Contrastive Language-Image Pre-training (CLIP) is a technique for training a pair of neural network models, one for image understanding and one for text understanding, using a contrastive objective. This method has enabled broad applications across multiple domains, including cross-modal retrieval, text-to-image generation, and aesthetic ranking. == Algorithm == The CLIP method trains a pair of models contrastively. One model takes in a piece of text as input and outputs a single vector representing its semantic content. The other model takes in an image and similarly outputs a single vector representing its visual content. The models are trained so that the vectors corresponding to semantically similar text-image pairs are close together in the shared vector space, while those corresponding to dissimilar pairs are far apart. To train a pair of CLIP models, one would start by preparing a large dataset of image-caption pairs. During training, the models are presented with batches of N {\displaystyle N} image-caption pairs. Let the outputs from the text and image models be respectively v 1 , . . . , v N , w 1 , . . . , w N {\displaystyle v_{1},...,v_{N},w_{1},...,w_{N}} . Two vectors are considered "similar" if their dot product is large. The loss incurred on this batch is the multi-class N-pair loss, which is a symmetric cross-entropy loss over similarity scores: − 1 N ∑ i ln e v i ⋅ w i / T ∑ j e v i ⋅ w j / T − 1 N ∑ j ln e v j ⋅ w j / T ∑ i e v i ⋅ w j / T {\displaystyle -{\frac {1}{N}}\sum _{i}\ln {\frac {e^{v_{i}\cdot w_{i}/T}}{\sum _{j}e^{v_{i}\cdot w_{j}/T}}}-{\frac {1}{N}}\sum _{j}\ln {\frac {e^{v_{j}\cdot w_{j}/T}}{\sum _{i}e^{v_{i}\cdot w_{j}/T}}}} In essence, this loss function encourages the dot product between matching image and text vectors ( v i ⋅ w i {\displaystyle v_{i}\cdot w_{i}} ) to be high, while discouraging high dot products between non-matching pairs. The parameter T > 0 {\displaystyle T>0} is the temperature, which is parameterized in the original CLIP model as T = e − τ {\displaystyle T=e^{-\tau }} where τ ∈ R {\displaystyle \tau \in \mathbb {R} } is a learned parameter. Other loss functions are possible. For example, Sigmoid CLIP (SigLIP) proposes the following loss function: L = 1 N ∑ i , j ∈ 1 : N f ( ( 2 δ i , j − 1 ) ( e τ w i ⋅ v j + b ) ) {\displaystyle L={\frac {1}{N}}\sum _{i,j\in 1:N}f((2\delta _{i,j}-1)(e^{\tau }w_{i}\cdot v_{j}+b))} where f ( x ) = ln ( 1 + e − x ) {\displaystyle f(x)=\ln(1+e^{-x})} is the negative log sigmoid loss, and the Dirac delta symbol δ i , j {\displaystyle \delta _{i,j}} is 1 if i = j {\displaystyle i=j} else 0. == CLIP models == While the original model was developed by OpenAI, subsequent models have been trained by other organizations as well. === Image model === The image encoding models used in CLIP are typically vision transformers (ViT). The naming convention for these models often reflects the specific ViT architecture used. For instance, "ViT-L/14" means a "vision transformer large" (compared to other models in the same series) with a patch size of 14, meaning that the image is divided into 14-by-14 pixel patches before being processed by the transformer. The size indicator ranges from B, L, H, G (base, large, huge, giant), in that order. Other than ViT, the image model is typically a convolutional neural network, such as ResNet (in the original series by OpenAI), or ConvNeXt (in the OpenCLIP model series by LAION). Since the output vectors of the image model and the text model must have exactly the same length, both the image model and the text model have fixed-length vector outputs, which in the original report is called "embedding dimension". For example, in the original OpenAI model, the ResNet models have embedding dimensions ranging from 512 to 1024, and for the ViTs, from 512 to 768. Its implementation of ViT was the same as the original one, with one modification: after position embeddings are added to the initial patch embeddings, there is a LayerNorm. Its implementation of ResNet was the same as the original one, with 3 modifications: In the start of the CNN (the "stem"), they used three stacked 3x3 convolutions instead of a single 7x7 convolution, as suggested by. There is an average pooling of stride 2 at the start of each downsampling convolutional layer (they called it rect-2 blur pooling according to the terminology of ). This has the effect of blurring images before downsampling, for antialiasing. The final convolutional layer is followed by a multiheaded attention pooling. ALIGN a model with similar capabilities, trained by researchers from Google used EfficientNet, a kind of convolutional neural network. === Text model === The text encoding models used in CLIP are typically Transformers. In the original OpenAI report, they reported using a Transformer (63M-parameter, 12-layer, 512-wide, 8 attention heads) with lower-cased byte pair encoding (BPE) with 49152 vocabulary size. Context length was capped at 76 for efficiency. Like GPT, it was decoder-only, with only causally-masked self-attention. Its architecture is the same as GPT-2. Like BERT, the text sequence is bracketed by two special tokens [SOS] and [EOS] ("start of sequence" and "end of sequence"). Take the activations of the highest layer of the transformer on the [EOS], apply LayerNorm, then a final linear map. This is the text encoding of the input sequence. The final linear map has output dimension equal to the embedding dimension of whatever image encoder it is paired with. These models all had context length 77 and vocabulary size 49408. ALIGN used BERT of various sizes. == Dataset == === WebImageText === The CLIP models released by OpenAI were trained on a dataset called "WebImageText" (WIT) containing 400 million pairs of images and their corresponding captions scraped from the internet. The total number of words in this dataset is similar in scale to the WebText dataset used for training GPT-2, which contains about 40 gigabytes of text data. The dataset contains 500,000 text-queries, with up to 20,000 (image, text) pairs per query. The text-queries were generated by starting with all words occurring at least 100 times in English Wikipedia, then extended by bigrams with high mutual information, names of all Wikipedia articles above a certain search volume, and WordNet synsets. The dataset is private and has not been released to the public, and there is no further information on it. ==== Data preprocessing ==== For the CLIP image models, the input images are preprocessed by first dividing each of the R, G, B values of an image by the maximum possible value, so that these values fall between 0 and 1, then subtracting by [0.48145466, 0.4578275, 0.40821073], and dividing by [0.26862954, 0.26130258, 0.27577711]. The rationale was that these are the mean and standard deviations of the images in the WebImageText dataset, so this preprocessing step roughly whitens the image tensor. These numbers slightly differ from the standard preprocessing for ImageNet, which uses [0.485, 0.456, 0.406] and [0.229, 0.224, 0.225]. If the input image does not have the same resolution as the native resolution (224×224 for all except ViT-L/14@336px, which has 336×336 resolution), then the input image is first scaled by bicubic interpolation, so that its shorter side is the same as the native resolution, then the central square of the image is cropped out. === Others === ALIGN used over one billion image-text pairs, obtained by extracting images and their alt-tags from online crawling. The method was described as similar to how the Conceptual Captions dataset was constructed, but instead of complex filtering, they only applied a frequency-based filtering. Later models trained by other organizations had published datasets. For example, LAION trained OpenCLIP with published datasets LAION-400M, LAION-2B, and DataComp-1B. == Training == In the original OpenAI CLIP report, they reported training 5 ResNet and 3 ViT (ViT-B/32, ViT-B/16, ViT-L/14). Each was trained for 32 epochs. The largest ResNet model took 18 days to train on 592 V100 GPUs. The largest ViT model took 12 days on 256 V100 GPUs. All ViT models were trained on 224×224 image resolution. The ViT-L/14 was then boosted to 336×336 resolution by FixRes, resulting in a model. They found this was the best-performing model. In the OpenCLIP series, the ViT-L/14 model was trained on 384 A100 GPUs on the LAION-2B dataset, for 160 epochs for a total of 32B samples seen. == Applications == === Cross-modal retrieval === CLIP's cross-modal retrieval enables the alignment of visual and textual data in a shared latent space, allowing users to retrieve images based on text descriptions and vice versa, without the need for explicit image annotations. In text-to-image retrieval, users input descriptive text, and CLIP retrieves images with matching embeddings. In image-to-text retrieval, images are used to find related text content. CLIP’s ability to connect vis
Kernel embedding of distributions
In machine learning, the kernel embedding of distributions (also called the kernel mean or mean map) comprises a class of nonparametric methods in which a probability distribution is represented as an element of a reproducing kernel Hilbert space (RKHS). A generalization of the individual data-point feature mapping done in classical kernel methods, the embedding of distributions into infinite-dimensional feature spaces can preserve all of the statistical features of arbitrary distributions, while allowing one to compare and manipulate distributions using Hilbert space operations such as inner products, distances, projections, linear transformations, and spectral analysis. This learning framework is very general and can be applied to distributions over any space Ω {\displaystyle \Omega } on which a sensible kernel function (measuring similarity between elements of Ω {\displaystyle \Omega } ) may be defined. For example, various kernels have been proposed for learning from data which are: vectors in R d {\displaystyle \mathbb {R} ^{d}} , discrete classes/categories, strings, graphs/networks, images, time series, manifolds, dynamical systems, and other structured objects. The theory behind kernel embeddings of distributions has been primarily developed by Alex Smola, Le Song, Arthur Gretton, and Bernhard Schölkopf. A review of recent works on kernel embedding of distributions can be found in. The analysis of distributions is fundamental in machine learning and statistics, and many algorithms in these fields rely on information theoretic approaches such as entropy, mutual information, or Kullback–Leibler divergence. However, to estimate these quantities, one must first either perform density estimation, or employ sophisticated space-partitioning/bias-correction strategies which are typically infeasible for high-dimensional data. Commonly, methods for modeling complex distributions rely on parametric assumptions that may be unfounded or computationally challenging (e.g. Gaussian mixture models), while nonparametric methods like kernel density estimation (Note: the smoothing kernels in this context have a different interpretation than the kernels discussed here) or characteristic function representation (via the Fourier transform of the distribution) break down in high-dimensional settings. Methods based on the kernel embedding of distributions sidestep these problems and also possess the following advantages: Data may be modeled without restrictive assumptions about the form of the distributions and relationships between variables Intermediate density estimation is not needed Practitioners may specify the properties of a distribution most relevant for their problem (incorporating prior knowledge via choice of the kernel) If a characteristic kernel is used, then the embedding can uniquely preserve all information about a distribution, while thanks to the kernel trick, computations on the potentially infinite-dimensional RKHS can be implemented in practice as simple Gram matrix operations Dimensionality-independent rates of convergence for the empirical kernel mean (estimated using samples from the distribution) to the kernel embedding of the true underlying distribution can be proven. Learning algorithms based on this framework exhibit good generalization ability and finite sample convergence, while often being simpler and more effective than information theoretic methods Thus, learning via the kernel embedding of distributions offers a principled drop-in replacement for information theoretic approaches and is a framework which not only subsumes many popular methods in machine learning and statistics as special cases, but also can lead to entirely new learning algorithms. == Definitions == Let X {\displaystyle X} denote a random variable with domain Ω {\displaystyle \Omega } and distribution P {\displaystyle P} . Given a symmetric, positive-definite kernel k : Ω × Ω → R {\displaystyle k:\Omega \times \Omega \rightarrow \mathbb {R} } the Moore–Aronszajn theorem asserts the existence of a unique RKHS H {\displaystyle {\mathcal {H}}} on Ω {\displaystyle \Omega } (a Hilbert space of functions f : Ω → R {\displaystyle f:\Omega \to \mathbb {R} } equipped with an inner product ⟨ ⋅ , ⋅ ⟩ H {\displaystyle \langle \cdot ,\cdot \rangle _{\mathcal {H}}} and a norm ‖ ⋅ ‖ H {\displaystyle \|\cdot \|_{\mathcal {H}}} ) for which k {\displaystyle k} is a reproducing kernel, i.e., in which the element k ( x , ⋅ ) {\displaystyle k(x,\cdot )} satisfies the reproducing property ⟨ f , k ( x , ⋅ ) ⟩ H = f ( x ) ∀ f ∈ H , ∀ x ∈ Ω . {\displaystyle \langle f,k(x,\cdot )\rangle _{\mathcal {H}}=f(x)\qquad \forall f\in {\mathcal {H}},\quad \forall x\in \Omega .} One may alternatively consider x ↦ k ( x , ⋅ ) {\displaystyle x\mapsto k(x,\cdot )} as an implicit feature mapping φ : Ω → H {\displaystyle \varphi :\Omega \rightarrow {\mathcal {H}}} (which is therefore also called the feature space), so that k ( x , x ′ ) = ⟨ φ ( x ) , φ ( x ′ ) ⟩ H {\displaystyle k(x,x')=\langle \varphi (x),\varphi (x')\rangle _{\mathcal {H}}} can be viewed as a measure of similarity between points x , x ′ ∈ Ω . {\displaystyle x,x'\in \Omega .} While the similarity measure is linear in the feature space, it may be highly nonlinear in the original space depending on the choice of kernel. === Kernel embedding === The kernel embedding of the distribution P {\displaystyle P} in H {\displaystyle {\mathcal {H}}} (also called the kernel mean or mean map) is given by: μ X := E [ k ( X , ⋅ ) ] = E [ φ ( X ) ] = ∫ Ω φ ( x ) d P ( x ) {\displaystyle \mu _{X}:=\mathbb {E} [k(X,\cdot )]=\mathbb {E} [\varphi (X)]=\int _{\Omega }\varphi (x)\ \mathrm {d} P(x)} If P {\displaystyle P} allows a square integrable density p {\displaystyle p} , then μ X = E k p {\displaystyle \mu _{X}={\mathcal {E}}_{k}p} , where E k {\displaystyle {\mathcal {E}}_{k}} is the Hilbert–Schmidt integral operator. A kernel is characteristic if the mean embedding μ : { family of distributions over Ω } → H {\displaystyle \mu :\{{\text{family of distributions over }}\Omega \}\to {\mathcal {H}}} is injective. Each distribution can thus be uniquely represented in the RKHS and all statistical features of distributions are preserved by the kernel embedding if a characteristic kernel is used. === Empirical kernel embedding === Given n {\displaystyle n} training examples { x 1 , … , x n } {\displaystyle \{x_{1},\ldots ,x_{n}\}} drawn independently and identically distributed (i.i.d.) from P , {\displaystyle P,} the kernel embedding of P {\displaystyle P} can be empirically estimated as μ ^ X = 1 n ∑ i = 1 n φ ( x i ) {\displaystyle {\widehat {\mu }}_{X}={\frac {1}{n}}\sum _{i=1}^{n}\varphi (x_{i})} === Joint distribution embedding === If Y {\displaystyle Y} denotes another random variable (for simplicity, assume the co-domain of Y {\displaystyle Y} is also Ω {\displaystyle \Omega } with the same kernel k {\displaystyle k} which satisfies ⟨ φ ( x ) ⊗ φ ( y ) , φ ( x ′ ) ⊗ φ ( y ′ ) ⟩ = k ( x , x ′ ) k ( y , y ′ ) {\displaystyle \langle \varphi (x)\otimes \varphi (y),\varphi (x')\otimes \varphi (y')\rangle =k(x,x')k(y,y')} ), then the joint distribution P ( x , y ) ) {\displaystyle P(x,y))} can be mapped into a tensor product feature space H ⊗ H {\displaystyle {\mathcal {H}}\otimes {\mathcal {H}}} via C X Y = E [ φ ( X ) ⊗ φ ( Y ) ] = ∫ Ω × Ω φ ( x ) ⊗ φ ( y ) d P ( x , y ) {\displaystyle {\mathcal {C}}_{XY}=\mathbb {E} [\varphi (X)\otimes \varphi (Y)]=\int _{\Omega \times \Omega }\varphi (x)\otimes \varphi (y)\ \mathrm {d} P(x,y)} By the equivalence between a tensor and a linear map, this joint embedding may be interpreted as an uncentered cross-covariance operator C X Y : H → H {\displaystyle {\mathcal {C}}_{XY}:{\mathcal {H}}\to {\mathcal {H}}} from which the cross-covariance of functions f , g ∈ H {\displaystyle f,g\in {\mathcal {H}}} can be computed as Cov ( f ( X ) , g ( Y ) ) := E [ f ( X ) g ( Y ) ] − E [ f ( X ) ] E [ g ( Y ) ] = ⟨ f , C X Y g ⟩ H = ⟨ f ⊗ g , C X Y ⟩ H ⊗ H {\displaystyle \operatorname {Cov} (f(X),g(Y)):=\mathbb {E} [f(X)g(Y)]-\mathbb {E} [f(X)]\mathbb {E} [g(Y)]=\langle f,{\mathcal {C}}_{XY}g\rangle _{\mathcal {H}}=\langle f\otimes g,{\mathcal {C}}_{XY}\rangle _{{\mathcal {H}}\otimes {\mathcal {H}}}} Given n {\displaystyle n} pairs of training examples { ( x 1 , y 1 ) , … , ( x n , y n ) } {\displaystyle \{(x_{1},y_{1}),\dots ,(x_{n},y_{n})\}} drawn i.i.d. from P {\displaystyle P} , we can also empirically estimate the joint distribution kernel embedding via C ^ X Y = 1 n ∑ i = 1 n φ ( x i ) ⊗ φ ( y i ) {\displaystyle {\widehat {\mathcal {C}}}_{XY}={\frac {1}{n}}\sum _{i=1}^{n}\varphi (x_{i})\otimes \varphi (y_{i})} === Conditional distribution embedding === Given a conditional distribution P ( y ∣ x ) , {\displaystyle P(y\mid x),} one can define the corresponding RKHS embedding as μ Y ∣ x = E [ φ ( Y ) ∣ X ] = ∫ Ω φ ( y ) d P ( y ∣ x ) {\displaystyle \mu _{Y\mid x}=\mathbb {E} [\varphi (Y)\mid X]=\int _{\Omega
17776
17776 (also known as What Football Will Look Like in the Future) is a serialized speculative fiction multimedia narrative by Jon Bois, published online through SB Nation. Set in the distant future in which all humans have become immortal and infertile, the series follows three sapient space probes that watch humanity play an evolved form of American football in which games can be played for millennia over distances of thousands of miles. The series debuted on July 5, 2017, and new chapters were published daily until the series concluded with its twenty-fifth chapter on July 15, 2017. Bois began developing 17776 in 2016. Because the story incorporates text, animated GIFs, still images, and videos hosted on YouTube, new tools were developed to allow it to be hosted efficiently on the SB Nation website. The work explores themes of consciousness, hope, despair, and why humans play sports. 17776 was well received by critics, who praised it for its innovative use of its medium and for the depth of emotion it evoked. In 2018, the story won a National Magazine Award for Digital Innovation and was longlisted for both the Hugo Awards for Best Novella and Best Graphic Story. It is followed by a sequel series: 20020, released from September to October 2020. The sequel series follows a 111-team game of college football on fields spanning 130,000 miles (210,000 km) across the United States. Bois originally intended to follow up with a further series entitled 20021; however, it was postponed indefinitely. In May 2025, Bois announced that the series would be continued with a novel titled 50007: An American Football Odyssey. == Premise == The story takes place on a future Earth where humans stopped dying, aging, and being born on April 7, 2026. All social ills were subsequently eliminated, and technology preventing humans from any injury was developed. In the United States, American football evolved to include new rules, including those that allow fields thousands of miles long, hundreds of in-game players, and games millennia long. Over time, computers gained sentience due to constant exposure to broadcast human data. By the year 17776, the space probe Pioneer 9 (called Nine) has gained sentience and made contact with Pioneer 10 (called Ten) and the Jupiter Icy Moons Explorer (called Juice). As Nine adjusts to a world radically different from that of the 20th century, the three space probes watch multiple football games occurring across the United States: a game using the entirety of Nebraska as a field in which the next point scored wins the game; a game in which players strive to possess every existing football autographed by obscure NFL player Koy Detmer; a game played between the Canadian border and the Mexican border deadlocked for 13,000 years at the bottom of a gorge in Arizona; an NFL regulation game between the Denver Broncos and the Pittsburgh Steelers that changed over 15,000 years into 58 playing teams owning and capitalizing upon portions of Sports Authority Field at Mile High while the ball is lost; a 500 game that results in the destruction of the Centennial Light; and a game in which the possessing player is attempting to score an automatic win by hiding in his team's end zone for 10,000 years. == Format == 17776 is read by scrolling through web pages occupied by large GIF images and colored dialogue text, interspersed with occasional YouTube videos. The story is divided into chapters, which were originally published in daily installments between July 5 and 15, 2017. Much of the GIF and video content of the series uses Google Earth satellite imagery, 3D buildings, and other tools within Google Earth to create animations and visual effects. == Development == Bois wrote and illustrated 17776 for Vox Media's sports news website SB Nation, of which he is creative director. Aside from 17776, Bois produces two other recurring, humorous video essay programs for the site: Pretty Good, which focuses on unusual sports topics and stories, and Chart Party, which focuses on statistics and has an emphasis on Bois' use of visual art in his journalism and storytelling. Bois is also known for the Breaking Madden series, in which he attempted unusual scenarios in the Madden NFL series of video games. In early 2016, Bois began developing an "anti-sci fi" project as a possible sequel to The Tim Tebow CFL Chronicles, an earlier work for SB Nation, and set the story in a year far enough in the future that "nobody ever thinks about it." Although he liked the concept and the visuals, he believed the project would not connect with readers and shelved it. Later, he realized that the story needed a centering character; he wrote one in the form of a small town, AM radio talk show host before coming up with the characters of the probes. Development renewed in May 2016, and the project solidified after SB Nation published its article "The Future of Football." Bois described it as the biggest project he ever attempted. The series was developed by Graham MacAree, who used a Vox Media tool that creates custom packages from standard article sets to give Bois creative leeway and to accommodate the series' weight on the SB Nation website. MacAree found that there were few resources online for achieving the desired effects. == Themes == Bois has stated that he had "conceived [17776] to give the reader a good time," asserting that this "was literally the whole point." William Hughes writing for The A.V. Club described 17776 as concerned with why humans play sports: "That is, given the massive resources, time, and information at our disposal (not to mention those available to our descendants), why does communal game-playing still hold such an important place in society?" He also listed consciousness, hope, and despair as among the work's themes. Beth Elderkin of io9 described it as "a deep thought experiment into what we consider humanly possible". She also felt that Ten and Juice take on the role of angel and devil, and she suggested the two may be unreliable narrators. Ian Crouch of The New Yorker felt that the work had a "tonal echo" of Don DeLillo's 1972 novel End Zone due to thematic similarities "with the way that the order and logic of football might act as a counterbalance to the chaos of the real world". == Reception == According to the communications director at Vox Media, 17776 garnered over 2.3 million pageviews by July 10. Two days later, it had received more than 2.9 million pageviews. Average engagement time was over nine minutes, and 43 percent of readers finished each installment of the series published by July 7. On July 19, Bois claimed that 17776 received 700,000 unique visitors and 4 million total pageviews, with an average engagement time of 11 minutes. Thu-Huong Ha for Quartz described 17776 as "part Italo Calvino, part Peter Heller [author of The Dog Stars], with humor seemingly from within the depths of Reddit," saying that the story would appeal to fans of both sports and literature. Tor.com described the first chapter as full of tension and felt that receiving answers is a "surprisingly heartbreaking" experience "lessened by a gleeful bouncing immaturity" one would not expect from the characters. Beth Elderkin at io9 said the series is "akin to Homestuck" and described it as "weird, complex, and pretty spectacular". William Hughes writing for The A.V. Club felt that 17776 is a "truly innovative piece of work". After reading the first three chapters, Agatha French of the Los Angeles Times stated that she was "impressed and excited by the innovation" of what she saw, and that she was intrigued despite not knowing what the work is or is saying. She felt the work took full advantage of its online medium and suggested that it "may also be a glimpse into the future of reading on the Internet". Ian Crouch of The New Yorker described the series as, "despite its seemingly meagre parts, a thing of startling beauty". Of the chapters published by July 12, he felt "the most striking chapter" to be one that used audio of Verne Lundquist calling the end of a 2013 game between the University of Alabama and Auburn University over a video panning over Earth. He also noted that the series was compared to Homestuck and relayed additional comparisons to Thomas Pynchon novels and "a Reddit thread hijacked by robot trolls". The series won the inaugural National Magazine Award for Digital Innovation from the American Society of Magazine Editors; this was the first National Magazine Award nomination and win for SB Nation. It was described by the judges as "an extraordinary combination of art, fiction and technology, an online acid trip that had to be experienced to be believed." It was also longlisted for the Hugo Awards for Best Novella and Best Graphic Story in 2018, ultimately finishing in 11th place in both categories. == Sequel series == On September 28, 2020, a sequel titled 20020 was launched on Secret Base, a branch of SB Nation; on October 13, it was revea
Interim Measures for the Management of Generative AI Services
The Interim Measures for the Management of Generative AI Services (Chinese: 生成式人工智能服务管理暂行办法; pinyin: Shēngchéng shì réngōng zhìnéng fúwù guǎnlǐ zànxíng bànfǎ) are a set of regulations governing public-facing generative artificial intelligence services in China. Issued on 10 July 2023 and effective from 15 August 2023, they were China's first binding regulation specifically targeting generative AI. They have been described as among the earliest such regulations adopted by any country. The measures were jointly issued by the Cyberspace Administration of China (CAC) and six other national bodies: the National Development and Reform Commission, the Ministry of Education, the Ministry of Science and Technology, the Ministry of Industry and Information Technology, the Ministry of Public Security, and the National Radio and Television Administration. Among the measures' most prominent requirements is that generative AI services must uphold Core Socialist Values and must not generate content that could subvert state power, harm national security, or undermine social stability. The measures also require providers of public-facing generative AI services to undergo security assessments and register their algorithms with the CAC. As of December 2025, 748 generative AI services had completed the filing process at the national level. == Background == The Interim Measures build on two earlier sets of regulations targeting specific algorithm applications. The Administrative Provisions on Algorithm Recommendation for Internet Information Services, effective from March 2022, established China's algorithm registry and required providers of recommendation algorithms with "public opinion properties or social mobilization capabilities" to file with the CAC and undergo security assessments. The Administrative Provisions on Deep Synthesis of Internet Information Services, effective from January 2023, extended similar requirements to algorithms used for generating synthetic media such as deepfakes. In April 2023, the CAC released a draft of the generative AI regulation for public comment. The draft included several requirements that attracted attention, including that generated content should "embody Core Socialist Values" and that training data should be "true and accurate". The public consultation period ran until May 2023. The final version, published in July 2023, was substantially revised from the draft. According to an analysis by the Future of Privacy Forum, changes appeared to reflect feedback from industry stakeholders including Baidu, Xiaomi, SenseTime, and others, as well as input from government-affiliated research institutes. The final measures adopted a more permissive tone, with the CAC describing its approach as "inclusive and prudent" (包容审慎) and emphasising "classified and graded" (分类分级) supervision. == Scope == The measures apply to services that use generative AI technology to provide text, images, audio, video, or other content to the public within mainland China (Article 2). They do not apply to organisations that develop or use generative AI internally without offering services to the domestic public, such as industry associations, enterprises, and research institutions. Overseas providers whose services are accessible to users in China are also subject to the measures. == Key provisions == === Content requirements === Article 4 sets out the core content obligations. Providers and users of generative AI services must uphold the Core Socialist Values. The measures prohibit generating content that incites subversion of national sovereignty or the socialist system, endangers national security or the nation's image, incites separatism, promotes terrorism or extremism, promotes ethnic hatred or discrimination, or contains violence, obscenity, or false information prohibited by law. These content prohibitions largely mirror those in Article 12 of the Cybersecurity Law and in prior regulations governing online content. Article 4 also requires that models be designed and trained to avoid discrimination, that services respect intellectual property rights, and that providers take effective measures to improve the transparency and accuracy of generated content. === Training data and labelling === Article 7 requires providers to ensure that training data is of high quality and legitimately sourced, and that it does not infringe upon intellectual property rights. Where personal information is used, consent must be obtained. The final version of this provision removed language from the draft that would have held providers responsible for the "legitimacy" of all pretraining data, replacing it with a requirement to "employ effective measures to improve the quality of training data". Article 8 requires providers to establish labelling rules for training data and to conduct quality assessments of data annotations. Article 12 requires that generated images, videos, and other synthetic content be labelled as AI-generated. === User rights and privacy === Article 11 requires providers to protect user privacy, to minimise the collection and retention of personal data, and to refrain from unlawfully sharing user information. Users have the right to request review, correction, or deletion of their personal information. Article 10 requires providers to take measures to prevent excessive dependence on or addiction to generative AI services by minors. === Security assessment and algorithm filing === Article 17 requires that providers of generative AI services with "public opinion properties or the capacity for social mobilization" (具有舆论属性或者社会动员能力) carry out security assessments and complete algorithm filing procedures in accordance with the Administrative Provisions on Algorithm Recommendation for Internet Information Services. == Implementation == === Algorithm filing process === In practice, the filing requirements under the Interim Measures have developed into a two-tier process. The first tier is the standard algorithm filing (算法备案) under the pre-existing Algorithm Recommendation Provisions, which involves submitting information about an algorithm's design, purpose, and data sources to the CAC. This process is primarily a registration mechanism. For public-facing generative AI products, there is an additional, more rigorous process commonly referred to as the "large model filing" (大模型备案). This involves submitting a security self-assessment report, data annotation rules, a keyword blocking list, and evaluation test question sets. The process includes technical testing at the provincial level, followed by review at the national CAC level. The algorithm filing targets specific algorithms, while the large model filing evaluates the broader system architecture, training data, model parameters, and potential social impact. The CAC publishes lists of generative AI services that have successfully completed the filing process. The first such list was published on 2 April 2024. According to the CAC's year-end announcements, 302 generative AI services had completed national-level filing by the end of 2024 (of which 238 were new that year), alongside 105 applications that completed local-level registration. By the end of 2025, the cumulative total had risen to 748 national-level filings and 435 local-level registrations. === Content compliance and testing === According to the Carnegie Endowment, the CAC has conducted compliance audits of generative AI services with a particular focus on ensuring appropriate responses to queries about politically sensitive topics. The large model filing process requires providers to pass both provincial-level and national-level technical testing before their services can be made available to the public. On 1 March 2024, the National Technical Committee 260 on Cybersecurity (TC260) published TC260-003, the Basic Security Requirements for Generative AI Services (生成式人工智能服务安全基本要求), a technical standard that provides detailed guidance on the security assessments required under the Interim Measures. The standard covers requirements for training data safety, model security, and content safety evaluation, and is used as a reference for the filing process. == Analysis == === Relationship to broader Chinese internet regulation === The content requirements in the Interim Measures extend China's existing framework for online information control to generative AI. Legal scholars have noted that the "Core Socialist Values" provision and the specific content prohibitions are consistent with longstanding requirements imposed on internet platforms under the Cybersecurity Law and related regulations. The Asia Society Policy Institute has described the Chinese government's highest regulatory priority in this area as retaining control of information, noting that content-related obligations receive stricter enforcement than other provisions. === Nature of the filing system === The character of the filing system has been debated by scholars. Angela Huyue Zh
Perry Rhodan
Perry Rhodan is a German space opera franchise, named after its hero. It commenced in 1961 and has been ongoing for decades, written by an ever-changing team of authors. Having sold approximately two billion copies (in novella format) worldwide (including over one billion in Germany alone), it is the most successful science fiction book series ever written. The first billion of worldwide sales was celebrated in 1986. The series has spun off into comic books, audio dramas, video games and the like. A reboot, Perry Rhodan NEO, was launched in 2011 and began publication in English in April 2021. == Print publication == The series has spun off into many different forms of media, but originated as a serial novella published weekly since 8 September 1961 in the Romanheft (Meaning "Magazine novel") format. These are digest-sized booklets, usually containing 66 pages, the German equivalent of the now-defunct (and generally longer) American pulp magazine. They are published by Pabel-Moewig Verlag, a subsidiary of Bauer Media Group headquartered in Hamburg. As of February 2019, 3000 booklet novels of the original series, 850 spinoff novels of the sister series Atlan and over 400 paperbacks and 200 hardcover editions have been published, totalling over 300,000 pages. == English translation == The first 126 novels (plus five novels of the spinoff series Atlan) were translated into English and published by Ace Books between 1969 and 1978, with the same translations used for the British edition published by Futura Publications which issued only 39 novels. When Ace cancelled its translation of the series, translator Wendayne Ackerman self-published the following 19 novels (under the business name 'Master Publications') and made them available by subscription only. Financial disputes with the German publishers led to the cancellation of the American translation in 1979. An attempt to revive the series in English was made in 1997–1998 by Vector Publications of the US, which published translations of four issues (1800–1803) from the current storyline being published in Germany at the time. The series and its spin-offs have captured a substantial fraction of the original German science fiction output and exert influence on many German writers in the field. == Structure == The series is told in an arc storyline structure. An arc—called a "cycle"—would have anywhere from 25 to 100 issues devoted to it. Similar subsequent cycles are referred to as a "grand-cycle". == History == ‘Perry Rhodan, der Erbe des Universums’ (Eng: ‘The Heir to the Universe’, though the American/British editions instead used the subtitle 'Peacelord of the Universe') was created by German science fiction authors K. H. Scheer and Walter Ernsting and launched in 1961 by German publishing house Arthur Moewig Verlag (now Pabel-Moewig Verlag). Originally planned as a 30 to 50 volume series, it has been published continuously every week since, celebrating the 3000th issue in 2019. Written by an ever-changing team of authors, many of whom, however, remained with the series for decades or life, Perry Rhodan is issued in weekly novella-size installments in the traditional German Heftroman (pulp booklet) format. Unlike most German Heftromane, Perry Rhodan consists not of unconnected novels but is a series with a continuous, increasingly complex plotline, with frequent back references to events. In addition to its original Heftroman form, the series now also appears in hardcovers, paperbacks, e-books, comics and audiobooks. Over the decades there have also been comic strips, numerous collectibles, several encyclopedias, audio plays, inspired music, etc. The series has seen partial translations into several languages. It also spawned the German-Italian-Spanish 1967 movie Mission Stardust, which is widely considered so terrible that many fans of the series pretend it never existed. Coinciding with the 50th-anniversary World Con, on 30 September 2011, a new series named Perry Rhodan Neo began publication, attracting new readers with a reboot of the story, starting in the year 2036 instead of 1971, and a related but independent story-line. On 2 April 2021, light novel and manga publisher J-Novel Club announced Perry Rhodan NEO as a launch title for its new J-Novel Pulp imprint, making this the first ongoing English release of new Perry Rhodan serials in over 20 years. It has become the most popular science fiction book series of all time. == Overview == === Fictional history === The story begins in 1971. During the first human Moon landing by US Space Force Major Perry Rhodan and his crew, they discover a marooned extraterrestrial space ship from the fictional planet Arkon, located in the (real) M13 cluster. Appropriating the Arkonide technology, they proceed to unify Terra and carve out a place for humanity in the galaxy and the cosmos. Two of the accomplishments that enable them to do so are positronic brains and starship drives for near-instantaneous hyperspatial translation. These were directly borrowed from Isaac Asimov's science fiction. As the series progresses, major characters, including the title character, are granted relative immortality. They are immune to age and disease, but not to violent death. The story continues over the course of millennia and includes flashbacks thousands and even millions of years into the past. The scope widens to encompass other galaxies, even more remote regions of space, parallel universes and cosmic structures, time travel, paranormal powers, a variety of aliens ranging from threatening to endearing, and bodiless entities, some of which have godlike powers. === Multiverse === The universe in which the main plot generally takes place is called the Einstein Universe (or "Meekorah"). Its laws are for the most part identical to those of the real universe, as known by late 20th century science. Newer theories about dark matter and dark energy are currently not used in the series. The laws of nature follow old theories that have been disproven, in order to protect series continuity. There are many other universes, each to a greater or lesser extent different from the familiar one, in which, for example one in which time runs slower, an anti-matter universe, a shrinking universe, etc. Each universe possesses its owntimelines, which are for the most part unreachable from each other but may be accessed by special means, thereby itself creating many more parallel timelines. The Einstein Universe is embedded in a high-dimensional manifold, called Hyperspace. This hyperspace consists of several subspaces use for faster-than-light travel by technological means. The exact traits of those higher dimensions are got yhr mode pity unexplained. The border of the universe is a dimension called the deep, once used for construction of the gigantic disc-shaped world Deepland. === Psionic Web and Moral Code === The Psionic Web crosses the whole universe, constantly emitting "vital energy" and "psionic energy", guaranteeing normal (organic among others) life and the wellbeing of higher entities. The Moral Code crosses through all universes, and is linked to the Psionic Web. It is subdivided into the Cosmogenes, which are again subdivided into the Cosmonucleotids. The Cosmonucleotids determine reality and fate for their respective parts of a given universe, via messengers. Higher beings are trying to gain control of this Code to rule reality. The Moral Code itself was not installed by the higher beings, the higher powers by themselves have no clue why or by whom the Code was made. Once the Cosmocrats ordered Perry Rhodan to find the answer to the third ultimate question: "Who initiated the LAW and what does it accomplish?" Perry Rhodan had the chance to receive the answer at the mountain of creation, but refused, as he knew that the answer would destroy his mind. The negative Superintelligence Koltoroc had received the answer to the last ultimate question, 69 million years BC at Negane Mountain, but it is not known if it made any use of the information. === Onion-shell model === An evolutionary schema, similar to the Great Chain of Being, called the "onion-shell model" is employed in relationship to all life. Here, continuous evolution is from lower to higher lifeforms, culminating in bodiless entities. Later in the series, further lifeforms, representing stages between the known shells, were introduced. The main shells are: Lifeless matter Bacteria Higher animals Intelligent species Intelligent species that have contacted other species Superintelligences (SI) Matter sources/ Matter sinks Cosmocrats / Chaotarchs (High Powers) Powers close to the "Horizon of the LAW", the essence of the Multiverse The Superintelligences are the next step above normal minds. They can be born, for example, when a species collectively gives up its bodies and unites their spirits. Such Superintelligences may claim as their domain areas consisting of up to several galaxies (the entity known as "E
Fabric Connect
Fabric Connect, in computer networking usage, is the name used by Extreme Networks to market an extended implementation of the IEEE 802.1aq and IEEE 802.1ah-2008 standards. The Fabric Connect technology was originally developed by the Enterprise Solutions R&D department within Nortel Networks. In 2009, Avaya, Inc acquired Nortel Networks Enterprise Business Solutions; this transaction included the Fabric Connect intellectual property together with all of the Ethernet Switching platforms that supported it. Subsequently, the Fabric Connect technology became part of the Extreme Networks portfolio by virtue of their 2017 purchase of the Avaya Networking business and assets. It was during the Avaya era that this technology was promoted as the lead element of the Virtual Enterprise Network Architecture (VENA). == Technologies == === Fabric Connect === Fabric Connect's provides network-wide, end-to-end, multi-layer virtualization. A network virtualization capability, based on an enhanced implementation of the IEEE 802.1aq Shortest Path Bridging (SPB) standard, Fabric Connect offers the ability to create a simplified network that can dynamically virtualize elements to efficiently provision and utilize resources, thus reducing the strain on the network and personnel. Extreme Networks base the Fabric Connect technology on the SPB standard, including support for RFC 6329, and have integrated IP Routing and IP Multicast support; this unified technology allows for the replacement of multiple conventional protocols such as Spanning Tree, RIP and/or OSPF, ECMP, and PIM. === Fabric Attach === An adjunct to the Fabric Connect technology, Fabric Attach allows network operators to extend network virtualization directly into conventional wiring closets (using existing non-Fabric Ethernet switches) and automate the provisioning of devices to their appropriate virtual network. This is particularly relevant for the mass of unattended network end-point that are now appearing, such as IP Phones, Wireless Access Points, and IP Cameras. Fabric Attach standardized protocols such as 802.1AB LLDP to exchange credentials and obtain provisioning information that allows "Client" Switches to be automatically re-configured on the fly with parameters that let Traffic Flows Map through to Fabric Connect Edge Switches (aka "Backbone Edge Bridge" in SPB definition) functioning as a Fabric Attach "Server" Switch. This method is described by an IETF "Internet Draft", pending further standardization activity. Fabric Attach is typically used to automate Wiring Closet connectivity, but has the potential to be extensible for use in the Data Center, with Virtual Machines being able to dynamically request VLAN/VSN (Virtual Service Network) assignment based upon application requirements. == Hardware products == === Virtual Services Platform 9000 Series === A range of modular chassis-based products, featuring a carrier-grade Linux operation system, and designed for high-performance deployment scenarios that need to scale to multiple terabits of switching capacity and support 10 and 40 gigabit Ethernet connections, and is designed eventually to support 100 gigabit Ethernet. === Virtual Services Platform 8000 Series === A compact form-factor platform delivering high-density 10/40 gigabit Ethernet connectivity, and targeted at mid-market through to mid-size enterprise core switch applications. === Virtual Services Platform 7000 Series === A range of high-end 10 gigabit Ethernet stackable switches that extend fabric-based networking to the data center top-of-rack. They support 40 gigabit Ethernet via the MDA Slot. === Virtual Services Platform 4000 Series === A range of high-end gigabit Ethernet stackable switches that extend Fabric-based networking to branch and metro locations. === Ethernet Routing Switch 5000 Series === A range of high-end gigabit Ethernet stackable switches that provides enterprise-class desktop features, including PoE, and offers 10 Gbit/s uplink connections. Each Switch supports up to 144 Gbit/s of virtual backplane capacity, delivering up to 1.152 Tbit/s for a system of eight, creating a virtual backplane through a stacking configuration. === Ethernet Routing Switch 4000 Series === A range of gigabit Ethernet stackable switches that provide enterprise-class desktop features, including PoE/PoE+, and offer 1/10 Gbit/s uplink connections. Each switch supports up to 48 Gbit/s of virtual backplane capacity, delivering up to 384 Gbit/s for a system of 8, creating a virtual backplane through a stacking configuration. === Ethernet Routing Switch 3500 Series === These entry-level gigabit Ethernet stackable switches provide enterprise-class desktop features, including PoE/PoE+, and 1 Gbit/s uplink connections.
Mars Plus
Mars Plus is a 1994 science fiction novel by American writer Frederik Pohl and Thomas T. Thomas. It is the sequel to Pohl's 1976 novel Man Plus, which is about a cyborg, Roger Torraway, who is designed to operate in the harsh Martian environment, so that humans can start to colonize Mars. Mars Plus is set fifty years after the first novel. Young Demeter Coghlan travels to Mars, now settled by humans and cyborgs, and finds herself amidst a rebellion by the colonists. == Plot == In Man Plus, set in the not-too-distant future, with threat of the Cold War becoming a fighting war, people plan for the colonization of Mars to escape the seemingly-inevitable Armageddon. The American government begins a cyborg program to create a being capable of surviving the harsh Martian environment: a "Man Plus" called Roger Torraway who is converted from man to cyborg. While his cyborg body is adapted to Mars, he feels strange at first. As more nations develop cyborgs, the computer networks of Earth become sentient. Mars Plus is set fifty years after the first novel, when Mars is settled by humans and cyborgs. The cyborg Torroway is in the novel, but he is not the main character. The protagonist is Demeter Coghlan, a young woman from Earth who travels to Mars. Demeter is seeking information about a canyon that she believes may be significant if the colonists begin to convert Mars to an Earth-like planet. Amidst a backdrop of spies and newly dispatched Earth diplomats, the inexperienced Demeter senses that tensions are rising on the planet. She is further disoriented due to recovering from an accident. Despite the risks in the region, Demeter has intense sexual encounters with some of the local colonists. When the locals rebel against the surveillance set up by the computer network, Demeter is kidnapped by the computer network. == Reception == The reviewer from SFBook Reviews criticizes the book, saying "nothing really happens" and stating that there is no linkage to Man Plus apart from the presence of the cyborg Torraway; moreover, the reviewer states that the questions posed in the first novel are not answered. SF Reviews calls Mars Plus "...not as good as Man Plus but...not bad", and it is praised for "...some nice touches: Demeter continuously forgetting to think about geology; her careless dictation to the computer and her irresistible urges for wild sex." SF Reviews criticizes the writing in Mars Plus for being "...a little careless in places" and in need of more "...more crafting and pruning."