Deep learning (also known as deep structured learning or hierarchical learning ) is part of a broader family of machine learning methods based on learning data representations , as opposed to task-specific algorithms. Learning can be supervised , partially supervised or unsupervised . [6] [9] [11]

Some representations are loosely based on interpretation of information processing and communication patterns in a biological nervous system , such as neural coding that attempts to define a relationship between various stimuli and associated neuronal responses in the brain . [14]

Deep learning architectures such as , deep belief networks and recurrent neural networks have been applied to fields including computer vision , speech recognition , natural language processing , audio recognition, social network filtering, machine translation , bioinformatics and drug design [15] , where they have produced results comparable to and in some cases superior [17] to human experts. [18]


Deep learning is a class of machine learning algorithms that: [20] ( pp199–200 )

Layers that have been used in deep learning include hidden layers of an artificial neural network and sets of propositional formulas . [22] They may also include latent variables organized layer-wise in deep generative models such as the nodes in Deep Belief Networks and Deep Boltzmann Machines .

Credit assignment

  • Credit assignment path (CAP) [9] – A chain of transformations from input to output. CAPs describe potentially causal connections between input and output.
  • CAP depth – for a feedforward neural network , the depth of the CAPs is that of the network and is the number of hidden layers plus one (as the output layer is also parameterized), but for recurrent neural networks , in which a signal may propagate through a layer more than once, the CAP depth is potentially unlimited.
  • Deep/shallow – No universally agreed upon threshold of depth divides shallow learning from deep learning, but most researchers agree that deep learning involves CAP depth > 2.


The assumption underlying distributed representations is that observed data are generated by the interactions of layered factors.

Deep learning adds the assumption that these layers of factors correspond to levels of abstraction or composition. Varying numbers of layers and layer sizes can provide different degrees of abstraction. [6]

Deep learning exploits this idea of hierarchical explanatory factors where higher level, more abstract concepts are learned from the lower level ones. . [23] [24]

Deep learning architectures are often constructed with a greedy layer-by-layer method. Deep learning helps to disentangle these abstractions and pick out which features are useful for improving performance. [6]

For supervised learning tasks, deep learning methods obviate feature engineering , by translating the data into compact intermediate representations akin to principal components , and derive layered structures that remove redundancy in representation.

Deep learning algorithms can be applied to unsupervised learning tasks. This is an important benefit because unlabeled data are more abundant than labeled data. Examples of deep structures that can be trained in an unsupervised manner are neural history compressors [25] and deep belief networks . [6] [26]


Deep neural networks are generally interpreted in terms of the universal approximation theorem [28] [29] [30] [31] or probabilistic inference . [20] [22] [6] [9] [26] [32]

The universal approximation theorem concerns the capacity of feedforward neural networks with a single hidden layer of finite size to approximate continuous functions . [28] [29] [30] [31] In 1989, the first proof was published by Cybenko for sigmoid activation functions [28] and was generalised to feed-forward multi-layer architectures in 1991 by Hornik. [29]

The probabilistic interpretation [32] derives from the field of machine learning . It features inference, [20] [22] [6] [9] [26] [32] as well as the optimization concepts of training and testing , related to fitting and generalization , respectively. More specifically, the probabilistic interpretation considers the activation nonlinearity as a cumulative distribution function . [32] The probabilistic interpretation led to the introduction of dropout as regularizer in neural networks. [34] The probabilistic interpretation was introduced by researchers including Hopfield , Widrow and Narendra and popularized in surveys such as the one by Bishop . [35]


The term Deep Learning was introduced to the machine learning community by Rina Dechter in 1986, [36] [25] and to Artificial Neural Networks by Igor Aizenberg and colleagues in 2000, in the context of Boolean threshold neurons. through neural networks for reinforcement learning . In 2006, a publication by Geoff Hinton , Osindero and Teh [39] [41] showed how a many-layered feedforward neural network could be effectively pre-trained one layer at a time, treating each layer in turn as an unsupervised restricted Boltzmann machine , then fine-tuning it using supervised backpropagation . The paper referred to learning for deep belief nets.

The first general, working learning algorithm for supervised, deep, feedforward, multilayer perceptrons was published by Alexey Ivakhnenko and Lapa in 1965. [23] A 1971 paper described a deep network with 8 layers trained by the group method of data handling algorithm. [24]

Other deep learning working architectures, specifically those built for computer vision , began with the Neocognitron introduced by Kunihiko Fukushima in 1980. [43] In 1989, Yann LeCun et al. applied the standard backpropagation algorithm, which had been around as the reverse mode of automatic differentiation since 1970, [44] [45] [46] to a deep neural network with the purpose of recognizing handwritten ZIP codes on mail. While the algorithm worked, training required 3 days.

By 1991 such systems were used for recognizing isolated 2-D hand-written digits, while recognizing 3-D objects was done by matching 2-D images with a handcrafted 3-D object model. Weng et al. suggested that a human brain does not use a monolithic 3-D object model and in 1992 they published Cresceptron, [47] [48] [49] a method for performing 3-D object recognition in cluttered scenes. Cresceptron is a cascade of layers similar to Neocognitron. But while Neocognitron required a human programmer to hand-merge features, Cresceptron learned an open number of features in each layer without supervision, where each feature is represented by a convolution kernel . Cresceptron segmented each learned object from a cluttered scene through back-analysis through the network. Max pooling , now often adopted by deep neural networks (e.g. ImageNet tests), was first used in Cresceptron to reduce the position resolution by a factor of (2x2) to 1 through the cascade for better generalization.

In 1994, André C. P. L. F. de Carvalho, together with Fairhurst and Bisset, published experimental results of a multi-layer boolean neural network, also known as a weightless neural network, composed of a self-organising feature extraction neural network module followed by a classification neural network module, which were independently trained. [50]

In 1995, Brendan Frey demonstrated that it was possible to train (over two days) a network containing six fully connected layers and several hundred hidden units using the wake-sleep algorithm , co-developed with Peter Dayan and Hinton . [51] Many factors contribute to the slow speed, including the vanishing gradient problem analyzed in 1991 by Sepp Hochreiter . [52] [53]

Simpler models that use task-specific handcrafted features such as Gabor filters and support vector machines (SVMs) were a popular choice in the 1990s and 2000s, because of ANNs' computational cost and a lack of understanding of how the brain wires its biological networks.

Both shallow and deep learning (e.g., recurrent nets) of ANNs have been explored for many years. [14] [57] [14] These methods never outperformed non-uniform internal-handcrafting Gaussian mixture model / Hidden Markov model (GMM-HMM) technology based on generative models of speech trained discriminatively. [61] Key difficulties have been analyzed, including gradient diminishing [52] and weak temporal correlation structure in neural predictive models. [62] [63] Additional difficulties were the lack of training data and limited computing power.

Most speech recognition researchers moved away from neural nets to pursue generative modeling. An exception was at SRI International in the late 1990s. Funded by the US government's NSA and DARPA , SRI studied deep neural networks in speech and speaker recognition. Heck's speaker recognition team achieved the first significant success with deep neural networks in speech processing in the 1998 National Institute of Standards and Technology Speaker Recognition evaluation. [64] While SRI experienced success with deep neural networks in speaker recognition, they were unsuccessful in demonstrating similar success in speech recognition. One decade later, Hinton and Deng collaborated with each other and then with colleagues across groups at University of Toronto, Microsoft, Google and IBM, igniting a renaissance of deep feedforward neural networks in speech recognition. [65] [66] [67]

The principle of elevating "raw" features over hand-crafted optimization was first explored successfully in the architecture of deep autoencoder on the "raw" spectrogram or linear filter-bank features in the late 1990s, [64] showing its superiority over the Mel-Cepstral features that contain stages of fixed transformation from spectrograms. The raw features of speech, waveforms , later produced excellent larger-scale results. [15]

Many aspects of speech recognition were taken over by a deep learning method called Long short-term memory (LSTM), a recurrent neural network published by Hochreiter and Schmidhuber in 1997. [72] LSTM RNNs avoid the vanishing gradient problem and can learn "Very Deep Learning" tasks [9] that require memories of events that happened thousands of discrete time steps before, which is important for speech. In 2003, LSTM started to become competitive with traditional speech recognizers on certain tasks. [73] Later it was combined with connectionist temporal classification (CTC) [74] in stacks of LSTM RNNs. In 2015, Google's speech recognition reportedly experienced a dramatic performance jump of 49% through CTC-trained LSTM, which they made available through Google Voice Search . [75]

In the early 2000s, CNNs processed an estimated 10% to 20% of all the checks written in the US. [76]

In 2006, Hinton and Salakhutdinov showed how a many-layered feedforward neural network could be effectively pre-trained one layer at a time, treating each layer in turn as an unsupervised restricted Boltzmann machine, then fine-tuning it using supervised backpropagation. [15]

Deep learning is part of state-of-the-art systems in various disciplines, particularly computer vision and automatic speech recognition (ASR). Results on commonly used evaluation sets such as TIMIT (ASR) and MNIST ( image classification ), as well as a range of large-vocabulary speech recognition tasks have steadily improved. [65] [81] [83] Convolutional neural networks (CNNs) were superseded for ASR by CTC [74] for LSTM. [72] [75] [84] [85] [86] [88] but are more successful in computer vision.

The impact of deep learning in industry began in the early 2000s, when CNNs already processed an estimated 10% to 20% of all the checks written in the US. [76] Industrial applications of deep learning to large-scale speech recognition started around 2010.

In late 2009, Li Deng invited Hinton to work with him and colleagues to apply deep learning to speech recognition. They co-organized the 2009 NIPS Workshop on Deep Learning for Speech Recognition. The workshop was motivated by the limitations of deep generative models of speech, and the possibility that given more capable hardware and large-scale data sets that deep neural nets (DNN) might become practical. It was believed that pre-training DNNs using generative models of deep belief nets (DBN) would overcome the main difficulties of neural nets. However, they discovered that replacing pre-training with large amounts of training data for straightforward backpropagation when using DNNs with large, context-dependent output layers produced error rates dramatically lower than then-state-of-the-art Gaussian mixture model (GMM)/Hidden Markov Model (HMM) and also than more-advanced generative model-based systems. [65] The nature of the recognition errors produced by the two types of systems was found to be characteristically different, [66] offering technical insights into how to integrate deep learning into the existing highly efficient, run-time speech decoding system deployed by all major speech recognition systems. [20] [89] [90] Analysis around 2009-2010, contrasted the GMM (and other generative speech models) vs. DNN models, stimulated early industrial investment in deep learning for speech recognition, [66] eventually leading to pervasive and dominant use in that industry. That analysis was done with comparable performance (less than 1.5% in error rate) between discriminative DNNs and generative models.

In 2010, researchers extended deep learning from TIMIT to large vocabulary speech recognition, by adopting large output layers of the DNN based on context-dependent HMM states constructed by decision trees . [91] [92] [93] [89]

Advances in hardware enabled the renewed interest. In 2009, Nvidia was involved in what was called the “big bang” of deep learning, “as deep-learning neural networks were trained with Nvidia graphics processing units (GPUs).” [94] That year, Google Brain used Nvidia GPUs to create capable DNNs. While there, Ng determined that GPUs could increase the speed of deep-learning systems by about 100 times. [95] In particular, GPUs are well-suited for the matrix/vector math involved in machine learning. [96] GPUs speed up training algorithms by orders of magnitude, reducing running times from weeks to days. [98] [101] Specialized hardware and algorithm optimizations can be used for efficient processing. [102]

In 2012, a team led by Dahl won the "Merck Molecular Activity Challenge" using multi-task deep neural networks to predict the biomolecular target of one drug. [103] [104] In 2014, Hochreiter's group used deep learning to detect off-target and toxic effects of environmental chemicals in nutrients, household products and drugs and won the "Tox21 Data Challenge" of NIH , FDA and NCATS . [105] [107]

Significant additional impacts in image or object recognition were felt from 2011 to 2012. Although CNNs trained by backpropagation had been around for decades, and GPU implementations of NNs for years, including CNNs, fast implementations of CNNs with max-pooling on GPUs in the style of Ciresan and colleagues were needed to progress on computer vision. [96] [109] [9] In 2011, this approach achieved for the first time superhuman performance in a visual pattern recognition contest. Also in 2011, it won the ICDAR Chinese handwriting contest, and in May 2012, it won the ISBI image segmentation contest. [110] Until 2011, CNNs did not play a major role at computer vision conferences, but in June 2012, a paper by Ciresan et al. at the leading conference CVPR [17] showed how max-pooling CNNs on GPU can dramatically improve many vision benchmark records. In October 2012, a similar system by Krizhevsky and Hinton [18] won the large-scale ImageNet competition by a significant margin over shallow machine learning methods. In November 2012, Ciresan et al.'s system also won the ICPR contest on analysis of large medical images for cancer detection, and in the following year also the MICCAI Grand Challenge on the same topic. [111] In 2013 and 2014, the error rate on the ImageNet task using deep learning was further reduced, following a similar trend in large-scale speech recognition. The Wolfram Image Identification project publicized these improvements. [112]

Image classification was then extended to the more challenging task of generating descriptions (captions) for images, often as a combination of CNNs and LSTMs. [113] [114] [115] [117]

Artificial neural networks

Artificial neural networks ( ANNs ) or connectionist systems are computing systems inspired by the biological neural networks that constitute animal brains. Such systems learn (progressively improve their ability) to do tasks by considering examples, generally without task-specific programming. For example, in image recognition, they might learn to identify images that contain cats by analyzing example images that have been manually labeled as "cat" or "no cat" and using the analytic results to identify cats in other images. They have found most use in applications difficult to express with a traditional computer algorithm using rule-based programming .

An ANN is based on a collection of connected units called artificial neurons , (analogous to axons in a biological brain ). Each connection ( synapse ) between neurons can transmit a signal to another neuron. The receiving (postsynaptic) neuron can process the signal(s) and then signal downstream neurons connected to it. Neurons may have state, generally represented by real numbers , typically between 0 and 1. Neurons and synapses may also have a weight that varies as learning proceeds, which can increase or decrease the strength of the signal that it sends downstream.

Typically, neurons are organized in layers. Different layers may perform different kinds of transformations on their inputs. Signals travel from the first (input), to the last (output) layer, possibly after traversing the layers multiple times.

The original goal of the neural network approach was to solve problems in the same way that a human brain would. Over time, attention focused on matching specific mental abilities, leading to deviations from biology such as backpropagation, or passing information in the reverse direction and adjusting the network to reflect that information.

Neural networks have been used on a variety of tasks, including computer vision, speech recognition , machine translation , social network filtering, playing board and video games and medical diagnosis.

As of 2017, neural networks typically have a few thousand to a few million units and millions of connections. Despite this number being several order of magnitude less than the number of neurons on a human brain, these networks can perform many tasks at a level beyond that of humans (e.g., recognizing faces, playing "Go").

Deep neural networks

A deep neural network (DNN) is an ANN with multiple hidden layers between the input and output layers. [22] [9] Similar to shallow ANNs, DNNs can model complex non-linear relationships. DNN architectures generate compositional models where the object is expressed as a layered composition of primitives . [118] The extra layers enable composition of features from lower layers, potentially modeling complex data with fewer units than a similarly performing shallow network. [22]

Deep architectures include many variants of a few basic approaches. Each architecture has found success in specific domains. It is not always possible to compare the performance of multiple architectures, unless they have been evaluated on the same data sets.

DNNs are typically feedforward networks in which data flows from the input layer to the output layer without looping back.

Recurrent neural networks (RNNs), in which data can flow in any direction, are used for applications such as language modeling . [119] [120] [121] [122] [123] Long short-term memory is particularly effective for this use. [72] [124]

Convolutional deep neural networks (CNNs) are used in computer vision. [125] CNNs also have been applied to acoustic modeling for automatic speech recognition (ASR). [88]


As with ANNs, many issues can arise with naively trained DNNs. Two common issues are overfitting and computation time.

DNNs are prone to overfitting because of the added layers of abstraction, which allow them to model rare dependencies in the training data. Regularization methods such as Ivakhnenko's unit pruning [24] or weight decay ( -regularization) or sparsity ( -regularization) can be applied during training to combat overfitting. [127] Alternatively dropout regularization randomly omits units from the hidden layers during training. This helps to exclude rare dependencies. [128] Finally, data can be augmented via methods such as cropping and rotating such that smaller training sets can be increased in size to reduce the chances of overfitting. [129]

DNNs must consider many training parameters, such as the size (number of layers and number of units per layer), the learning rate and initial weights. Sweeping through the parameter space for optimal parameters may not be feasible due to the cost in time and computational resources. Various tricks such as batching (computing the gradient on several training examples at once rather than individual examples) [130] speed up computation. The large processing throughput of GPUs has produced significant speedups in training, because the matrix and vector computations required are well-suited for GPUs. [9]


Automatic speech recognition

Large-scale automatic speech recognition is the first and most convincing successful case of deep learning. LSTM RNNs can learn "Very Deep Learning" tasks [9] that involve multi-second intervals containing speech events separated by thousands of discrete time steps, where one time step corresponds to about 10 ms. LSTM with forget gates [124] is competitive with traditional speech recognizers on certain tasks. [73]

The initial success in speech recognition was based on small-scale recognition tasks based on TIMIT. The data set contains 630 speakers from eight major dialects of American English , where each speaker reads 10 sentences. Its small size allows many configurations to be tried. More importantly, the TIMIT task concerns phone-sequence recognition, which, unlike word-sequence recognition, allows weak language models (without a strong grammar). This allows the weaknesses in acoustic modeling aspects of speech recognition to be more easily analyzed. The error rates listed below, including these early results and measured as percent phone error rates (PER), have been summarized over the past 20 years:

Method PER (%)
Randomly Initialized RNN 26.1
Bayesian Triphone GMM-HMM 25.6
Hidden Trajectory (Generative) Model 24.8
Monophone Randomly Initialized DNN 23.4
Monophone DBN-DNN 22.4
Triphone GMM-HMM with BMMI Training 21.7
Monophone DBN-DNN on fbank 20.7
Convolutional DNN [131] 20.0
Convolutional DNN w. Heterogeneous Pooling 18.7
Ensemble DNN/CNN/RNN [132] 18.2
Bidirectional LSTM 17.9

The debut of DNNs for speaker recognition in the late 1990s and speech recognition around 2009-2011 and of LSTM around 2003-2007, accelerated progress in eight major areas: [20] [67] [89]

  • Scale-up/out and acclerated DNN training and decoding
  • Sequence discriminative training
  • Feature processing by deep models with solid understanding of the underlying mechanisms
  • Adaptation of DNNs and related deep models
  • Multi-task and transfer learning by DNNs and related deep models
  • CNNs and how to design them to best exploit domain knowledge of speech
  • RNN and its rich LSTM variants
  • Other types of deep models including tensor-based models and integrated deep generative/discriminative models.

All major commercial speech recognition systems (e.g., Microsoft Cortana , Xbox , Skype Translator , Amazon Alexa , Google Now , Apple Siri , Baidu and iFlyTek voice search, and a range of Nuance speech products, etc.) are based on deep learning. [20] [133] [134] [135]

Image recognition

A common evaluation set for image classification is the MNIST database data set. MNIST is composed of handwritten digits and includes 60,000 training examples and 10,000 test examples. As with TIMIT, its small size allows multiple configurations to be tested. A comprehensive list of results on this set is available. [136]

Deep learning-based image recognition has become "superhuman", producing more accurate results than human contestants. This first occurred in 2011. [138]

Deep learning-trained vehicles now interpret 360° camera views. [139] Another example is Facial Dysmorphology Novel Analysis (FDNA) used to analyze cases of human malformation connected to a large database of genetic syndromes.

Visual Art Processing

Closely related to the progress that has been made in image recognition is the increasing application of deep learning techniques to various visual art tasks. DNNs have proven themselves capable, for example, of a) identifying the style period of a given painting, b) "capturing" the style of a given painting and applying it in a visually pleasing manner to an arbitrary photograph, and c) generating striking imagery based on random visual input fields. [140] [141]

Natural language processing

Neural networks have been used for implementing language models since the early 2000s. [119] [143] LSTM helped to improve machine translation and language modeling. [120] [121] [122]

Other key techniques in this field are negative sampling [144] and word embedding . Word embedding, such as word2vec , can be thought of as a representational layer in a deep learning architecture that transforms an atomic word into a positional representation of the word relative to other words in the dataset; the position is represented as a point in a vector space . Using word embedding as an RNN input layer allows the network to parse sentences and phrases using an effective compositional vector grammar. A compositional vector grammar can be thought of as probabilistic context free grammar (PCFG) implemented by an RNN. [145] Recursive auto-encoders built atop word embeddings can assess sentence similarity and detect paraphrasing. [145] Deep neural architectures provide the best results for constituency parsing , [146] sentiment analysis , [147] information retrieval, [148] [149] spoken language understanding, [150] machine translation, [120] [151] contextual entity linking, [151] writing style recognition [152] , categorizing clinical narratives [154] , and others. [155]

Google Translate (GT) uses a large end-to-end long short-term memory network. [156] [157] [158] [160] [161] [162] GNMT uses an example-based machine translation method in which the system "learns from millions of examples." [157] It translates "whole sentences at a time, rather than pieces. Google Translate supports over one hundred languages. [157] The network encodes the "semantics of the sentence rather than simply memorizing phrase-to-phrase translations". [157] [163] GT can translate directly from one language to another, rather than using English as an intermediate. [163]

Drug discovery and toxicology

A large percentage of candidate drugs fail to win regulatory approval. These failures are caused by insufficient efficacy (on-target effect), undesired interactions (off-target effects), or unanticipated toxic effects . [165] [167] Research has explored use of deep learning to predict biomolecular target, [103] [104] off-target and toxic effects of environmental chemicals in nutrients, household products and drugs. [105] [107]

AtomNet is a deep learning system for structure-based rational drug design . [168] AtomNet was used to predict novel candidate biomolecules for disease targets such as the Ebola virus [169] and multiple sclerosis . [170]

Customer relationship management

Deep reinforcement learning has been used to approximate the value of possible direct marketing actions, defined in terms of RFM variables. The estimated value function was shown to have a natural interpretation as customer lifetime value . [171]

Recommendation systems

Recommendation systems have used deep learning to extract meaningful features for a latent factor model for content-based music recommendations. [172] Multiview deep learning has been applied for learning user preferences from multiple domains. [173] The model uses a hybrid collaborative and content-based approach and enhances recommendations in multiple tasks.


An autoencoder ANN was used in bioinformatics , to predict gene ontology annotations and gene-function relationships. [175]

In medical informatics, deep learning was used to predict sleep quality based on data from wearables [177] and predictions of health complications from electronic health record data. [181]

An extension of Word2Vec was used to create semantic labels for radiological images by exploiting the clinical narratives.

Mobile Advertising

Finding the appropriate mobile audience for mobile advertising [182] is always challenging since there are many data points that need to be considered and assimilated before a target segment can be created and used in ad serving by any ad server. Deep learning has been used to interpret large, many-dimensioned advertising datasets. Many data points are collected during the request/serve/click internet advertising cycle. This information can form the basis of machine learning to improve ad selection.

Relation to human development

Deep learning is closely related to a class of theories of brain development (specifically, neocortical development) proposed by cognitive neuroscientists in the early 1990s. [183] [184] [185] [186] These developmental theories were instantiated in computational models, making them predecessors of deep learning systems. These developmental models share the property that various proposed learning dynamics in the brain (e.g., a wave of nerve growth factor ) support the self-organization somewhat analogous to the neural networks utilized in deep learning models. Like the neocortex , neural networks employ a hierarchy of layered filters in which each layer considers information from a prior layer (or the operating environment), and then passes its output (and possibly the original input), to other layers. This process yields a self-organizing stack of transducers , well-tuned to their operating environment. A 1995 description stated, "...the infant's brain seems to organize itself under the influence of waves of so-called trophic-factors ... different regions of the brain become connected sequentially, with one layer of tissue maturing before another and so on until the whole brain is mature."

Commercial activity

Many organizations employ deep learning for particular applications. Facebook 's AI lab performs tasks such as automatically tagging uploaded pictures with the names of the people in them. [187]

Google's DeepMind Technologies developed a system capable of learning how to play Atari video games using only pixels as data input. In 2015 they demonstrated their AlphaGo system, which learned the game of Go well enough to beat a professional Go player. [188] [192] [193] Google Translate uses an LSTM to translate between more than 100 languages.

In 2015, Blippar demonstrated a mobile augmented reality application that uses deep learning to recognize objects in real time. [194]

Criticism and comment

Deep learning has attracted both criticism and comment, in some cases from outside the field of computer science.


A main criticism concerns the lack of theory surrounding the methods. Learning in the most common deep architectures is implemented using well-understood gradient descent. However, the theory surrounding other algorithms, such as contrastive divergence is less clear. (e.g., Does it converge? If so, how fast? What is it approximating?) Deep learning methods are often looked at as a black box , with most confirmations done empirically, rather than theoretically. [195]

Others point out that deep learning should be looked at as a step towards realizing strong AI, not as an all-encompassing solution. Despite the power of deep learning methods, they still lack much of the functionality needed for realizing this goal entirely. Research psychologist Gary Marcus noted:

"Realistically, deep learning is only part of the larger challenge of building intelligent machines. Such techniques lack ways of representing causal relationships (...) have no obvious ways of performing logical inferences , and they are also still a long way from integrating abstract knowledge, such as information about what objects are, what they are for, and how they are typically used. The most powerful A.I. systems, like Watson (...) use techniques like deep learning as just one element in a very complicated ensemble of techniques, ranging from the statistical technique of Bayesian inference to deductive reasoning ." [196]

As an alternative to this emphasis on the limits of deep learning, one author speculated that it might be possible to train a machine vision stack to perform the sophisticated task of discriminating between "old master" and amateur figure drawings, and hypothesized that such a sensitivity might represent the rudiments of a non-trivial machine empathy. [197] This same author proposed that this would be in line with anthropology, which identifies a concern with aesthetics as a key element of behavioral modernity . [198]

In further reference to the idea that artistic sensitivity might inhere within relatively low levels of the cognitive hierarchy, a published series of graphic representations of the internal states of deep (20-30 layers) neural networks attempting to discern within essentially random data the images on which they were trained [199] demonstrate a visual appeal: the original research notice received well over 1,000 comments, and was the subject of what was for a time the most frequently accessed article on The Guardian 's [200] web site.


Some deep learning architectures display problematic behaviors, [201] such as confidently classifying unrecognizable images as belonging to a familiar category of ordinary images [202] and misclassifying minuscule perturbations of correctly classified images. [203] Goertzel hypothesized that these behaviors are due to limitations in their internal representations and that these limitations would inhibit integration into heterogeneous multi-component AGI architectures. [201] These issues may possibly be addressed by deep learning architectures that internally form states homologous to image-grammar [204] decompositions of observed entities and events. [201] Learning a grammar (visual or linguistic) from training data would be equivalent to restricting the system to commonsense reasoning that operates on concepts in terms of grammatical production rules and is a basic goal of both human language acquisition and AI. [205]


As deep learning moves from the lab into the world, artificial neural networks have been shown to be vulnerable to hacks and deception. By identifying patterns that these systems use to function, attackers can modify inputs to ANNs in such a way that the ANN finds a match that human observers would not recognize. For example, an attacker can make subtle changes to an image such that the ANN finds a match even though the image looks to a human nothing like the search target. Such a manipulation is termed an “adversarial attack.” In 2016 researchers used one ANN to doctor images in trial and error fashion, identify another's focal points and thereby generate images that deceived it. The modified images looked no different to human eyes. Another group showed that printouts of doctored images then photographed successfully tricked an image classification system. [206] One defense is reverse image search, in which a possible fake image is submitted to a site such as TinEye that can then find other instances of it. A refinement is to search using only parts of the image, to identify images from which that piece may have been taken . [208]

Another group showed that certain psychedelic spectacles could fool a facial recognition system into thinking ordinary people were celebrities, potentially allowing one person to impersonate another. In 2017 researchers added stickers to stop signs and caused an ANN to misclassify them. [206]

ANNs can however be further trained to detect attempts at deception, potentially leading attackers and defenders into an arms race similar to the kind that already defines the malware defense industry. ANNs have been trained to defeat ANN-based anti-malware software by repeatedly attacking a defense with malware that was continually altered by a genetic algorithm until it tricked the anti-malware while retaining its ability to damage the target. [206]

Another group demonstrated that certain sounds could make the Google Now voice command system open a particular web address that would download malware. [206]

In “data poisoning”, false data is continually smuggled into a machine learning system’s training set to prevent it from achieving mastery. [206]

See also