions is warranted." scottdodelson.substack.com

"30 years ago when a cosmological constant was invoked to resolve a different set of tensions: I thought it was unserious. But I was wrong...

Please consider subscribing [for free!] to received emails linking to posts. scottdodelson.substack.com

"DESI is now state-of-the-art, giving us insight into how fast the universe has been expanding over the past 5-10 billion years. The fiducial cosmological model predicts the distance-redshift relation. So, these measurements allow us to test ΛCDM. And the model seems to fail ..."

...one or more of its precise predictions is wrong. We then stand on its shoulders and move beyond it, to a higher level of truth." open.substack.com/pub/scottdodel…

"A great model of reality makes predictions for phenomena that are otherwise unexplained. And that means that measurements can be made to pin down any freedom in the model... and then, ironically, this precision will sometimes lead to the realization that the model is incomplete

For those of you looking for yet another reason to go to Aspen, see the pointer in my substack.

@McdarghZach nice to hear from you Zach; hope you are doing well

@ScottDodelson scottdodelson.substack.com/p/evolving-dar… Posting the link for other folks. Btw, hello from a former CMU grad student, didn’t realize you were on here or S*bstack 👋

"A few days ago, I typed into codex: “i wrote a paper 25 years ago with a model for evolving dark energy. Can you write code that chooses parameters in that model so that it gets distances consistent with the latest DESI results?” In about 5 minutes, it produced ..."

@Yi_Zen_Chu my thinking on this has been evolving rapidly over the past few days. Your question is one of the key questions. We can do this though by asking them for verification: show me that your result agrees with X in the literature. My first go at this was not encouraging.

@ScottDodelson In the long run, how do we ensure that research results are robust? That is--in your view--how do we ensure that AI-assisted results are independently verifiable?

@NikoSarcevic @skdh ha ha ... i think[?]

@ScottDodelson @skdh Last year Troxel told me he tried something similar — DES Y3 results if i remember correctly

@ourtown2 These are facts about the CMB. Would they have led AI to introduce the story that there was a time before galaxies/stars/people/computers? That there was only e,p,gamma then? That recombination took place at z=1080?

Angular coherence: large-scale isotropy with specific multipole structure inconsistent with local sources. Spectral form: near-perfect blackbody spectrum at 2.725 K (CMB), matching equilibrium radiation Dipole anisotropy: consistent with Earth’s motion through a cosmological rest frame. Frequency–intensity scaling: follows expansion-redshift relations across independent measurements.

"How we do know that the photons detected today in the microwave range of the spectrum are cosmic, that they come from the earliest moments of time, that they are indeed the oldest things ever seen? This requires a bit of story telling and modeling ..."

@WilmsRolf Not sure I agree: a few people predicted it, but not many. It was not an obvious prediction of the expanding universe. In any event, I'm not sure the fact that some people predicted it implies that AI could have.

@ScottDodelson The CMB was predicted, only years later it was detected. That’s a pretty convincing story.

read more on my subst@ck

"... last week @mattshumer_ posted about the accelerating power of AI. I wonder how much of science will remain a human endeavor: could AI have figured out that the microwave background is cosmic? Could AI have come up with the very simple story/model I am about to share?"

After the destruction of the Temple, the rabbis of the Talmud confronted a parallel problem: how to preserve a tradition while fundamentally transforming it." Read more here: open.substack.com/pub/scottdodel…

"In cosmology today ... observational tensions suggest that the standard model may not be the final story. A new paradigm may be needed. What does it mean to reinvent a scientific framework without losing what made it successful in the first place?