研究:人腦處理語言的方式與AI模型相似

研究:人腦處理語言的方式與AI模型相似

Hacker News·

一項新研究表明,人腦處理口語的順序與先進的AI語言模型高度相似,這挑戰了基於規則的語言理解理論,並為相關研究引入了一個新的神經數據集。

Image

American Friends of the Hebrew University

Post navigation

Image

December 22, 2025 – The human brain processes spoken language in a sequence that closely mirrors the structure of advanced artificial intelligence (AI) language models, according to a new study that challenges rule-based theories of language comprehension and introduces a publicly available neural dataset for studying how the brain constructs meaning.

In a study published in Nature Communications, researchers led by Dr. Ariel Goldstein of the Hebrew University of Jerusalem Department of Cognitive and Brain Sciences, uncovered a surprising connection between the way our brains make sense of spoken language and the way advanced AI models analyze text.

What the Study Found

When we listen to someone speak, our brain transforms each incoming word through a cascade of neural computations. Goldstein’s team discovered that these transformations unfold over time in a pattern that parallels the tiered layers of AI language models. Early AI layers track simple features of words, while deeper layers integrate context, tone, and meaning. The study found that human brain activity follows a similar progression: early neural responses aligned with early model layers, and later neural responses aligned with deeper layers.

This alignment was especially clear in high-level language regions such as Broca’s area, where the peak brain response occurred later in time for deeper AI layers. According to Dr. Goldstein, “What surprised us most was how closely the brain’s temporal unfolding of meaning matches the sequence of transformations inside large language models. Even though these systems are built very differently, both seem to converge on a similar step-by-step buildup toward understanding.”

The findings suggest that AI is not just a tool for generating text. It may also offer a new window to understand how the human brain processes meaning. For decades, scientists believed that language comprehension relied on symbolic rules and rigid linguistic hierarchies. However, this study challenges that view. Instead, it supports a more dynamic and statistical approach to language, in which meaning emerges gradually through layers of contextual processing.

The researchers also found that classical linguistic features such as phonemes and morphemes did not predict the brain’s real-time activity as well as AI-derived contextual embeddings. This strengthens the idea that the brain integrates meaning in a more fluid and context-driven way than previously believed.

A New Benchmark for Neuroscience

To advance the field, the team publicly released the full dataset of neural recordings paired with linguistic features. This new resource enables scientists worldwide to test competing theories of how the brain understands natural language, paving the way for computational models that more closely resemble human cognition.

The research paper titled “Temporal structure of natural language processing in the human brain corresponds to layered hierarchy of large language models” is now available in Nature Communications and can be accessed here.

Researchers:

Ariel Goldstein 1,2,3, Eric Ham4,5, Mariano Schain3, Samuel A. Nastase 4, Bobbi Aubrey4,6, Zaid Zada 4, Avigail Grinstein-Dabush3, Harshvardhan Gazula4, Amir Feder3, Werner Doyle6, Sasha Devore6, Patricia Dugan6, Daniel Friedman6, Michael Brenner 3,7, Avinatan Hassidim3, Yossi Matias 3, Orrin Devinsky 6, Noam Siegelman 1,8, Adeen Flinker6,9, Omer Levy10, Roi Reichart11, Uri Hasson 3,4

Institutions:

Image

Image

Image

Image

Image

Image

Image

Image

Image

Image

Image

Image

Post navigation

Related articles

New Findings on Antibiotic Resistance Could Provide New Approaches to Bacterial Treatments, According to a Hebrew University Study

January 16, 2026 – A new study reveals that bacteria can survive antibiotic treatment through two fundamentally different modes, resulting in antibiotic persistence, a major cause of treatment failure and difficulty in curing.   For years, persistence has largely been blamed on bacteria

New, Flexible Solar Cell That Works on Windows and Flexible Surfaces Developed by Hebrew University Nanotechnology Researchers

January 12, 2026 – A new solar cell that is designed to work in locations like windows and flexible surfaces where traditional panels can’t, has been developed by nanotechnology researchers at the Hebrew University of Jerusalem (HU).   The research was led by Prof. Shlomo Magdassi

Dentin Inside Wolffish Teeth is a Rare Material: When Compressed Along its Length–it Also Shrinks in Width: An Inspiration for Designing Tougher Materials

January 12, 2026 – The Atlantic wolffish is known for its powerful bite, capable of crushing hard-shelled prey with ease. Now, researchers have discovered that the fish teeth don’t just withstand these extreme forces; they respond in a way that almost no natural hard tissue does. In a new

Image

Image

Image

Keep In Touch

Contact Us

[email protected]
212.607.8500

Careers

[email protected]

Privacy Policy
Financial Disclosure

Newsletter

Social

Subscribe

Topic Interests

(function($) {window.fnames = new Array(); window.ftypes = new Array();fnames[2]='FNAME';ftypes[2]='text';fnames[3]='LNAME';ftypes[3]='text';fnames[0]='EMAIL';ftypes[0]='email';fnames[17]='MMERGE17';ftypes[17]='text';fnames[30]='MMERGE30';ftypes[30]='text';fnames[29]='MMERGE29';ftypes[29]='text';fnames[28]='ADDRESS';ftypes[28]='address';fnames[14]='AGE';ftypes[14]='radio';fnames[12]='MMERGE12';ftypes[12]='text';fnames[11]='MMERGE11';ftypes[11]='text';fnames[6]='MMERGE6';ftypes[6]='text';fnames[27]='MMERGE27';ftypes[27]='text';fnames[26]='MMERGE26';ftypes[26]='text';fnames[25]='MMERGE25';ftypes[25]='text';fnames[23]='MMERGE23';ftypes[23]='text';fnames[21]='MMERGE21';ftypes[21]='text';fnames[18]='MMERGE18';ftypes[18]='text';fnames[16]='MMERGE16';ftypes[16]='radio';fnames[1]='SIGNUP';ftypes[1]='text';fnames[24]='MMERGE24';ftypes[24]='text';fnames[22]='MMERGE22';ftypes[22]='text';fnames[20]='MMERGE20';ftypes[20]='text';fnames[19]='MMERGE19';ftypes[19]='text';fnames[15]='MMERGE15';ftypes[15]='text';fnames[13]='MMERGE13';ftypes[13]='text';fnames[10]='MMERGE10';ftypes[10]='text';fnames[9]='MMERGE9';ftypes[9]='text';fnames[8]='MMERGE8';ftypes[8]='text';fnames[7]='MMERGE7';ftypes[7]='text';fnames[5]='MMERGE5';ftypes[5]='text';fnames[4]='DONOR_ID';ftypes[4]='text';}(jQuery));var $mcj = jQuery.noConflict(true);

Position Summary:

The CEO of AFHU will join the organization at a pivotal moment in time, succeeding a successful leader and building on a legacy of tremendous growth.  In partnership with Hebrew University, a high-performing team, and a committed Board, the new CEO will grow the donor base, build new collaborations, and position the organization to achieve even more ambitious goals.

The CEO is accountable for the financial and operational management of AFHU, and for executing a shared strategic vision to meet Hebrew University’s priorities and fundraising goals. The CEO of AFHU works closely with the Hebrew University administration and faculty and will lead a culture of cooperation and transparency with colleagues across the University. The trusted partnership between the University and AFHU is critical to success. The CEO will work closely with the Board of AFHU to build a best-in-class fundraising apparatus and infrastructure across the United States, lead a high-performing executive management team, and galvanize resources from individual donors, foundations, and beyond. A central part of the CEO’s role is the recruitment, training, and retention of development staff. It will be important for the future success of AFHU that the organization creates a dynamic, supportive, and effective development team that is impactful with donors across the United States.

The Chief Executive will:

Monitor the fundraising team’s progress against performance targets, review and approve service strategies, metrics, and budgets; and ensure that succession plans are in place and current for all resource development leadership positions.

Have overall responsibility for the day-to-day management of the organization, including strategic, fiscal, and operational accountability.

Lead the organization’s management team in prioritizing, implementing and enhancing its resources in order to create impact.

Ensure that all activities and programs meet the highest standards of ethics, governance, and transparency and are aligned with AFHU’s core values.

The CEO will be a front-facing leader who is responsible for managing and soliciting a portfolio of major and principal prospects and cultivating senior-level relationships with a variety of donors, including individuals, foundations, and other partners within and outside of the Hebrew University alumni community. The CEO will lead a team of major gifts staff managing their own portfolios and will ensure that, at all times, AFHU works collaboratively with the Hebrew University.

Overview:

The CEO will be an inspiring and engaging individual with proven ability to build organizational capacity, strong leadership and management skills, and excellent fundraising acumen.

They will also possess a commitment to the mission and values of the Hebrew University paired with strong interpersonal skills to cultivate new and build on existing partnerships. As a senior spokesperson for Hebrew University, the CEO must be an outstanding communicator with the stature, gravitas, and credibility to cultivate relationships at the highest levels across a variety of donors and partners.

Pivotal Experience & Expertise:

Strategic Leadership – A proven track record of developing and/or implementing a strategic plan and enhancing an organization’s capacity and impact, ideally in a fundraising organization.

Someone who quickly grasps the opportunities and challenges of the organization and the role and understands how to work effectively with the Board and the broader AFHU structure, has the capacity to develop innovative and creative fundraising plans that integrate multiple priorities and goals, and has a strong analytical mindset and uses data to support thinking and decision-making; the inclination to seek and analyze data from a variety of sources to support decisions and to align others with the organization’s overall strategy.

Team Leadership & Management – Demonstrated strong leadership and people management skills, including a previous track record of leading a fundraising team to be successful.

Someone who has the personal inclination and professional ability to be a positive and unifying figure who can lead by influence and example, in order to continue strengthening a collaborative development culture between AFHU and Hebrew University. A leader who is viewed by others as having a high degree of integrity and forethought in his/her approach to making decisions.

Fundraising – Strong experience and ability to raise substantial sums of money from a variety of sources. Track record of coaching a team of fundraisers to execute campaigns.

Culture Fit & Impact:

Mission Orientation – Strong connection to and passion for the purpose of AFHU: to connect the passions of Americans to the talent at the Hebrew University of Jerusalem.

Relationship-Builder – Strong interpersonal skills and demonstrated ability building relationships, networks, and partnerships both internally and externally. Proven commitment to building consensus across the organization, but also the professional courage to make and communicate tough decisions when necessary.

Ambassador – Articulate communicator and ambassador for AFHU and HU with all audiences; knowledgeable of the issues affecting the university with the ability to advocate and communicate these issues.

Engagement Team:

Nominations and applications can be directed to: [email protected]

J.J. Cutler: Phone: + 1 215 527 0658 Email: [email protected]

Jackie Zavitz: Phone: + 1 215 901 9098 Email: [email protected]

Erin Kneeley: Phone: + 1 215 636 5066 Email: [email protected]

Ana Rosado: Phone: +1 212 699 3007 Email: [email protected]

Hacker News

相關文章

  1. 不同的AI模型在編碼現實的方式上似乎趨於一致

    4 個月前

  2. 趨同演化:不同語言模型如何學習相似的數字表示法

    5 天前

  3. 隆重推出 Falcon-H1-Arabic:以混合架構突破阿拉伯語 AI 的界限

    Huggingface · 4 個月前

  4. 機器是否終將具備智慧?

    Microsoft Research · 大約 1 個月前

  5. AI 模型是進行推理還是僅僅鸚鵡學舌?

    3 個月前

其他收藏 · 0