Enterprise AI Analysis
AI Needs Physics More Than Physics Needs AI
By Peter Coveney* and Roger Highfield
Artificial intelligence (AI) is commonly depicted as transformative. Yet, after more than a decade of hype, its measurable impact remains modest outside a few high-profile scientific and commercial successes. The 2024 Nobel Prizes in Chemistry and Physics recognized Al's potential, but broader assessments indicate the impact to date is often more promotional than technical. We argue that while current Al may influence physics, physics has significantly more to offer this generation of Al. Current architectures—large language models, reasoning models, and agentic Al—can depend on trillions of meaningless parameters, suffer from distributional bias, lack uncertainty quantification, provide no mechanistic insights, and fail to capture even elementary scientific laws. We review critiques of these limits, highlight opportunities in quantum Al and analogue computing, and lay down a roadmap for the adoption of ‘Big Al': a synthesis of theory-based rigour with the flexibility of machine learning.
Executive Impact: Key Metrics & Opportunities
Understanding the current landscape and future potential of AI's integration with fundamental science.
0
Measurable Impact of Current AI
0
BenevolentAI Valuation Collapse
0
Potential Reliability Boost with Physics-AI
Deep Analysis & Enterprise Applications
Select a topic to dive deeper, then explore the specific findings from the research, rebuilt as interactive, enterprise-focused modules.
The Scale of AI's Underspecification
Trillions Meaningless Parameters in LLMsCurrent AI architectures, including large language models, rely on an astronomical number of parameters, often in the trillions. These parameters are frequently meaningless in real-world terms, leading to models that suffer from distributional bias and a lack of mechanistic insight.
Pure AI vs. Physics-Informed AI: A Feature Comparison
| Feature | Pure AI | Physics-Informed AI |
|---|---|---|
| Interpretability | Low | High |
| Uncertainty Qualification | Weak | Strong |
| Scalability | High | Moderate |
| Mechanistic Insight | None | Embedded |
| Data Dependence | High | Lower |
Physics-Informed AI addresses key weaknesses of traditional pure AI, offering enhanced interpretability, robust uncertainty quantification, and embedded mechanistic insights, reducing over-reliance on massive datasets.
Path to Degenerative AI
Distributional Failures
→
Lack of Physical Understanding
→
Digital Pathologies
→
Spurious Correlations
→
Degenerative AI
This flowchart illustrates how inherent limitations in current AI architectures, from training data biases to fundamental mathematical challenges, can lead to a state of 'degenerative AI' where predictions degrade, especially with increased data.
Computational Performance Trade-offs
Inevitable Reduction in High-Fidelity Simulation PerformanceTo boost computational speed and cut memory usage, many AI implementations prioritize lower precisions (e.g., half-precision). This choice, however, leads to an inevitable reduction in the computational performance of high-fidelity simulations, impacting scientific accuracy.
Case Study: AlphaFold's Physical Discrepancies
Despite significant breakthroughs in protein structure prediction, models like AlphaFold3, Boltz-1, Chai-1 and RoseTTAFold All-Atom sometimes generate predictions that defy fundamental laws of physics and chemistry. This highlights the crucial need for grounding AI models in scientific principles to ensure their physical plausibility and trustworthiness, preventing fantastical or unworkable outputs. The article points out that critics found some of Google DeepMind's "digital crystals fantastical or unworkable," emphasizing that machine learning still needs the help of human chemists.
Key Takeaway: AI must move beyond mere pattern recognition and be grounded in scientific laws to achieve true reliability and prevent physically impossible results.
Roadmap for Big AI Adoption
Near-term (1-3 years): Improve scaling, VVUQ, benchmarks, understand spurious correlations.
→
Medium-term (3-7 years): Scale QIML, develop hybrid analogue, integrate causal inference.
→
Long-term (7-15 years): Achieve fault-tolerant quantum, high-precision analogue, AGI grounded in physical law.
This roadmap outlines the progressive integration of physics-informed methods, quantum inspiration, and analogue computing to build a more robust, reliable, and interpretable "Big AI" system over different timescales.
Calculate Your Potential AI-Driven ROI
Estimate the efficiency gains and cost savings your enterprise could achieve by adopting physics-informed AI methodologies.
Estimated Annual Savings
$0
Employee Hours Reclaimed Annually
0
Your Roadmap to "Big AI"
A strategic overview for adopting physics-informed, reliable, and interpretable AI in your enterprise.
Near-term (1-3 years)
Develop new AI and open physics-informed algorithms with improved scaling performance and more reliable error metrics; Standardize VVUQ protocols for ML in scientific computing; Establish community-defined benchmarks comparing physics-informed with pure ML predictions across multiple domains; Understand in more quantitative terms the impact of the deluge of spurious correlations to determine efficient means of dealing with it.
Medium-term (3-7 years)
Scale quantum-inspired machine-learning algorithms to industrially relevant problems; Develop hybrid analogue computing architectures or practical use; Integrate causal inference with physics constraints for improved generalization and reliable reasoning.
Long-term (7-15 years)
Achieve fault-tolerant quantum computing for scientific ML; Develop high accuracy, high precision analogue systems for general purpose computing; Establish an AGI theory grounded in physical law, not in the more speculative promise—championed by AI maximalists—that an algorithm might one day infer or even invent the universe's rules from pure data.
Ready to Transform Your AI Strategy?
Don't settle for "black box" AI. Partner with us to build intelligent systems grounded in scientific rigor, interpretability, and true reliability.
Ready to Get Started?
Book Your Free Consultation.
Let's Discuss Your AI Strategy!
Lets Discuss Your Needs
Select a Date
Sun
Mon
Tue
Wed
Thu
Fri
Sat