Back
47· Steady
Health & Science3h 46m ago
A new study published in the Journal of Cosmology and Astroparticle Physics (JCAP) found that while AI can accelerate the search for new laws of physics, its reliance on previous training can make it struggle to recognize genuinely new phenomena.
Archive Window: 30 Days Left
Princeton University
Who
Veena Krishnaraj, Adrian Bayer, Christian Kragh Jespersen, Peter Melchior
What
A new study published in the Journal of Cosmology and Astroparticle Physics (JCAP) found that while AI can accelerate the search for new laws of physics, its reliance on previous training can make it struggle to recognize genuinely new phenomena.
When
Sun, 14 Jun 2026 19:21:04 GMT · 3h 46m ago
Where
Princeton University ·
Why
AI can become so dependent on its prior training in established cosmological models that it overlooks or misinterprets new physics that present similar signals.
The Frontline Impact
How this affects you
This research suggests AI's pre-training, similar to how large language models function, could both expedite and impede the discovery of new physics in cosmology by making it efficient but also prone to 'negative transfer' when encountering truly novel data. This has implications for how researchers will develop and utilize AI in science.
Story chain
8 events in this thread- Health & Science3h 46m agoA new study published in the Journal of Cosmology and Astroparticle Physics (JCAP) found that while AI can accelerate the hunt for new physics, it sometimes struggles to recognize genuinely new phenomena due to its previous training.Open article
- Health & Science3h 46m agoA new study published in the Journal of Cosmology and Astroparticle Physics (JCAP) found that while AI can accelerate the search for new physics, its pre-training can sometimes make it difficult to recognize genuinely new phenomena.Open article
- Currently Reading3h 46m agoA new study published in the Journal of Cosmology and Astroparticle Physics (JCAP) found that while AI can accelerate the search for new laws of physics, its reliance on previous training can make it struggle to recognize genuinely new phenomena.
- Health & Science3h 46m agoAI could make it much cheaper and faster to search for new laws of physics, but in some situations, AI can become so dependent on its previous training that it struggles to recognize genuinely new phenomena.Open article
- Health & Science3h 46m agoA new study reveals that while AI can accelerate the search for new physics, its dependency on previous training can sometimes prevent it from recognizing genuinely new phenomena.Open article
- Health & Science3h 46m agoA new study found that while AI can accelerate the hunt for new physics, it sometimes struggles to recognize genuinely new phenomena due to over-reliance on previous training.Open article
- Health & Science3h 46m agoA new study in the Journal of Cosmology and Astroparticle Physics (JCAP) indicates that while AI can accelerate the search for new physics, it may struggle to recognize genuinely new phenomena when over-reliant on previous training.Open article
- Health & Science3h 46m agoA new study in JCAP suggests AI could accelerate the hunt for new physics, but may struggle to recognize genuinely new phenomena due to prior training.Open article