Secret London is a Facebook group started by 21-year-old Bristol University graduate, Tiffany Philippou, on 19 January 2010 in response to a Saatchi & Saatchi competition. The group grew rapidly (180,000 members as of 8 February 2010) and is composed mostly of Londoners who use the site to share suggestions and photos of London. After the group's early success, the founder announced her intention to launch a website of the same name by crowdsourcing the design and development. The website was launched on 16 February 2010. == Other secret cities == Following the initial success of Secret London, a number of other secret groups were independently started around the world, some of which already have over 100,000 users. As of 19 February 2010, the list of other groups includes: Secret Frankfurt, Secret Tel Aviv, Secret Paris, Secret New York, Secret Tokyo, Secret Toronto, Secret Los Angeles, Secret Exeter, Secret Boston, Secret Norwich, Secret Singapore, Secret Brighton, Secret Minneapolis, Secret Sydney, Secret Canberra, Secret Brisbane, Secret Wellington, Secret Christchurch, Secret Madeira, Secret Funchal, Secret Bristol and Secret Cardiff. == Controversy == Some commentators have questioned whether it possible to share secrets without compromising them, and whether sharing tips publicly will lead to over-exposure of the businesses who are recommended.
Texture atlas
In computer graphics, a texture atlas (also called a spritesheet or an image sprite in 2D game development) is an image containing multiple smaller images, usually packed together to reduce overall dimensions. An atlas can consist of uniformly-sized images or images of varying dimensions. A sub-image is drawn using custom texture coordinates to pick it out of the atlas. == Benefits == In an application where many small textures are used frequently, it is often more efficient to store the textures in a texture atlas which is treated as a single unit by the graphics hardware. This reduces both the disk I/O overhead and the overhead of a context switch by increasing memory locality. Careful alignment may be needed to avoid bleeding between sub textures when used with mipmapping and texture compression. In web development, images are packed into a sprite sheet to reduce the number of image resources that need to be fetched in order to display a page. == Gallery ==
Defuzzification
Defuzzification is the process of producing a quantifiable result in crisp logic, given fuzzy sets and corresponding membership degrees. It is the process that maps a fuzzy set to a crisp set. It is typically needed in fuzzy control systems. These systems will have a number of rules that transform a number of variables into a fuzzy result, that is, the result is described in terms of membership in fuzzy sets. For example, rules designed to decide how much pressure to apply might result in "Decrease Pressure (15%), Maintain Pressure (34%), Increase Pressure (72%)". Defuzzification is interpreting the membership degrees of the fuzzy sets into a specific decision or real value. The simplest but least useful defuzzification method is to choose the set with the highest membership, in this case, "Increase Pressure" since it has a 72% membership, and ignore the others, and convert this 72% to some number. The problem with this approach is that it loses information. The rules that called for decreasing or maintaining pressure might as well have not been there in this case. A common and useful defuzzification technique is center of gravity. First, the results of the rules must be added together in some way. The most typical fuzzy set membership function has the graph of a triangle. Now, if this triangle were to be cut in a straight horizontal line somewhere between the top and the bottom, and the top portion were to be removed, the remaining portion forms a trapezoid. The first step of defuzzification typically "chops off" parts of the graphs to form trapezoids (or other shapes if the initial shapes were not triangles). For example, if the output has "Decrease Pressure (15%)", then this triangle will be cut 15% the way up from the bottom. In the most common technique, all of these trapezoids are then superimposed one upon another, forming a single geometric shape. Then, the centroid of this shape, called the fuzzy centroid, is calculated. The x coordinate of the centroid is the defuzzified value. == Methods == There are many different methods of defuzzification available, including the following: AI (adaptive integration) BADD (basic defuzzification distributions) BOA (bisector of area) CDD (constraint decision defuzzification) COA (center of area) COG (center of gravity) ECOA (extended center of area) EQM (extended quality method) FCD (fuzzy clustering defuzzification) FM (fuzzy mean) FOM (first of maximum) GLSD (generalized level set defuzzification) ICOG (indexed center of gravity) IV (influence value) LOM (last of maximum) MeOM (mean of maxima) MOM (middle of maximum) QM (quality method) RCOM (random choice of maximum) SLIDE (semi-linear defuzzification) WFM (weighted fuzzy mean) The maxima methods are good candidates for fuzzy reasoning systems. The distribution methods and the area methods exhibit the property of continuity that makes them suitable for fuzzy controllers.
Hyperion Cantos
The Hyperion Cantos is a series of science fiction novels by Dan Simmons. The title was originally used for the collection of the first pair of books in the series, Hyperion and The Fall of Hyperion, and later came to refer to the overall storyline, including Endymion, The Rise of Endymion, and a number of short stories. More narrowly, inside the fictional storyline, after the first volume, the Hyperion Cantos is an epic poem written by the character Martin Silenus covering in verse form the events of the first two books. Of the four novels, Hyperion received the Hugo and Locus Awards in 1990; The Fall of Hyperion won the Locus and British Science Fiction Association Awards in 1991; and The Rise of Endymion received the Locus Award in 1998. All four novels were also nominated for various science fiction awards. == Works == === Hyperion (1989) === First published in 1989, Hyperion has the structure of a frame story, similar to Geoffrey Chaucer's Canterbury Tales and Giovanni Boccaccio's Decameron. The story weaves the interlocking tales of a diverse group of travelers sent on a pilgrimage to the Time Tombs on Hyperion. The travelers have been sent by the Hegemony (the government of the human star systems), the All Thing, and the Church of the Final Atonement, alternately known as the Shrike Church, to make a request of the Shrike. As they progress in their journey, each of the pilgrims tells their tale. === The Fall of Hyperion (1990) === This book concludes the story begun in Hyperion. It abandons the storytelling frame structure of the first novel, and is instead presented primarily as a series of dreams by John Keats. === Endymion (1996) === The story commences 274 years after the events in the previous novel. Few main characters from the first two books are present in the later two. The main character is Raul Endymion, an ex-soldier who receives a death sentence after an unfair trial. He is rescued by Martin Silenus and asked to perform a series of rather extraordinarily difficult tasks. The main task is to rescue and protect the daughter of Brawne Lamia (one of the main characters of Hyperion), Aenea, a messiah coming from the time period just after the first books via time travel. The Catholic Church has become a dominant force in the human universe and views Aenea as a potential threat to their power. The group of Aenea, Endymion, and A. Bettik (an android) evades the Church's forces on several worlds through use of the Consul's spaceship, ending the story on Earth. === The Rise of Endymion (1997) === This final novel in the series finishes the story begun in Endymion, expanding on the themes in Endymion, as Raul and Aenea battle the Church and meet their respective destinies. === Short stories === The series also includes three short stories: "Remembering Siri" (1983, included almost verbatim in Hyperion) "The Death of the Centaur" (1990) "Orphans of the Helix" (1999) == Development == The Hyperion universe originated when Simmons was an elementary school teacher, as an extended tale he told at intervals to his young students; this is recorded in "The Death of the Centaur", and its introduction. It then inspired his short story "Remembering Siri", which eventually became the nucleus around which Hyperion and The Fall of Hyperion formed. After the quartet was published came the short story "Orphans of the Helix". "Orphans" is currently the final work in the Cantos, both chronologically and internally. The original Hyperion Cantos has been described as a novel published in two volumes, published separately at first for reasons of length. In his introduction to "Orphans of the Helix", Simmons elaborates: Some readers may know that I've written four novels set in the "Hyperion Universe"—Hyperion, The Fall of Hyperion, Endymion, and The Rise of Endymion. A perceptive subset of those readers—perhaps the majority—know that this so-called epic actually consists of two long and mutually dependent tales, the two Hyperion stories combined and the two Endymion stories combined, broken into four books because of the realities of publishing. == Influences == Much of the appeal of the series stems from its extensive use of references and allusions from a wide array of thinkers such as Teilhard de Chardin, John Muir, Norbert Wiener, and to the poetry of John Keats, the famous 19th-century English Romantic poet, Norse mythology, and the monk Ummon. A large number of technological elements are acknowledged by Simmons to be inspired by elements of Out of Control: The New Biology of Machines, Social Systems, and the Economic World. The Hyperion series has many echoes of Jack Vance, explicitly acknowledged in one of the later books. The title of the first novel, "Hyperion", is taken from one of Keats's poems, the unfinished epic Hyperion. Similarly, the title of the third novel is from Keats' poem Endymion. Quotes from actual Keats poems and the fictional Cantos of Martin Silenus are interspersed throughout the novels. Simmons goes so far as to have two artificial reincarnations of John Keats ("cybrids": artificial intelligences in human bodies) play a major role in the series. == Setting == Much of the action in the series takes place on the planet Hyperion. It is described as having one-fifth less gravity than Earth standard. Hyperion has a number of peculiar indigenous flora and fauna, notably Tesla trees, which are essentially large electricity-spewing trees. It is also a "labyrinthine" planet, which means that it is home to ancient subterranean labyrinths of unknown purpose. Most importantly, Hyperion is the location of the Time Tombs, large artifacts surrounded by "anti-entropic" fields that allow them to move backward through time. In the fictional universe of the Hyperion Cantos, the Hegemony of Man encompasses over 200 planets. Faster than light communications technology, Fatlines, are said to operate through tachyon bursts. However, in later books it is revealed that they operate through the Void Which Binds. The Farcaster network was given to humanity by the TechnoCore and again it was another use of the Void Which Binds that allowed this instantaneous travel between worlds. The Hawking Drive was developed by human scientists, allowing the faster than light travel which led to the Hegira (from the Arabic word هجرة Hijra, meaning 'migration'). The Gideon drive, a Core-provided starship drive, allows for near-instantaneous travel between any two points in human-occupied space. The drive's use kills any human on board a Gideon-propelled starship; thus, the technology is only of use with remote probes or when used in conjunction with the Pax's resurrection technology. The resurrection creche can regenerate someone carrying a cruciform from their remains. Treeships are living trees that are propelled by ergs (spider-like solid-state alien being that emits force fields) through space. === The Shrike === The region of the Tombs is also the home of the Shrike, a menacing half-mechanical, half-organic four-armed creature that features prominently in the series. It appears in all four Hyperion Cantos books and is an enigma in the initial two; its purpose is not revealed until the second book, but is still left nebulous. The Shrike appears to act both autonomously and as a servant of some unknown force or entity. In the first two Hyperion books, it exists solely in the area around the Time Tombs on the planet Hyperion. Its portrayal is changed significantly in the last two books, Endymion and The Rise of Endymion. In these novels, the Shrike appears effectively unfettered and protects the heroine Aenea against assassins of the opposing TechnoCore. Surrounded in mystery, the object of fear, hatred, and even worship by members of the Church of the Final Atonement (the Shrike Cult), the Shrike's origins are described as uncertain. It is portrayed as composed of razorwire, thorns, blades, and cutting edges, having fingers like scalpels and long, curved toe blades. It has the ability to control the flow of time, and may thus appear to travel infinitely fast. The Shrike may kill victims in a flash or it may transport them to an eternity of impalement upon an enormous artificial 'Tree of Thorns,' or 'Tree of Pain' in Hyperion's distant future. The Tree of Thorns is described as an unimaginably large, metallic tree, alive with the agonized writhing of countless human victims of all ages and races. It is also hinted in the second book that the Tree of Thorns is actually a simulation generated by a mystical interface which connects to human brains via a strong and pulsing (as if it were alive) cord. The name Shrike seems a reference to birds of the shrike family, a family of birds that impales their victims on thorns, spines, or twigs. === Worlds and Systems === In the fictional universe of the Hyperion Cantos, the Hegemony of Man encompasses over 200 planets. The following planets appear or are specifically mentioned in the Hyperion Cantos. Planets of
Deep learning speech synthesis
Deep learning speech synthesis refers to the application of deep learning models to generate natural-sounding human speech from written text (text-to-speech) or spectrum (vocoder). Deep neural networks are trained using large amounts of recorded speech and, in the case of a text-to-speech system, the associated labels and/or input text. == Formulation == Given an input text or some sequence of linguistic units Y {\displaystyle Y} , the target speech X {\displaystyle X} can be derived by X = arg max P ( X | Y , θ ) {\displaystyle X=\arg \max P(X|Y,\theta )} where θ {\displaystyle \theta } is the set of model parameters. Typically, the input text will first be passed to an acoustic feature generator, then the acoustic features are passed to the neural vocoder. For the acoustic feature generator, the loss function is typically L1 loss (Mean Absolute Error, MAE) or L2 loss (Mean Square Error, MSE). These loss functions impose a constraint that the output acoustic feature distributions must be Gaussian or Laplacian. In practice, since the human voice band ranges from approximately 300 to 4000 Hz, the loss function will be designed to have more penalty on this range: l o s s = α loss human + ( 1 − α ) loss other {\displaystyle loss=\alpha {\text{loss}}_{\text{human}}+(1-\alpha ){\text{loss}}_{\text{other}}} where loss human {\displaystyle {\text{loss}}_{\text{human}}} is the loss from human voice band and α {\displaystyle \alpha } is a scalar, typically around 0.5. The acoustic feature is typically a spectrogram or Mel scale. These features capture the time-frequency relation of the speech signal, and thus are sufficient to generate intelligent outputs. The Mel-frequency cepstrum feature used in the speech recognition task is not suitable for speech synthesis, as it reduces too much information. == History == In September 2016, DeepMind released WaveNet, which demonstrated that deep learning-based models are capable of modeling raw waveforms and generating speech from acoustic features like spectrograms or mel-spectrograms. Although WaveNet was initially considered to be computationally expensive and slow to be used in consumer products at the time, a year after its release, DeepMind unveiled a modified version of WaveNet known as "Parallel WaveNet," a production model 1,000 faster than the original. This was followed by Google AI's Tacotron 2 in 2018, which demonstrated that neural networks could produce highly natural speech synthesis but required substantial training data—typically tens of hours of audio—to achieve acceptable quality. Tacotron 2 used an autoencoder architecture with attention mechanisms to convert input text into mel-spectrograms, which were then converted to waveforms using a separate neural vocoder. When trained on smaller datasets, such as 2 hours of speech, the output quality degraded while still being able to maintain intelligible speech, and with just 24 minutes of training data, Tacotron 2 failed to produce intelligible speech. In 2019, Microsoft Research introduced FastSpeech, which addressed speed limitations in autoregressive models like Tacotron 2. FastSpeech utilized a non-autoregressive architecture that enabled parallel sequence generation, significantly reducing inference time while maintaining audio quality. Its feedforward transformer network with length regulation allowed for one-shot prediction of the full mel-spectrogram sequence, avoiding the sequential dependencies that bottlenecked previous approaches. The same year saw the release of HiFi-GAN, a generative adversarial network (GAN)-based vocoder that improved the efficiency of waveform generation while producing high-fidelity speech. In 2020, the release of Glow-TTS introduced a flow-based approach that allowed for fast inference and voice style transfer capabilities. In March 2020, the free text-to-speech website 15.ai was launched. 15.ai gained widespread international attention in early 2021 for its ability to synthesize emotionally expressive speech of fictional characters from popular media with minimal amount of data. The creator of 15.ai (known pseudonymously as 15) stated that 15 seconds of training data is sufficient to perfectly clone a person's voice (hence its name, "15.ai"), a significant reduction from the previously known data requirement of tens of hours. 15.ai is credited as the first platform to popularize AI voice cloning in memes and content creation. 15.ai used a multi-speaker model that enabled simultaneous training of multiple voices and emotions, implemented sentiment analysis using DeepMoji, and supported precise pronunciation control via ARPABET. The 15-second data efficiency benchmark was later corroborated by OpenAI in 2024. == Semi-supervised learning == Currently, self-supervised learning has gained much attention through better use of unlabelled data. Research has shown that, with the aid of self-supervised loss, the need for paired data decreases. == Zero-shot speaker adaptation == Zero-shot speaker adaptation is promising because a single model can generate speech with various speaker styles and characteristic. In June 2018, Google proposed to use pre-trained speaker verification models as speaker encoders to extract speaker embeddings. The speaker encoders then become part of the neural text-to-speech models, so that it can determine the style and characteristics of the output speech. This procedure has shown the community that it is possible to use only a single model to generate speech with multiple styles. == Neural vocoder == In deep learning-based speech synthesis, neural vocoders play an important role in generating high-quality speech from acoustic features. The WaveNet model proposed in 2016 achieves excellent performance on speech quality. Wavenet factorised the joint probability of a waveform x = { x 1 , . . . , x T } {\displaystyle \mathbf {x} =\{x_{1},...,x_{T}\}} as a product of conditional probabilities as follows p θ ( x ) = ∏ t = 1 T p ( x t | x 1 , . . . , x t − 1 ) {\displaystyle p_{\theta }(\mathbf {x} )=\prod _{t=1}^{T}p(x_{t}|x_{1},...,x_{t-1})} where θ {\displaystyle \theta } is the model parameter including many dilated convolution layers. Thus, each audio sample x t {\displaystyle x_{t}} is conditioned on the samples at all previous timesteps. However, the auto-regressive nature of WaveNet makes the inference process dramatically slow. To solve this problem, Parallel WaveNet was proposed. Parallel WaveNet is an inverse autoregressive flow-based model which is trained by knowledge distillation with a pre-trained teacher WaveNet model. Since such inverse autoregressive flow-based models are non-auto-regressive when performing inference, the inference speed is faster than real-time. Meanwhile, Nvidia proposed a flow-based WaveGlow model, which can also generate speech faster than real-time. However, despite the high inference speed, parallel WaveNet has the limitation of needing a pre-trained WaveNet model, so that WaveGlow takes many weeks to converge with limited computing devices. This issue has been solved by Parallel WaveGAN, which learns to produce speech through multi-resolution spectral loss and GAN learning strategies.
Excalidraw
Excalidraw is an open-source, web-based virtual whiteboard and diagramming application. It is used to create diagrams, wireframes, and sketches within a web browser without requiring account registration. The software features a characteristic hand-drawn visual style and supports real-time multi-user collaboration using client-side end-to-end encryption. Excalidraw is released under the MIT License and is maintained by Excalidraw s.r.o., a company based in Brno, Czech Republic. == History == Excalidraw was created on 1 January 2020 by Christopher Chedeau, a software engineer at Meta Platforms. Chedeau, who previously co-created React Native and Prettier, initially developed the application as a personal project before registering the domain on 3 January 2020. Within its first months, the project attracted open-source contributors who assisted in expanding its features and rewriting the codebase into TypeScript and React. By early 2021, day-to-day operations moved to Czech developers David Luzar and Milos Vetesnik. In May 2021, the team incorporated Excalidraw s.r.o. in Brno and launched a commercial cloud-based version named Excalidraw+ to fund the open-source project's development. By May 2026, the main open-source repository on GitHub had accumulated over 123,000 stars. == Features and architecture == The application provides an infinite canvas for geometric shapes, lines, arrows, text, and freehand drawing. Its visual presentation relies on Rough.js, a JavaScript graphics library that alters standard vector paths to mimic irregular, hand-drawn lines. Excalidraw operates as a Progressive web application (PWA), allowing local installation and offline usage, saving data natively to local browser storage. Files use a native, JSON-based extension format (.excalidraw), and canvases can be exported to PNG or SVG formats. Real-time collaboration sessions are executed using Socket.IO via a relay server. Data transmission uses the browser's native Web Cryptography API to achieve end-to-end encryption. A symmetric AES key is generated on the client side and appended to the sharing URL as a fragment identifier (following the # character). Because web browsers do not transmit URL fragments to HTTP servers, the data remains unreadable to the distribution server. == Ecosystem == Excalidraw is distributed as an npm package, allowing third-party developers to embed the whiteboard component directly into external React web applications. Community-developed extensions integrate the application's file format into text editors and note-taking systems, including Visual Studio Code and Obsidian. The platform also has native integrations in commercial platforms such as Notion and HackerRank. == Reception == Google's developer relations team published a technical case study on Excalidraw as a reference implementation for Progressive Web Apps. The analysis highlighted the software's adoption of advanced web platform capabilities, specifically its utilization of the File System Access API and native Clipboard API to replicate desktop software behavior within a web browser environment.
2023 Bilderberg Conference
The 2023 Bilderberg Conference or Bilderberg Club was held between May 18–21, 2023 at the Pestana Palace hotel in Lisbon, Portugal. The 2023 meeting was the 69th edition of the event. A Bilderberg Group press release stated that there were approximately 130 participants from 23 countries. Established in 1954 by Prince Bernhard of the Netherlands, Bilderberg conferences (or meetings) are an annual private gathering of the European and North American political and business elite. Events are attended by between 120 and 150 people each year invited by the Bilderberg Group's steering committee; including prominent politicians, CEOs, national security experts, academics and journalists. The 2023 conference received some media attention due to the participation of several major players in the artificial intelligence space, such as OpenAI CEO Sam Altman, Microsoft CEO Satya Nadella, Google DeepMind chief Demis Hassabis and former Google CEO Eric Schmidt. Bilderberg conferences operate under Chatham House Rule, meaning that participants are cannot disclose the identity or affiliation of any particular speaker. There were no press conferences during or after the event, as is customary. According to The Guardian, the paper's journalists were able to approach one high-ranking attendee, economist Victor Halberstadt, in a Lisbon pharmacy, but he denied his identity before jumping into a car and heading back to his hotel. == Agenda == The key topics for discussion at the 2023 Bilderberg Conference were announced on the Bilderberg website shortly before the meeting. These topics included: == Participants == A list of 128 participants was published on the Bilderberg website. This list may not be complete, as a source connected to the Bilderberg group told The Daily Telegraph in 2013 that some attendees do not have their names publicized. Oscar Stenström, Sweden’s chief negotiator for NATO membership, was reported to have been seen at the venue despite his name not being on the list.