ClaimsPeople

Yann LeCun

Yann LeCun is a computer scientist who shared the 2018 ACM Turing Award, is a Jacob T. Schwartz Professor at NYU's Courant Institute, was Meta's Chief AI Scientist and founding director of FAIR, and is Executive Chairman of AMI Labs.

1,175 claims collected, 197 researched, 84 resolved. Report of 4 November 2026, method 0.1.0+c876030c.

Scores by kind of claim

Each kind of claim is scored on its own, out of 100: the share that held up, with the range it likely falls in. Fewer resolved claims give a wider range.

  • Factual claims

    Statements about the present or the past, checked against the record: did they hold up?

    78 72 83

    High confidence, 79 factual claims resolved. From 16 documents, 2006 to 2026; 19 documents sampled from 33 candidates found.

    Not scored: 43 unresolvable.

  • Predictions

    Statements about events outside the person's control, checked once they fell due: did they come true?

    64 43 83

    Low confidence, 5 predictions resolved. From 4 documents, 2010 to 2024; 19 documents sampled from 33 candidates found.

    Not scored: 66 unresolvable. 3 more fall due later.

    Does not yet allow for how likely each prediction was when made.

  • Promises

    Commitments to the person's or organisation's own future action, with something to deliver: was it delivered?

    Insufficient record: 0 promises resolved, 5 needed for a score.

    Not scored: 1 unresolvable.

All types combined: 77 (likely 71 to 82, high confidence), from 84 claims: 79 factual claims, 5 predictions. It mixes kinds of claim in different proportions for different people, so compare people kind by kind.

Yann LeCun's claims

1,175 claims, sorted by newest said. Choose a claim to read it in full.
VerdictKindSaidClaimTopic
Not checkable Not checkable 26 May 2014 Only the first is interesting to me (the other uses are interesting too, just not on my own radar screen). Not checkable · 26 May 2014
Not checkable Not checkable 26 May 2014 Proofs always come later. Not checkable · 26 May 2014
Not checkable Not checkable 26 May 2014 I like the idea of Boltzmann machines (it’s a beautifully simple concept) but it doesn’t scale well. Not checkable · 26 May 2014
Not checkable Not checkable 26 May 2014 The theory of deep learning is a wide open field. Not checkable · 26 May 2014
Not checkable Not checkable 26 May 2014 Ask them what error rate they get on MNIST or ImageNet. Not checkable · 26 May 2014
Not checkable Not checkable 26 May 2014 You can construct pathological cases where the objective function is like a golf course: flat with a tiny hole somewhere. Not checkable · 26 May 2014
Not checkable Not checkable 26 May 2014 DeconvNets are the generative counterpart of feed-forward ConvNets. Not checkable · 26 May 2014
Not checkable Not checkable 26 May 2014 Learning with temporal/sequential signals: language, video, speech. Not checkable · 26 May 2014
Not checkable Not checkable 26 May 2014 Seriously, I don’t like the phrase “Big Data”. Not checkable · 26 May 2014
Not checkable Not checkable 26 May 2014 They have the advantage of scaling well. Not checkable · 26 May 2014
Not checkable Not checkable 26 May 2014 Jeff Hawkins has the right intuition and the right philosophy. Not checkable · 26 May 2014
Not checkable Not checkable 26 May 2014 Few of us believe that we have found a good solution to unsupervised learning. Not checkable · 26 May 2014
Unresolvable Factual 15 May 2014 Theoretical results suggest that learning good parameters for a small neural network can be as computationally hard as breaking RSA. Factual · 15 May 2014 Philosophy, ethics and society
Mostly true Factual 15 May 2014 The notions of partition functions, entropy, free energy and variational methods prevalent in the graphical models literature come from statistical physics. Factual · 15 May 2014 Science and research methods
Not researched Factual 15 May 2014 SVMs do not naturally emerge from VC theory; they are simple enough for VC theory to give specific results about them. Factual · 15 May 2014 Science and research methods
Unresolvable Factual 15 May 2014 The limitations the questioner pointed out concern not just backprop but all learning algorithms that use gradient-based optimization. Factual · 15 May 2014 AI capabilities
Mostly true Factual 15 May 2014 The first commercially viable check reading system, deployed by AT&T/NCR in 1996, used a ConvNet rather than an SVM. Factual · 15 May 2014 AI capabilities
Mostly false Factual 15 May 2014 Convolutional nets were always the record holder on MNIST, and SVMs never really managed to beat them there. Factual · 15 May 2014 AI capabilities
Unresolvable Prediction 15 May 2014 Researchers will eventually figure out how to merge ConvNets and DeconvNets into a feed-forward plus feed-back system trainable supervised or unsupervised. Prediction · 15 May 2014 AI capabilities
Mostly true Factual 15 May 2014 No one uses generalization bounds to do model selection; practitioners use cross-validation. Factual · 15 May 2014 Science and research methods
Unresolvable Prediction 15 May 2014 Now or in the near future, most of the world's knowledge will be extracted by machines and reside in machines. Prediction · 15 May 2014 AI timelines and forecasting
Not researched Factual 15 May 2014 LeCun said in May 2014 that, as far as he could tell, HTM had not been demonstrated to get anywhere close to state-of-the-art results on any serious task. Factual · 15 May 2014 AI capabilities
Unresolvable Prediction 15 May 2014 As more data becomes available, machine learning methods will increasingly rely on learning. Prediction · 15 May 2014 AI capabilities
Not researched Factual 15 May 2014 Deep learning has become the dominant method for acoustic modeling in speech recognition and is quickly becoming dominant for vision tasks such as object recognition, detection and semantic segmentation. Factual · 15 May 2014 AI capabilities
Mostly true Factual 15 May 2014 Speech is a domain with access to ridiculously large amounts of data and a very large number of categories. Factual · 15 May 2014 AI capabilities
Unresolvable Prediction 15 May 2014 Natural language processing is the next frontier for deep learning. Prediction · 15 May 2014 AI timelines and forecasting
Mostly true Factual 15 May 2014 Facebook AI Research uses Torch7 for many projects, as do Deep Mind and several groups at Google. Factual · 15 May 2014 Technology
Not checkable Not checkable 2010 Implementing ConvNets in software is best achieved using the modular, object-oriented approach suggested in [2]. Not checkable · 2010
Not checkable Not checkable 2010 It is clear that abs and normalization are cruciala for achieving good performance. Not checkable · 2010
Not researched Factual 2010 To the authors' knowledge, 0.53% is the lowest error rate reported on the original MNIST dataset without distortions or preprocessing; the best previously reported error rate was 0.60%. Factual · 2010 AI capabilities
Not checkable Not checkable 2010 We describe new unsupervised learning algorithms, and new non-linear stages that allow ConvNets to be trained with very few labeled samples. Not checkable · 2010
Not researched Factual 2010 Training deep multi-stage architectures with supervised gradient back-propagation requires many labeled samples. Factual · 2010 AI capabilities
Not researched Factual 2010 Works [5] and [1] simplified the architecture and used back-propagation to train the entire system in a supervised fashion. Factual · 2010 AI capabilities
Not researched Factual 2010 The supervised ConvNet approach was very successful for OCR and handwriting recognition. Factual · 2010 AI capabilities
Not researched Factual 2010 SIFT and HoG are composed of filter banks followed by non-linearities and pooling (histogramming). Factual · 2010 AI capabilities
Mostly true Prediction 2010 In the next decade, a major challenge of computer vision, machine learning and AI will be devising methods that learn good feature hierarchies from unlabeled and labeled data in an integrated fashion. Prediction · 2010 AI capabilities
Not researched Citation 2010 Recent deep learning research [30],[31],[32] has shown unsupervised learning can train each stage one after another using only unlabeled data, significantly reducing the need for labeled samples. Citation · 2010 AI capabilities
Unresolvable Prediction 2010 The authors expected to see many new embedded vision systems based on ConvNets in the next few years. Prediction · 2010 AI compute, hardware and data
Not checkable Not checkable 2010 The Convolutional Network architecture is a remarkably versatile, yet conceptually simple paradigm that can be applied to a wide spectrum… Not checkable · 2010
Resolved true Factual 2010 Each stage of a ConvNet is composed of a filter bank, some non-linearities, and feature pooling layers. Factual · 2010 AI capabilities
Resolved true Factual 2010 The idea behind multi-stage Hubel-Wiesel architectures is rooted in Hubel and Wiesel's 1962 work on the cat's primary visual cortex. Factual · 2010 Science and research methods
Not researched Factual 2010 In the reported experiments, comparing rows R vs R+ and U vs U+, supervised fine tuning consistently improves performance, particularly with weak non-linearities. Factual · 2010 AI capabilities
Not researched Factual 2010 A two-stage ConvNet with unsupervised pre-training and supervised refinement achieved a 0.53% test error rate on MNIST. Factual · 2010 AI capabilities
Not researched Factual 2010 Accuracy of 65.5% was obtained using unsupervised pre-training and supervised refinement with abs and normalization non-linearities. Factual · 2010 AI capabilities
Not checkable Not checkable 2010 Current and future research will focus on performing unsupervised learning on multiple stages simultaneously, on the integration of… Not checkable · 2010
Not checkable Not checkable 2010 The unsupervised method we propose, to learn the filter coefficients in the filter bank layers, is called Predictive Sparse Decomposition… Not checkable · 2010
Mostly true Factual 2010 A face detection application on the CNP system was demonstrated achieving 10 fps on VGA images. Factual · 2010 AI compute, hardware and data
Not researched Factual 2010 Predictive Sparse Decomposition trains a feed-forward encoder C(X,K) to quickly approximate the sparse solution Z*. Factual · 2010 AI capabilities
Not checkable Not checkable 2010 Good internal representations are hierarchical. Not checkable · 2010
Not checkable Not checkable 2010 This is an extremely important fact for users of convolutional networks, which traditionally only use tanh(). Not checkable · 2010
Not researched Factual 2010 In traditional computer vision and pattern recognition, internal representations are produced by a hand-crafted feature extractor whose output is fed to a trainable classifier. Factual · 2010 AI capabilities
Not checkable Not checkable 2010 Convolutional Networks (ConvNets) are an answer to the first question. Not checkable · 2010
Mixed Factual 2010 ConvNets have been successfully deployed in many commercial applications from OCR to video surveillance, but require large amounts of labeled training samples. Factual · 2010 AI capabilities
Not researched Factual 2010 Random filters without any filter learning achieve decent performance (62.9% for R) as long as abs and normalization are present. Factual · 2010 AI capabilities
Mostly true Factual 2010 Convolutional Networks are a biologically inspired trainable architecture that can learn invariant features. Factual · 2010 AI capabilities
Unresolvable Factual 2010 A big advantage of ConvNets for detection is their computational efficiency. Factual · 2010 AI capabilities
Not checkable Not checkable 2010 One of the key questions of Vision Science (natural and artificial) is how to produce good internal representations of the visual world. Not checkable · 2010
Not researched Factual 2010 Using rectification, contrast normalization and unsupervised pre-training of each filter bank considerably reduces the need for labeled samples in ConvNets. Factual · 2010 AI capabilities
Not researched Factual 2010 The CNP developed by the authors' group achieves 10 GOPS for 7x7 kernels and implements entire programmable ConvNets including pre/postprocessing. Factual · 2010 AI compute, hardware and data
Not checkable Not checkable 2010 Convolutional Networks [1], [2] are trainable multistage architectures composed of multiple stages. Not checkable · 2010
Not checkable Not checkable 2010 This suggests that recognition architectures for vision (and for other modalities such as audio and natural language) should have multiple… Not checkable · 2010
Not checkable Not checkable 2010 Intelligent tasks, such as visual perception, auditory perception, and language understanding require the construction of good internal… Not checkable · 2010
Not checkable Not checkable 2010 ConvNets are perfect candidates for hardware implementations, and embedded applications, as demonstrated by the increasing amount of work… Not checkable · 2010
Not checkable Not checkable 2010 A major question for Machine Learning is how to learn such good features automatically. Not checkable · 2010
Not checkable Not checkable 2010 It seems that unsupervised pre-training (U , U +) is crucial when newly proposed non-linearities are not in place. Not checkable · 2010
Not researched Factual 2010 Spatially translating the input of a ConvNet feature detection layer translates the output but leaves it otherwise unchanged. Factual · 2010 AI capabilities
Resolved true Factual 2010 In many problems labeled data is scarce whereas unlabeled data is abundant. Factual · 2010 AI capabilities
Not checkable Not checkable 2006 In the following, we use the convention that small energy values correspond to highly compatible configurations of the variables, while… Not checkable · 2006
Not researched Factual 2006 Late normalization solves several problems associated with the internal normalization of HMMs and Bayesian nets. Factual · 2006 AI capabilities
Not checkable Not checkable 2006 For decision-making tasks, such as steering a robot, it is merely necessary that the system give the lowest energy to the correct answer. Not checkable · 2006
Not researched Factual 2006 Probabilistic models must be properly normalized, which sometimes requires evaluating intractable integrals over all possible variable configurations. Factual · 2006 Science and research methods
Not checkable Not checkable 2006 Among the many loss functions proposed in the literature, some are good (with a non-zero margin), and some can be bad. Not checkable · 2006
Not researched Factual 2006 In the experiment, learning with the NLL loss was much faster than with the square-square loss. Factual · 2006 AI capabilities
Not researched Factual 2006 The loss is minimized by placing the V's apex at Y=X^2, which raises the energies of all other answers because the V has a single minimum. Factual · 2006 AI capabilities
Not checkable Not checkable 2006 However, the convexity of the loss function, while mathematically satisfying, does not seem to be a significant practical advantage. Not checkable · 2006
Not researched Factual 2006 Under the NLL loss, the desired answer may have different, non-zero energies for different X. Factual · 2006 AI capabilities
Not checkable Not checkable 2006 For example, there is no reason to believe that the model in Figure 7 is normalizable over Y . Not checkable · 2006
Not researched Factual 2006 Most probabilistic models can be viewed as energy-based models whose energy function satisfies normalizability conditions and whose loss has a particular form. Factual · 2006 Science and research methods
Not researched Factual 2006 A linear model with the NLL loss is logistic regression. Factual · 2006 AI capabilities
Not researched Factual 2006 The minimum classification error (MCE) loss is non-convex. Factual · 2006 AI capabilities
Not checkable Not checkable 2006 A common approach is to view variational methods and energy-based loss functions as approximations to the probabilistic method. Not checkable · 2006
Not researched Factual 2006 The GTN system trained with NLL loss in LeCun et al. 1998a assigns a well-defined probability distribution over label sequences, contrary to the claim in Lafferty et al. 2001. Factual · 2006 AI capabilities
Not researched Factual 2006 In speech recognition, segmentation into words and phonemes must occur simultaneously with recognition, while the correct phoneme segmentation is rarely available during training. Factual · 2006 AI capabilities
Not checkable Not checkable 2006 In a sense, variational methods could be interpreted in the context of EBM as a way to choose a subset of energies to pull up. Not checkable · 2006
Not researched Factual 2006 The perceptron, hinge and NLL losses can be directly translated to the multiclass case. Factual · 2006 AI capabilities
Not checkable Not checkable 2006 One advantage of the energy-based approach is that it puts very little restrictions on the nature of E. Not checkable · 2006
Not researched Factual 2006 The perceptron loss has no mechanism for creating an energy gap between the correct answer and incorrect ones. Factual · 2006 Science and research methods
Not checkable Not checkable 2006 To some observers, the recent interest in the linear structured model looks like somewhat of a throw-back to the past, and a regression on… Not checkable · 2006
Not researched Factual 2006 The energy loss does not satisfy condition 3 for a general architecture. Factual · 2006 AI capabilities
Not researched Factual 2006 The quality of probability estimates obtained from energies depends on how the energy is estimated from data, not on how energies are converted to probabilities. Factual · 2006 Science and research methods
Mostly true Factual 2006 The hinge loss with G_W(X)=W^T X and a ||W||^2 regularizer gives the linear support vector machine. Factual · 2006 AI capabilities
Mostly true Factual 2006 The hinge loss systematically chooses the most offending incorrect answer as the one whose energy is pulled up. Factual · 2006 AI capabilities
Not researched Factual 2006 The negative log-likelihood loss pulls up the energies of all incorrect answers at each iteration, including those unlikely to have lower energy than the correct answer. Factual · 2006 AI capabilities
Not researched Factual 2006 EBMs manipulate energies, so no normalization is necessary. Factual · 2006 AI capabilities
Not checkable Not checkable 2006 Energy minimization is a convenient way to represent the general process of reasoning and inference. Not checkable · 2006
Not checkable Not checkable 2006 Any prior knowledge we may have about the task at hand is embedded into the architecture and into the loss function (particularly the… Not checkable · 2006
Not researched Factual 2006 Because the MEE loss saturates the contribution of any single error, it is more robust to label noise and outliers but is non-convex. Factual · 2006 AI capabilities
Not researched Factual 2006 The lack of a margin in the perceptron loss, which may theoretically cause stability problems, was overlooked in LeCun et al. 1998a. Factual · 2006 Their own work or organisation
Not researched Factual 2006 The quality of an answer in an energy-based model cannot be measured by its energy alone without considering the energies of all other answers. Factual · 2006 AI capabilities
Not checkable Not checkable 2006 We should emphasize again that this is inconsequential if the model is to be used for prediction, classification, or decision-making. Not checkable · 2006
Back to the list

Yann LeCun

Open its page

Factual AI capabilities

“The approach was very successful for such tasks as OCR and handwriting recognition.”

Yann LeCun, co-author of paper 2010 Convolutional Networks and Applications in Vision (yann.lecun.com)

The supervised ConvNet approach was very successful for OCR and handwriting recognition.

Not researched Checkable, but outside the sample that was researched.

Cite this claim

Yann LeCun, “The supervised ConvNet approach was very successful for OCR and handwriting recognition.” (said 2010). Verdict: not researched. Reliability API, claim c_244a04ad5455, method 0.1.0+c876030c. https://epistacklabs.com/claims/c_244a04ad5455

Claims without a verdict

Is a verdict here wrong, or do you speak for Yann LeCun? Reply on the claim's own page, with the form at its foot (how replies work). Verdicts are made automatically; how verdicts and scores are made.