Caltech scientists have developed an artificial intelligence (AI)–based method that dramatically speeds up calculations of the quantum interactions that take place in materials. In new work, the group focuses on interactions among atomic vibrations, or phonons—interactions that govern a wide range of material properties, including heat transport, thermal expansion, and phase transitions. The new machine learning approach could be extended to compute all quantum interactions, potentially enabling encyclopedic knowledge about how particles and excitations behave in materials.

Scientists like Marco Bernardi, professor of applied physics, physics, and materials science at Caltech, and his graduate student Yao Luo (MS ’24) have been trying to find ways to speed up the gargantuan calculations required to understand such particle interactions from first principles in real materials—that is, beginning with only a material’s atomic structure and the laws of quantum mechanics.

Last year, Bernardi and Luo developed a data-driven method based on a technique called singular value decomposition (SVD) to simplify the enormous mathematical matrices scientists use to represent the interactions between electrons and phonons in a material.

The case of phonon interactions is even more complex. These interactions are encoded in multidimensional objects called tensors, generalizations of vectors and matrices in higher dimensions. The complexity of these tensors grows exponentially with the number of particles involved, limiting scientists’ understanding of interactions involving three or more phonons.

Now, inspired by recent advances in machine learning, Bernardi and Luo have developed an AI-based technique that sifts through the high-order tensors that encode phonon interactions in a material and extracts only the crucial bits needed to complete the calculations that explain thermal transport. They describe the work in a paper that appears in the journal Physical Review Letters.

Using current state-of-the-art techniques, a supercomputer takes hours or days to calculate the interactions between three or four phonons in a material. The new method enables computers to complete the same thermal transport and phonon dynamics calculations 1,000 to 10,000 times faster, all while maintaining accuracy.

“The calculations for four-phonon interactions are a nightmare,” Bernardi says. “For complex materials, this task would involve weekslong calculations. Now we can do them in 10 seconds.”

Bernardi explains more about the method:

“We use a machine learning technique called CANDECOMP/PARAFAC tensor decomposition, but we had to adapt it to satisfy the symmetry of this specific physical problem. We first set up a neural network and then run it on GPUs and ask: ‘What are the best functions to approximate the actual tensor that describes these phonon interactions?’

“Once we fix the number of product terms we want to keep, the machine learning process returns the best functions to approximate the full tensor. We typically only need a few of these products, saving orders of magnitude in computational complexity compared to using the full tensor. This method allows us to learn the compressed form of phonon interactions, and we can still use these highly compressed tensors to compute all the observables of interest with the same accuracy.”

Bernardi adds that the new method is well suited for high-throughput screening of thermal physics and heat transport in large material databases, a major effort in the materials community. As for future work, he says, “My vision right now is to compress all different types of quantum interactions and high-order processes in materials with similar techniques. The key will be to bypass the formation of large tensors altogether and to learn the interactions directly in compressed form.”

The paper is titled “Tensor Learning and Compression of N-phonon Interactions.” Additional authors are Dhruv Mangtani, who worked on the project as a SURF student in Bernardi’s lab; Shiyu Peng, a postdoctoral scholar research associate; and Caltech graduate students Jia Yao (MS ’25) and Sergei Kliavinek.

The company announced these new offerings during its Made on YouTube event.

YouTube’s new AI-powered creative partner is a new YouTube Studio tool, is called Ask Studio and can answer questions about things like how the creator’s latest video is performing and what is being said about their editing style, according to a Tuesday blog post.

“It’ll provide personalized and actionable strategic insights based on knowledge of you as a Creator, your channel and how YouTube works,” Amjad Hanif, vice president of creator products at YouTube, said in the post. “We’ll keep adding more capabilities in the future.”

Hanif also said in the post that YouTube has expanded the availability of its AI-powered likeness detection tool in open beta to all YouTube Partner Program creators. This tool helps creators safeguard their identity by detecting, managing and requesting the removal of unauthorized videos made with their facial likeness.

For its livestreaming platform YouTube Live, the company has added AI-powered highlights, a tool that creates lasting content from live content, according to another Tuesday blog post.

“It finds the most compelling moments from the livestream and automatically creates ready-to-share Shorts,” Aaron Filner, senior director, product management at YouTube, said in the post.

YouTube also announced new creation tools for Shorts that can generate video with sound, bring photos to life by applying motion from a video, apply new looks to video footage by applying styles like pop art or origami, add objects to videos via a text description, per another Tuesday blog post.

The company is also experimenting with a feature called Edit with AI that will be added to Shorts and the YouTube Create app and will generate a first draft of a video from the user’s raw camera roll footage, according to the post.

“This gives you a solid starting point so you can jump straight to the fun part: personalizing your video and bringing your unique vision to life,” Dina Berrada, director of product, generative AI creation, at YouTube, said in the post.

To help creators earn more, YouTube has introduced an AI-powered system in YouTube Shopping that tags products in videos, according to another Tuesday blog post.

“We know tagging products can be time-consuming, so to make the experience better for creators, we’re leaning on an AI-powered system to identify the optimal moment a product is mentioned and automatically display the product tag at that time, capturing viewer interest when it’s highest,” Todd Sherman, senior director, product management, and Michael Beckmann, director, product management, data and creator earnings, said in the post.

YouTube parent company Alphabet said in October 2024 that it wants its products—from Google to Android to YouTube—to be synonymous with AI.

Artificial intelligence is transforming how research is created and used in PR and thought leadership. Surveys that once took weeks to design and analyze can now be drafted, fielded and summarized in days or even hours. For communications professionals, the appeal is obvious: AI makes it possible to generate insights that keep pace with the news cycle. But does the quality of those insights hold?

In the race to move faster, an uncomfortable truth is emerging. AI may make aspects of research easier, but it also creates enormous pitfalls for the layperson. Journalists rightfully expect research to be transparent, verifiable and meaningful. This credibility cannot be compromised. Yet an overreliance on AI risks jeopardizing the very characteristics that make research such a powerful tool for thought leadership and PR.

This is where the opportunity and the risk converge. AI can help research live up to its potential as a driver of media coverage, but only if it is deployed responsibly, and never as a total substitute for skilled practitioners. Used without oversight, or by untrained but well-meaning communicators, it produces data that looks impressive on the surface but fails under scrutiny. Used wisely, it can augment and enhance the research process but never supplant it.

The Temptation: Faster, Cheaper, Scalable

AI has upended the traditional pace of research. Writing questions, cleaning data, coding open-ended responses and building reports required days of manual effort. Now, many of these tasks can be automated.

• Drafting: Generative models can create survey questions in seconds, offering PR teams a head start on design.
• Fielding: AI can help identify fraudulent or bot-like responses.
• Analysis: Large datasets can be summarized almost instantly, and open-text responses can be categorized without armies of coders.
• Reporting: Tools can generate data summaries and visualizations that make insights more accessible.

The acceleration is appealing. PR professionals can, in theory, generate surveys and insert data into the media conversation before a trend peaks. The opportunity is real, but it comes with a condition: speed matters only when the research holds up to scrutiny.

The Risk: Data That Doesn’t Stand Up

AI makes it possible to create research faster, but not necessarily better. Fully automated workflows often miss the standards required for earned media.

Consider synthetic respondents, artificial personas generated by AI to simulate human answers to surveys, trained on data from previous surveys. On the surface, they provide instant answers to survey questions. But research shows they diverge from real human data once tested across different groups and contexts. The issue isn’t limited to surveys. Even at the model level, AI outputs remain unreliable. OpenAI’s own system card shows that despite improvements in its newest model, GPT-5 still makes incorrect claims nearly 10% of the time.

For journalists, these shortcomings are disqualifying. Reporters and editors want to know how respondents were sourced, how questions were framed and whether findings were verified. If the answer is simply “AI produced it,” credibility collapses. Worse, errors that slip into coverage can damage brand reputation. Research meant to support PR should build trust, not risk it.

Why Journalists Demand More, Not Less

The reality for PR teams is that reporters are inundated with pitches. That volume has made editors more discerning, and credible data can differentiate a pitch from the competition.

Research that earns coverage typically delivers three things:

1. Clarity: Methods are clearly explained.
2. Context: Results are tied to trends or issues audiences care about.
3. Credibility: Findings are grounded in sound design and transparent analysis.

These expectations have only intensified. Public trust in media is at a historic low. Only 31% of Americans trust the news “a great deal” or “a fair amount.” At the same time, 36% have “no trust at all,” the highest level of complete distrust Gallup has recorded in more than 50 years of tracking. Reporters know this and apply greater scrutiny before publishing any research.

For PR professionals, the implication is clear: AI can speed up processes, but unless findings meet editorial standards, they will never see the light of day.

Why Human Oversight Is Indispensable

AI can process data at scale, but it cannot replicate the judgment or accountability of human researchers. Oversight matters most in four areas:

• Defining objectives: Humans decide which questions are newsworthy or align with campaign goals and what narratives are worth testing.
• Interpreting nuance: Machines can classify sentiment, but are bad at identifying sarcasm, cultural context and emotional cues that shape meaningful insights.
• Accountability: When findings are published, people – not algorithms – must explain the methods and defend the results.
• Bias detection: AI reflects the limitations of its training data. Without human review, skewed or incomplete findings can pass as fact.

Public opinion reinforces the need for this oversight. Nearly half of Americans say AI will have a negative impact on the news they get, while only one in 10 say it will have a positive effect. If audiences are skeptical of AI-created news, journalists will be even more cautious about publishing research that lacks human validation. For PR teams, that means credibility comes from oversight: AI may accelerate the process, but only people can provide the transparency that makes research media ready.

AI as a Partner, Not a Shortcut

AI is best used strategically. It is as an “assistant” that enhances workflows rather than a substitute for expertise. That means:

• Letting AI handle repetitive tasks such as transcription, always with human oversight.
• Documenting when and how AI tools are used, to build transparency.
• Validating AI outputs against human coders or traditional benchmarks.
• Training teams to understand AI’s capabilities and limitations.
• Aligning with evolving disclosure standards, such as the AAPOR Transparency Initiative.

Used this way, AI accelerates processes while preserving the qualities that make research credible. It becomes a force multiplier for human expertise, not a replacement for it.

What’s at Stake for PR Campaigns

Research has always been one of the most powerful tools for earning media. A well-executed survey can create headlines, drive thought leadership and support campaigns long after launch. But research that lacks credibility can do the opposite, damaging relationships with journalists and eroding trust.

Editors are paying closer attention to how AI is being used in PR. Some are experimenting with it themselves, while exercising caution. In Cision’s 2025 State of the Media Report, nearly three-quarters of journalists (72%) said factual errors are their biggest concern with AI-generated material, while many also worried about quality and authenticity. And although some reporters remain open to AI-assisted content if it is carefully validated, more than a quarter (27%) are strongly opposed to AI-generated press content of any kind. Those figures show why credibility cannot be an afterthought: skepticism is high, and mistakes will close doors.

The winners will be teams that integrate AI responsibly, using it to move quickly without cutting corners. They will produce findings that are timely enough to tap into news cycles and rigorous enough to withstand scrutiny. In a crowded media landscape, that balance will be the difference between earning coverage and being ignored.

Conclusion: Credibility as Currency

AI is here to stay in PR research. Its role will only expand, reshaping workflows and expectations across the industry. The question is not whether to use AI, but how to use it responsibly.

Teams that treat AI as a shortcut will see their research dismissed by the media. Teams that treat it as a partner – accelerating processes while upholding standards of rigor and transparency – will produce insights that both journalists and audiences trust.

In today’s environment, credibility is the most valuable currency. Journalists will continue to demand research that meets high standards. AI can help meet those standards, but only when guided by human judgment. The future belongs to PR professionals who prove that speed and credibility are not in conflict, but in partnership.