# Brenden Lake

Brenden Lake is an NYU professor and cognitive scientist researching human-level AI, focusing on learning and reasoning from minimal data. His work bridges machine learning and cognitive psychology to build more human-like artificial intelligence.

**Brenden Lake** is an associate professor of psychology and data science at New York University, where he leads the Human & Machine Learning Lab. His research sits at the intersection of [artificial-intelligence](https://www.wikiprompt.org/wiki/artificial-intelligence), [machine-learning](https://www.wikiprompt.org/wiki/machine-learning), and cognitive science, aiming to understand and replicate the remarkable efficiency with which humans learn and generalize from sparse examples. Lake is best known for his work on human-level concept learning, particularly the development of models that can learn from a single example, a capability that contrasts sharply with the data-hungry nature of most modern [deep-learning](https://www.wikiprompt.org/wiki/deep-learning) systems.

Lake's academic trajectory has been defined by a commitment to integrating computational models with behavioral experiments. He completed his Ph.D. in computational neuroscience at the Massachusetts Institute of Technology, where he was advised by [joshua-tenenbaum](https://www.wikiprompt.org/wiki/joshua-tenenbaum), a leading figure in computational cognitive science. This foundation shaped his perspective that progress in AI requires not only scaling up models but also understanding the underlying principles of human cognition.

## Early Career and Academic Background

Born in the United States, Lake pursued his undergraduate studies at the University of California, Berkeley, where he earned a degree in cognitive science. His interest in the intersection of psychology and computation led him to MIT, where he received his Ph.D. in 2014. His doctoral dissertation focused on learning concepts from few examples, a theme that would become central to his later research.

After completing his Ph.D., Lake conducted postdoctoral research at MIT before joining the faculty at New York University in 2017. At NYU, he holds appointments in both the Department of Psychology and the Center for Data Science, reflecting his interdisciplinary approach. His lab has become a hub for researchers interested in building AI systems that learn and reason like humans.

## Human-Level Concept Learning

One of Lake's most significant contributions is the development of computational models that can learn new concepts from a single example. In a landmark 2015 paper published in *Science*, Lake and his colleagues introduced a model called Bayesian Program Learning (BPL). This model could learn to recognize and generate novel handwritten characters after seeing just one instance, achieving performance comparable to human learners.

BPL works by decomposing a visual concept into a probabilistic program - a structured representation that captures the underlying strokes and their relationships. This approach contrasts with the [neural-network](https://www.wikiprompt.org/wiki/neural-network) models that dominated the field at the time, which typically require thousands of examples to achieve similar accuracy. Lake's work demonstrated that structured, compositional representations could enable far more efficient learning.

This research has had a lasting impact on the field, inspiring subsequent work on few-shot learning and meta-learning. It also highlighted the importance of incorporating inductive biases that reflect human cognitive priors, such as the tendency to decompose complex objects into simpler parts.

## The Abstraction and Reasoning Corpus

In 2019, Lake and his colleague François Chollet introduced the Abstraction and Reasoning Corpus (ARC), a benchmark designed to measure general fluid intelligence in AI systems. ARC consists of visual reasoning tasks that require the ability to infer abstract patterns and apply them to novel situations. Unlike many standard benchmarks, ARC tasks are designed to be easy for humans but challenging for machines, as they require core knowledge about objectness, spatial relations, and transformations.

ARC has become a widely used evaluation tool in the AI community, spurring research on systematic generalization and out-of-distribution reasoning. Lake has argued that current [large-language-model](https://www.wikiprompt.org/wiki/large-language-model)s and [transformer](https://www.wikiprompt.org/wiki/transformer)-based architectures, despite their impressive scale, often fail on ARC tasks because they lack the compositional and causal reasoning abilities that humans possess. This work has positioned Lake as a prominent critic of the idea that scaling alone will lead to human-level AI.

## Systematic Generalization and Compositionality

A recurring theme in Lake's research is the importance of systematic generalization - the ability to understand and generate novel combinations of known concepts. In a 2019 paper published in *Science*, Lake and his colleagues demonstrated that standard neural networks often fail to generalize systematically when tested on new combinations of words and meanings. For example, a model trained to understand the phrase "jump twice" and "spin left" might fail to understand "jump left" or "spin twice."

This work highlighted a fundamental limitation of many [deep-learning](https://www.wikiprompt.org/wiki/deep-learning) approaches and argued for the need to incorporate compositional inductive biases. Lake has explored various methods to address this, including meta-learning techniques that train models to learn new tasks quickly, and hybrid architectures that combine neural networks with symbolic reasoning. His research suggests that achieving human-level AI will require a synthesis of connectionist and symbolic approaches.

## Cognitive Science and AI Integration

Lake is a strong advocate for the idea that AI research should be informed by cognitive science. He has argued that studying how humans learn, reason, and generalize can provide valuable insights for building more robust and efficient AI systems. This perspective is reflected in his collaborations with psychologists and neuroscientists, as well as his involvement in initiatives that bridge the two fields.

In a widely cited 2017 paper titled "Building Machines That Learn and Think Like People," Lake and his co-authors outlined a roadmap for AI research based on cognitive principles. They proposed that future AI systems should be able to learn from few examples, reason causally, and use intuitive physics and psychology. This paper has been influential in shaping discussions about the limits of current approaches and the potential for more human-like AI.

## Recent Work and Future Directions

In recent years, Lake has extended his research to explore the capabilities and limitations of [generative-ai](https://www.wikiprompt.org/wiki/generative-ai) models. He has investigated whether large language models can exhibit systematic generalization when prompted appropriately, and whether they can learn new concepts from minimal examples. His findings have been mixed, suggesting that while these models are powerful, they still lack some core aspects of human cognition.

Lake has also been involved in efforts to develop new benchmarks and evaluation methods that better capture human-like intelligence. He has argued for the importance of testing AI systems on tasks that require active learning, causal reasoning, and the ability to adapt to novel environments. His ongoing work aims to build models that not only perform well on specific tasks but also exhibit the flexibility and robustness of human cognition.

## Impact and Recognition

Lake's research has been widely recognized for its originality and influence. He has received several awards, including the National Science Foundation CAREER Award and the Sloan Research Fellowship. His papers have been published in top journals such as *Science* and *Nature Human Behaviour*, and he is a frequent invited speaker at major AI and cognitive science conferences.

Beyond academia, Lake's ideas have influenced discussions in the AI industry about the future of the field. His critiques of scaling-based approaches have prompted researchers at companies like [openai](https://www.wikiprompt.org/wiki/openai) and [google-deepmind](https://www.wikiprompt.org/wiki/google-deepmind) to consider alternative architectures and training paradigms. He has also contributed to public discourse on AI safety and the importance of building systems that align with human values.

## Teaching and Mentorship

At NYU, Lake teaches courses on machine learning and cognitive science, mentoring a new generation of researchers who are interested in the intersection of these fields. His students have gone on to positions in academia and industry, continuing to push the boundaries of human-level AI research. Lake is known for his collaborative and open-minded approach, encouraging interdisciplinary thinking and rigorous experimentation.

Lake's vision for AI is one where machines are not just tools but partners that can learn and reason alongside humans. His work challenges the field to move beyond narrow benchmarks and toward a deeper understanding of intelligence itself. As AI continues to advance, Lake's contributions will likely remain central to the quest for machines that truly think and learn like people.

## Selected Publications

Lake has authored numerous influential papers, including:

- "Human-level concept learning through probabilistic program induction" (2015, *Science*)
- "Building machines that learn and think like people" (2017, *Behavioral and Brain Sciences*)
- "Generalization without systematicity: On the compositional skills of sequence-to-sequence recurrent networks" (2018, *ICML*)
- "The Abstraction and Reasoning Corpus" (2019, *arXiv*)
- "Systematic generalization: What is required and can it be learned?" (2019, *Science*)

These works have collectively shaped the research agenda for human-level AI, inspiring both academic and industrial efforts to build more capable and human-like systems.

## Conclusion

Brenden Lake stands as a leading figure in the effort to create AI that mirrors human learning and reasoning. His interdisciplinary approach, combining cognitive science with machine learning, has produced foundational insights into how machines can learn from few examples and generalize systematically. While the goal of human-level AI remains distant, Lake's research provides a principled path forward, emphasizing the importance of structure, compositionality, and cognitive realism. His work will continue to influence the field for years to come, as researchers grapple with the challenge of building machines that truly understand the world.

---
Source: https://www.wikiprompt.org/wiki/brenden-lake
License: CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/)
Last updated: 2026-09-12T22:24:15.798633+00:00