Deepti Gurdasani is a British-Indian clinical epidemiologist and statistical geneticist who is a senior lecturer in machine learning at the Queen Mary University of London. Her research considers the genetic diversity of African Populations. Throughout the COVID-19 pandemic, Gurdasani has provided the public with her analysis of the evolving situation mainly on the Twitter platform. == Early life and education == Gurdasani was an undergraduate and medical student at the Christian Medical College Vellore at Tamil Nadu Dr. M.G.R. Medical University. After earning her medical degree and qualifying in internal medicine, she moved to the United Kingdom, where she worked toward a research doctorate in genetic epidemiology at Wolfson College, Cambridge. Her doctoral research involved the design of strategies to understand complex diseases in diverse populations. == Research and career == In 2013, Gurdasani joined the Wellcome Sanger Institute as a postdoctoral fellow, where she worked on the genomic diversity of African populations and how this diversity impacts susceptibility to disease. She makes use of dense genotypes and whole genome sequences to better understand how population movements determined genetic structure. In particular, Gurdasani develops machine learning algorithms to large-scale clinical data sets. At the Sanger Gurdasani co-led the African Genome Variation Project and the Uganda Resource Project. Gurdasani moved to Queen Mary University of London in 2019, where she created deep learning approaches for clinical prediction and the identification of novel, genome-based drug targets. During the COVID-19 pandemic Gurdasani has provided public commentary on the pandemic, making use of both Twitter and print media to share information on the evolving situation. She has researched the incidence of long covid in the UK. In 2021 Gurdasani started to write for The Guardian. == Selected publications == Deepti Gurdasani; Tommy Carstensen; Fasil Tekola-Ayele; et al. (3 December 2014). "The African Genome Variation Project shapes medical genetics in Africa". Nature. 517 (7534): 327–332. doi:10.1038/NATURE13997. ISSN 1476-4687. PMC 4297536. PMID 25470054. Wikidata Q34979569. Nisreen A Alwan; Rochelle Ann Burgess; Simon Ashworth; et al. (15 October 2020). "Scientific consensus on the COVID-19 pandemic: we need to act now". The Lancet. doi:10.1016/S0140-6736(20)32153-X. ISSN 0140-6736. PMC 7557300. PMID 33069277. Wikidata Q100697134. Deepti Gurdasani; Inês Barroso; Eleftheria Zeggini; Manjinder S Sandhu (24 June 2019). "Genomics of disease risk in globally diverse populations". Nature Reviews Genetics. 20 (9): 520–535. doi:10.1038/S41576-019-0144-0. ISSN 1471-0056. PMID 31235872. Wikidata Q93000887. (erratum)
Neural style transfer
Neural style transfer (NST) software algorithms are able to manipulate digital images, or videos, in order to adopt the appearance or visual style of another image. NST algorithms are characterized by their use of deep neural networks for the sake of image transformation. Common uses for NST are the creation of artificial artwork from photographs, for example by transferring the appearance of famous paintings to user-supplied photographs. Several notable mobile apps use NST techniques for this purpose, including DeepArt and Prisma. This method has been used by artists and designers around the globe to develop new artwork based on existent style(s). == History == NST is an example of image stylization, a problem studied for over two decades within the field of non-photorealistic rendering. The first two example-based style transfer algorithms were image analogies and image quilting. Both of these methods were based on patch-based texture synthesis algorithms. Given a training pair of images–a photo and an artwork depicting that photo–a transformation could be learned and then applied to create new artwork from a new photo, by analogy. If no training photo was available, it would need to be produced by processing the input artwork; image quilting did not require this processing step, though it was demonstrated on only one style. NST was first published in the paper "A Neural Algorithm of Artistic Style" by Leon Gatys et al., originally released to ArXiv 2015, and subsequently accepted by the peer-reviewed CVPR conference in 2016. The original paper used a VGG-19 architecture that has been pre-trained to perform object recognition using the ImageNet dataset. In 2017, Google AI introduced a method that allows a single deep convolutional style transfer network to learn multiple styles at the same time. This algorithm permits style interpolation in real-time, even when done on video media. == Mathematics == This section closely follows the original paper. === Overview === The idea of Neural Style Transfer (NST) is to take two images—a content image p → {\displaystyle {\vec {p}}} and a style image a → {\displaystyle {\vec {a}}} —and generate a third image x → {\displaystyle {\vec {x}}} that minimizes a weighted combination of two loss functions: a content loss L content ( p → , x → ) {\displaystyle {\mathcal {L}}_{\text{content }}({\vec {p}},{\vec {x}})} and a style loss L style ( a → , x → ) {\displaystyle {\mathcal {L}}_{\text{style }}({\vec {a}},{\vec {x}})} . The total loss is a linear sum of the two: L NST ( p → , a → , x → ) = α L content ( p → , x → ) + β L style ( a → , x → ) {\displaystyle {\mathcal {L}}_{\text{NST}}({\vec {p}},{\vec {a}},{\vec {x}})=\alpha {\mathcal {L}}_{\text{content}}({\vec {p}},{\vec {x}})+\beta {\mathcal {L}}_{\text{style}}({\vec {a}},{\vec {x}})} By jointly minimizing the content and style losses, NST generates an image that blends the content of the content image with the style of the style image. Both the content loss and the style loss measures the similarity of two images. The content similarity is the weighted sum of squared-differences between the neural activations of a single convolutional neural network (CNN) on two images. The style similarity is the weighted sum of Gram matrices within each layer (see below for details). The original paper used a VGG-19 CNN, but the method works for any CNN. === Symbols === Let x → {\textstyle {\vec {x}}} be an image input to a CNN. Let F l ∈ R N l × M l {\textstyle F^{l}\in \mathbb {R} ^{N_{l}\times M_{l}}} be the matrix of filter responses in layer l {\textstyle l} to the image x → {\textstyle {\vec {x}}} , where: N l {\textstyle N_{l}} is the number of filters in layer l {\textstyle l} ; M l {\textstyle M_{l}} is the height times the width (i.e. number of pixels) of each filter in layer l {\textstyle l} ; F i j l ( x → ) {\textstyle F_{ij}^{l}({\vec {x}})} is the activation of the i th {\textstyle i^{\text{th}}} filter at position j {\textstyle j} in layer l {\textstyle l} . A given input image x → {\textstyle {\vec {x}}} is encoded in each layer of the CNN by the filter responses to that image, with higher layers encoding more global features, but losing details on local features. === Content loss === Let p → {\textstyle {\vec {p}}} be an original image. Let x → {\textstyle {\vec {x}}} be an image that is generated to match the content of p → {\textstyle {\vec {p}}} . Let P l {\textstyle P^{l}} be the matrix of filter responses in layer l {\textstyle l} to the image p → {\textstyle {\vec {p}}} . The content loss is defined as the squared-error loss between the feature representations of the generated image and the content image at a chosen layer l {\displaystyle l} of a CNN: L content ( p → , x → , l ) = 1 2 ∑ i , j ( A i j l ( x → ) − A i j l ( p → ) ) 2 {\displaystyle {\mathcal {L}}_{\text{content }}({\vec {p}},{\vec {x}},l)={\frac {1}{2}}\sum _{i,j}\left(A_{ij}^{l}({\vec {x}})-A_{ij}^{l}({\vec {p}})\right)^{2}} where A i j l ( x → ) {\displaystyle A_{ij}^{l}({\vec {x}})} and A i j l ( p → ) {\displaystyle A_{ij}^{l}({\vec {p}})} are the activations of the i th {\displaystyle i^{\text{th}}} filter at position j {\displaystyle j} in layer l {\displaystyle l} for the generated and content images, respectively. Minimizing this loss encourages the generated image to have similar content to the content image, as captured by the feature activations in the chosen layer. The total content loss is a linear sum of the content losses of each layer: L content ( p → , x → ) = ∑ l v l L content ( p → , x → , l ) {\displaystyle {\mathcal {L}}_{\text{content }}({\vec {p}},{\vec {x}})=\sum _{l}v_{l}{\mathcal {L}}_{\text{content }}({\vec {p}},{\vec {x}},l)} , where the v l {\displaystyle v_{l}} are positive real numbers chosen as hyperparameters. === Style loss === The style loss is based on the Gram matrices of the generated and style images, which capture the correlations between different filter responses at different layers of the CNN: L style ( a → , x → ) = ∑ l = 0 L w l E l , {\displaystyle {\mathcal {L}}_{\text{style }}({\vec {a}},{\vec {x}})=\sum _{l=0}^{L}w_{l}E_{l},} where E l = 1 4 N l 2 M l 2 ∑ i , j ( G i j l ( x → ) − G i j l ( a → ) ) 2 . {\displaystyle E_{l}={\frac {1}{4N_{l}^{2}M_{l}^{2}}}\sum _{i,j}\left(G_{ij}^{l}({\vec {x}})-G_{ij}^{l}({\vec {a}})\right)^{2}.} Here, G i j l ( x → ) {\displaystyle G_{ij}^{l}({\vec {x}})} and G i j l ( a → ) {\displaystyle G_{ij}^{l}({\vec {a}})} are the entries of the Gram matrices for the generated and style images at layer l {\displaystyle l} . Explicitly, G i j l ( x → ) = ∑ k F i k l ( x → ) F j k l ( x → ) {\displaystyle G_{ij}^{l}({\vec {x}})=\sum _{k}F_{ik}^{l}({\vec {x}})F_{jk}^{l}({\vec {x}})} Minimizing this loss encourages the generated image to have similar style characteristics to the style image, as captured by the correlations between feature responses in each layer. The idea is that activation pattern correlations between filters in a single layer captures the "style" on the order of the receptive fields at that layer. Similarly to the previous case, the w l {\displaystyle w_{l}} are positive real numbers chosen as hyperparameters. === Hyperparameters === In the original paper, they used a particular choice of hyperparameters. The style loss is computed by w l = 0.2 {\displaystyle w_{l}=0.2} for the outputs of layers conv1_1, conv2_1, conv3_1, conv4_1, conv5_1 in the VGG-19 network, and zero otherwise. The content loss is computed by w l = 1 {\displaystyle w_{l}=1} for conv4_2, and zero otherwise. The ratio α / β ∈ [ 5 , 50 ] × 10 − 4 {\displaystyle \alpha /\beta \in [5,50]\times 10^{-4}} . === Training === Image x → {\displaystyle {\vec {x}}} is initially approximated by adding a small amount of white noise to input image p → {\displaystyle {\vec {p}}} and feeding it through the CNN. Then we successively backpropagate this loss through the network with the CNN weights fixed in order to update the pixels of x → {\displaystyle {\vec {x}}} . After several thousand epochs of training, an x → {\displaystyle {\vec {x}}} (hopefully) emerges that matches the style of a → {\displaystyle {\vec {a}}} and the content of p → {\displaystyle {\vec {p}}} . As of 2017, when implemented on a GPU, it takes a few minutes to converge. == Extensions == In some practical implementations, it is noted that the resulting image has too much high-frequency artifact, which can be suppressed by adding the total variation to the total loss. Compared to VGGNet, AlexNet does not work well for neural style transfer. NST has also been extended to videos. Subsequent work improved the speed of NST for images by using special-purpose normalizations. In a paper by Fei-Fei Li et al. adopted a different regularized loss metric and accelerated method for training to produce results in real-time (three orders of magnitude faster than Gatys). Their idea was to use not the pixel-based loss defined above but rather a 'perceptual loss' measuring t
YouNoodle
YouNoodle, Inc. is a San Francisco-based company, with offices in Barcelona and Santiago, founded in 2010, building a platform for entrepreneurship competitions all over the world. YouNoodle matches entrepreneurs with competitions, accelerators, and startup programs, and provides a judging and voting SaaS platform to university, non-profit, government and enterprise clients organizing innovation challenges and competitions. Stanford's BASES, UC Berkeley LAUNCH, Start-Up Chile, Amazon Startup Challenge, and NASA are all running one or more competitions on YouNoodle's platform. == History and structure == YouNoodle was founded by Rebeca Hwang and Torsten Kolind in 2010. The company was spun off a project started by Bob Goodson (Quid) and Kirill Makharinsky (Enki) in 2007 with support from Peter Thiel (Founders Fund), Max Levchin (PayPal) and Charles Lho (Amicus Group), founding investor and Chairman of YouNoodle today. This project also spawned Quid (Goodson) and indirectly Ostrovok (Makharinsky). Although also named YouNoodle, this project/company was discontinued in 2010, when the three new entities started operations. The founders of the 2007-2010 entity were Goodson and Makharinsky, both former students of the University of Oxford. Goodson had studied medieval English literature before moving from Oxford to California when Levchin, the co-founder of PayPal, invited him to join a start-up there. Makharinsky's degree was in applied mathematics, and he was also encouraged to pursue opportunities in the United States by Levchin. Other significant employees included Hwang (co-founder of today's YouNoodle), a Stanford University doctoral student whose research is into social network theory. == Startup predictor == YouNoodle's now discontinued "Startup predictor", part of the 2007-2010 entity and developed by Makharinsky and Hwang, used mathematical models to predict the success of new businesses. The user fills in a questionnaire, which takes about half an hour to complete and concentrates on the business concept, finances, founders and advisers. Because the procedure was designed for new companies, questions on revenue and traffic are not included. The site then provided an estimate of what the company's value will be after three years and a score from 1 to 1000 representing its value as an investment. The service was free for the startups themselves, but YouNoodle intended to charge third parties for access to the results. The level of detail required by the questionnaire makes it difficult for people without inside knowledge of a company to provide the data for a prediction on their own. The company's founders have declined to explain the algorithm in detail, but state that it takes into account the entrepreneurs' experience, networks and mutual relations. Information provided by companies which use the site's networking features is used to improve the algorithm. As of August 2008, the algorithm was based on data from 3,000 startups. In the same month the company had four patents pending on the technology.
Neural architecture search
Neural architecture search (NAS) is a technique for automating the design of artificial neural networks (ANN), a widely used model in the field of machine learning. NAS has been used to design networks that are on par with or outperform hand-designed architectures. Methods for NAS can be categorized according to the search space, search strategy and performance estimation strategy used: The search space defines the type(s) of ANN that can be designed and optimized. The search strategy defines the approach used to explore the search space. The performance estimation strategy evaluates the performance of a possible ANN from its design (without constructing and training it). NAS is closely related to hyperparameter optimization and meta-learning and is a subfield of automated machine learning (AutoML). == Reinforcement learning == Reinforcement learning (RL) can underpin a NAS search strategy. Barret Zoph and Quoc Viet Le applied NAS with RL targeting the CIFAR-10 dataset and achieved a network architecture that rivals the best manually-designed architecture for accuracy, with an error rate of 3.65, 0.09 percent better and 1.05x faster than a related hand-designed model. On the Penn Treebank dataset, that model composed a recurrent cell that outperforms LSTM, reaching a test set perplexity of 62.4, or 3.6 perplexity better than the prior leading system. On the PTB character language modeling task it achieved bits per character of 1.214. Learning a model architecture directly on a large dataset can be a lengthy process. NASNet addressed this issue by transferring a building block designed for a small dataset to a larger dataset. The design was constrained to use two types of convolutional cells to return feature maps that serve two main functions when convoluting an input feature map: normal cells that return maps of the same extent (height and width) and reduction cells in which the returned feature map height and width is reduced by a factor of two. For the reduction cell, the initial operation applied to the cell's inputs uses a stride of two (to reduce the height and width). The learned aspect of the design included elements such as which lower layer(s) each higher layer took as input, the transformations applied at that layer and to merge multiple outputs at each layer. In the studied example, the best convolutional layer (or "cell") was designed for the CIFAR-10 dataset and then applied to the ImageNet dataset by stacking copies of this cell, each with its own parameters. The approach yielded accuracy of 82.7% top-1 and 96.2% top-5. This exceeded the best human-invented architectures at a cost of 9 billion fewer FLOPS—a reduction of 28%. The system continued to exceed the manually-designed alternative at varying computation levels. The image features learned from image classification can be transferred to other computer vision problems. E.g., for object detection, the learned cells integrated with the Faster-RCNN framework improved performance by 4.0% on the COCO dataset. In the so-called Efficient Neural Architecture Search (ENAS), a controller discovers architectures by learning to search for an optimal subgraph within a large graph. The controller is trained with policy gradient to select a subgraph that maximizes the validation set's expected reward. The model corresponding to the subgraph is trained to minimize a canonical cross entropy loss. Multiple child models share parameters, ENAS requires fewer GPU-hours than other approaches and 1000-fold less than "standard" NAS. On CIFAR-10, the ENAS design achieved a test error of 2.89%, comparable to NASNet. On Penn Treebank, the ENAS design reached test perplexity of 55.8. == Evolution == An alternative approach to NAS is based on evolutionary algorithms, which has been employed by several groups. An Evolutionary Algorithm for Neural Architecture Search generally performs the following procedure. First a pool consisting of different candidate architectures along with their validation scores (fitness) is initialised. At each step the architectures in the candidate pool are mutated (e.g.: 3x3 convolution instead of a 5x5 convolution). Next the new architectures are trained from scratch for a few epochs and their validation scores are obtained. This is followed by replacing the lowest scoring architectures in the candidate pool with the better, newer architectures. This procedure is repeated multiple times and thus the candidate pool is refined over time. Mutations in the context of evolving ANNs are operations such as adding or removing a layer, which include changing the type of a layer (e.g., from convolution to pooling), changing the hyperparameters of a layer, or changing the training hyperparameters. On CIFAR-10 and ImageNet, evolution and RL performed comparably, while both slightly outperformed random search. == Bayesian optimization == Bayesian Optimization (BO), which has proven to be an efficient method for hyperparameter optimization, can also be applied to NAS. In this context, the objective function maps an architecture to its validation error after being trained for a number of epochs. At each iteration, BO uses a surrogate to model this objective function based on previously obtained architectures and their validation errors. One then chooses the next architecture to evaluate by maximizing an acquisition function, such as expected improvement, which provides a balance between exploration and exploitation. Acquisition function maximization and objective function evaluation are often computationally expensive for NAS, and make the application of BO challenging in this context. Recently, BANANAS has achieved promising results in this direction by introducing a high-performing instantiation of BO coupled to a neural predictor. == Hill-climbing == Another group used a hill climbing procedure that applies network morphisms, followed by short cosine-annealing optimization runs. The approach yielded competitive results, requiring resources on the same order of magnitude as training a single network. E.g., on CIFAR-10, the method designed and trained a network with an error rate below 5% in 12 hours on a single GPU. == Multi-objective search == While most approaches solely focus on finding architecture with maximal predictive performance, for most practical applications other objectives are relevant, such as memory consumption, model size or inference time (i.e., the time required to obtain a prediction). Because of that, researchers created a multi-objective search. LEMONADE is an evolutionary algorithm that adopted Lamarckism to efficiently optimize multiple objectives. In every generation, child networks are generated to improve the Pareto frontier with respect to the current population of ANNs. Neural Architect is claimed to be a resource-aware multi-objective RL-based NAS with network embedding and performance prediction. Network embedding encodes an existing network to a trainable embedding vector. Based on the embedding, a controller network generates transformations of the target network. A multi-objective reward function considers network accuracy, computational resource and training time. The reward is predicted by multiple performance simulation networks that are pre-trained or co-trained with the controller network. The controller network is trained via policy gradient. Following a modification, the resulting candidate network is evaluated by both an accuracy network and a training time network. The results are combined by a reward engine that passes its output back to the controller network. == One-shot models == RL or evolution-based NAS require thousands of GPU-days of searching/training to achieve state-of-the-art computer vision results as described in the NASNet, mNASNet and MobileNetV3 papers. To reduce computational cost, many recent NAS methods rely on the weight-sharing idea. In this approach, a single overparameterized supernetwork (also known as the one-shot model) is defined. A supernetwork is a very large Directed Acyclic Graph (DAG) whose subgraphs are different candidate neural networks. Thus, in a supernetwork, the weights are shared among a large number of different sub-architectures that have edges in common, each of which is considered as a path within the supernet. The essential idea is to train one supernetwork that spans many options for the final design rather than generating and training thousands of networks independently. In addition to the learned parameters, a set of architecture parameters are learnt to depict preference for one module over another. Such methods reduce the required computational resources to only a few GPU days. More recent works further combine this weight-sharing paradigm, with a continuous relaxation of the search space, which enables the use of gradient-based optimization methods. These approaches are generally referred to as differentiable NAS and have proven very efficient in exploring the search space of ne
Noise-based logic
Noise-based logic (NBL) is a class of multivalued deterministic logic schemes, developed in the twenty-first century, where the logic values and bits are represented by different realizations of a stochastic process. The concept of noise-based logic and its name was created by Laszlo B. Kish. In its foundation paper it is noted that the idea was inspired by the stochasticity of brain signals and by the unconventional noise-based communication schemes, such as the Kish cypher. == The noise-based logic space and hyperspace == The logic values are represented by multi-dimensional "vectors" (orthogonal functions) and their superposition, where the orthogonal basis vectors are independent noises. By the proper combination (products or set-theoretical products) of basis-noises, which are called noise-bit, a logic hyperspace can be constructed with D(N) = 2N number of dimensions, where N is the number of noise-bits. Thus N noise-bits in a single wire correspond to a system of 2N classical bits that can express 22N different logic values. Independent realizations of a stochastic process of zero mean have zero cross-correlation with each other and with other stochastic processes of zero mean. Thus the basis noise vectors are orthogonal not only to each other but they and all the noise-based logic states (superpositions) are orthogonal also to any background noises in the hardware. Therefore, the noise-based logic concept is robust against background noises, which is a property that can potentially offer a high energy-efficiency. == The types of signals used in noise-based logic == In the paper, where noise-based logic was first introduced, generic stochastic-processes with zero mean were proposed and a system of orthogonal sinusoidal signals were also proposed as a deterministic-signal version of the logic system. The mathematical analysis about statistical errors and signal energy was limited to the cases of Gaussian noises and superpositions as logic signals in the basic logic space and their products and superpositions of their products in the logic hyperspace (see also. In the subsequent brain logic scheme, the logic signals were (similarly to neural signals) unipolar spike sequences generated by a Poisson process, and set-theoretical unifications (superpositions) and intersections (products) of different spike sequences. Later, in the instantaneous noise-based logic schemes and computation works, random telegraph waves (periodic time, bipolar, with fixed absolute value of amplitude) were also utilized as one of the simplest stochastic processes available for NBL. With choosing unit amplitude and symmetric probabilities, the resulting random-telegraph wave has 0.5 probability to be in the +1 or in the −1 state which is held over the whole clock period. == The noise-based logic gates == Noise-based logic gates can be classified according to the method the input identifies the logic value at the input. The first gates analyzed the statistical correlations between the input signal and the reference noises. The advantage of these is the robustness against background noise. The disadvantage is the slow speed and higher hardware complexity. The instantaneous logic gates are fast, they have low complexity but they are not robust against background noises. With either neural spike type signals or with bipolar random-telegraph waves of unity absolute amplitude, and randomness only in the sign of the amplitude offer very simple instantaneous logic gates. Then linear or analog devices unnecessary and the scheme can operate in the digital domain. However, whenever instantaneous logic must be interfaced with classical logic schemes, the interface must use correlator-based logic gates for an error-free signal. == Universality of noise-based logic == All the noise-based logic schemes listed above have been proven universal. The papers typically produce the NOT and the AND gates to prove universality, because having both of them is a satisfactory condition for the universality of a Boolean logic. == Computation by noise-based logic == The string verification work over a slow communication channel shows a powerful computing application where the methods is inherently based on calculating the hash function. The scheme is based on random telegraph waves and it is mentioned in the paper that the authors intuitively conclude that the intelligence of the brain is using similar operations to make a reasonably good decision based on a limited amount of information. The superposition of the first D(N) = 2N integer numbers can be produced with only 2N operations, which the authors call "Achilles ankle operation" in the paper. == Computer chip realization of noise-based logic == Preliminary schemes have already been published to utilize noise-based logic in practical computers. However, it is obvious from these papers that this young field has yet a long way to go before it will be seen in everyday applications.
Frameserver
A frameserver is any program that acts as a media source in the process called frameserving, which transfers digital video data from one computer program to another without intermediate files. The program that receives the data – the frameclient – could be any type of video application. The process is controlled by the frameclient: the frameclient requests audio/video frames and the frameserver serves them. The client can request frames in any order, allowing it to pause or jump to an arbitrary frame, just as a media player does with a file on disk. The client is most commonly a media encoder, a non-linear editing system, or a media player. == Frameservers == AviSynth VirtualDub VapourSynth Debugmode FrameServer
Use of artificial intelligence by the United States Department of Defense
The United States Department of Defense has been analyzing and employing military applications of artificial intelligence since at least 2014. The program initially focused on drones and other robots, but has also been using large language models for military research and analysis. The current US policy on lethal autonomous weapons is Department of Defense Directive 3000.09, updated in January 2023. == Background == The United States Department of Defense began developing lethal autonomous weapons as early as the Reagan administration. An early version of the Tomahawk missile could have been used to destroy Soviet ships without direct human control; the initiative was abandoned after the United States and the Soviet Union signed START I. By 2014, the United Kingdom, Israel, and Norway had already begun using missiles equipped with artificial intelligence systems. The Department of Defense established a policy on the use of artificial intelligence in 2012. == History == === 2016–2017: Carter secretaryship === In May 2016, secretary of defense Ash Carter stated that his Third Offset strategy would include utilizing artificial intelligence as a military advantage. The New York Times reported that year that the Department of Defense had tested an autonomous drone at an approximation of a Middle Eastern village at Camp Edwards. Deputy secretary of defense Robert O. Work, who advocated for developing artificial intelligence, told the Times that the United States needed to compete with China and Russia by having a tactical advantage they could not easily replicate. The initiative was developed by DARPA beginning in 2015. The use of artificial intelligence in the U.S. military was controversial within the department; in February, Paul Scharre, who worked for the Office of the Secretary of Defense in the secretaryships of Robert Gates and Leon Panetta, published a report about the risks of artificial intelligence for broad military applications. === 2017–2019: Mattis secretaryship === By 2017, the United States Air Force had already begun using artificial intelligence in military robots. The Air Force's use of Neurala, an artificial intelligence company, concerned officials in the Department of Defense after an investigation found that Neurala had accepted money from an investment firm with funding from a state-run Chinese company. The Department of Defense began heavily investing in artificial intelligence after Work established Project Maven, an initiative to encourage the development and integration of artificial intelligence in the military, in April 2017. In May 2018, secretary of defense Jim Mattis privately expressed to president Donald Trump that he needed to establish a national strategy on artificial intelligence, quoting an article from former secretary of state Henry Kissinger that called for a presidential commission on the technology. The Department of Defense established the Joint Artificial Intelligence Center the following month. Google began working with the Department of Defense on analyzing drone footage as early as March. Google's involvement in the initiative led to protests from employees and mass resignations. Seeking to quell internal unrest, Google stated it would not renew its contract with the Department of Defense in June. The Department of Defense announced an artificial intelligence contract with Microsoft in October. === 2025–present: Hegseth secretaryship === In December 2025, secretary of defense Pete Hegseth announced GenAI.mil, an artificial intelligence platform for the Department of Defense. In a video announcing the platform, Hegseth stated that Department of Defense workers would be able to "conduct deep research, format documents and even analyze video or imagery." The Department of Defense contracted first Gemini by Google, then ChatGPT by OpenAI, and finally Grok by xAI for the platform. Claude by Anthropic was also contracted by the Department of Defense and was in use on secure servers until it was revealed that Claude had been used in the 2026 operation to capture Nicolás Maduro, who was at the time the leader of Venezuela. This revelation sparked a high-profile dispute over Anthropic's ability to constrain Claude's useage, resulting in the termination of Anthropic's $200 million defense contract. The Department of Defense also moved to label Anthropic a supply chain risk, which was later blocked by a federal judge.