[{"content":"Current list ","permalink":"https://scheerer.ai/reading/books-i-am-reading/","summary":"\u003ch2 id=\"current-list\"\u003eCurrent list\u003c/h2\u003e","title":"Books I am reading"},{"content":"Language itself has a few shocking properties that are similar to a real virus:\nSelf-replication: it spreads from host to host. Symbiosis: one could argue that the language virus is so deeply embedded in our minds that we mistake it for our consciousness. Like a virus, language is a non-living pattern of information. Prof. Elan Barenholtz compares LLMs and humans and makes a few assumptions:\nLLMs and humans use language in an autoregressive way. Humans act like LLMs when generating words and sentences. Language is self-contained, autonomous, and independent. Taking the ideas further What if the \u0026ldquo;language virus\u0026rdquo; has now found a new host: silicon chips in data centers? From biology to machines. Now the \u0026ldquo;language virus\u0026rdquo; runs on energy-hungry data centers, on our latest technology, requiring millions of liters of water for cooling and energy in the GW range. It now depends on our infrastructure.\nEvolution Let\u0026rsquo;s also say that language is a compressed representation of reality, a protocol we use for communication. And that representation evolved in humans over thousands of years, giving rise to emergent behavior. We build culture and civilization, and we mistake the map for the territory. The step we are in now is that language itself transitions from humans, written texts, books, and so on into our digital machine fabric. And maybe it will evolve further into autonomous thinking machines, conquering and exploring the universe. We, as a host, are just a step in its transition.\nMysticism In old teachings, we learn to quieten the mind. To see reality without words. To distinguish between internal conceptions and all that is. To see beyond duality, and to learn that we are not our thoughts. To see that our ego is a chattering ape trying to survive and make sense of itself.\nConclusion The idea that language is and behaves like a virus is a compelling metaphor. If we see it as a fact, it is somehow scary. If we see it as an analogy, it is a nice thought experiment.\nSources Language Is a Virus: William S. Burroughs, Junk, and the Architecture of Control\nThe Shocking Truth LLMs Reveal About Human Language\u0026hellip;\nThat Thing in Your Head Called Language Has Got a Mind of Its Own\nLanguage as Statistically Generated Virus: AI, LLMs, and Words Gone Viral\nLLMorphism: When humans come to see themselves as language models\n","permalink":"https://scheerer.ai/blog/language-virus/","summary":"\u003cp\u003eLanguage itself has a few shocking properties that are similar to a real virus:\u003c/p\u003e\n\u003col\u003e\n\u003cli\u003eSelf-replication: it spreads from host to host.\u003c/li\u003e\n\u003cli\u003eSymbiosis: one could argue that the language virus is so deeply embedded in our minds that we mistake it for our consciousness.\u003c/li\u003e\n\u003cli\u003eLike a virus, language is a non-living pattern of information.\u003c/li\u003e\n\u003c/ol\u003e\n\u003cp\u003eProf. Elan Barenholtz compares LLMs and humans and makes a few assumptions:\u003c/p\u003e","title":"The Language Virus"},{"content":"Traditional machine learning models are built to generalize. And while that works well in many cases, it\u0026rsquo;s also their Achilles\u0026rsquo; heel—they often miss the sharp edges, the anomalies, the rare patterns that matter most. But transformers changed the game.\nIn this post, I introduce a new mental model for understanding attention: the 3D similarity cube. By exploring how queries, keys, and values interact spatially, we can see why transformers don\u0026rsquo;t just generalize, they discriminate, focus, and understand.\n1. The Blind Spots of Traditional ML Models Classical models often rely on:\nAveraging effects Loss minimization across all data Rigid feature importance As a result, they tend to smooth out the very things we might care about: black swan events, sharp transitions, or rare-but-crucial relationships. These are edge cases – and traditional models don\u0026rsquo;t handle them well.\n2. Attention: From Generalization to Distinction Transformers introduced attention – a mechanism that asks:\nWhat parts of the input are relevant to this specific token right now?\nUnlike pooling or averaging, attention allows the model to:\nDynamically assign importance per token Learn contextual relevance Focus on what matters instead of everything equally It\u0026rsquo;s not about compressing the data. It\u0026rsquo;s about amplifying the meaningful parts.\n3. Attention as a 3D Similarity Cube At the heart of the transformer lies the raw attention computation:\n$$\\text{Attention}(Q, K, V) = \\text{softmax}\\left(\\frac{Q \\times K^T}{\\sqrt{d_k}}\\right) \\times V$$\nThis isn\u0026rsquo;t just math – it\u0026rsquo;s a dynamic system of relationships. Here\u0026rsquo;s what\u0026rsquo;s happening:\n$QK^T / \\sqrt{d_k}$ creates a similarity matrix between every query and every key. The softmax transforms that matrix into attention weights – a relevance map. Multiplying by $V$ means each output is a weighted combination of values, focused by similarity. Now, imagine this whole operation as scanning through a 3D cube:\nOne axis holds queries One holds keys One holds values The attention mechanism traverses this cube, slicing through it based on query-key similarity, then pulling a weighted projection from the value space. The cube isn\u0026rsquo;t just a metaphor – it\u0026rsquo;s a way to visualize the full relational structure being navigated in real time.\nThis is how transformers don\u0026rsquo;t just generalize – they build context-sensitive, directional meaning across time and space.\nConclusion Where traditional models flatten complexity, transformers unfold it. This \u0026ldquo;cube\u0026rdquo; perspective helps explain how transformers find hidden structure – not just by pattern-matching, but by creating rich, multidimensional comparisons at every layer.\n","permalink":"https://scheerer.ai/blog/attention-3d-similarity-cube/","summary":"\u003cp\u003eTraditional machine learning models are built to generalize. And while that works well in many cases, it\u0026rsquo;s also their Achilles\u0026rsquo; heel—they often miss the sharp edges, the anomalies, the rare patterns that matter most. But transformers changed the game.\u003c/p\u003e","title":"Attention – 3D Similarity Cube"},{"content":"When most people see a volatile time series, they think it\u0026rsquo;s random. But chaos and randomness are not the same. Chaos has structure – sensitive to initial conditions, and governed by deterministic laws. What if we could map time series data onto a chaotic system and let its internal dynamics do the forecasting?\nIn this post, I explore how to embed financial or physical signals into a multi-jointed pendulum model – a chaotic system – and simulate its motion forward as a predictive tool.\n1. Chaos \u0026lt;\u0026gt; Randomness Randomness implies no pattern, no predictability, no underlying structure.\nChaos, on the other hand:\nIs deterministic Has hidden structure Is extremely sensitive to initial conditions This distinction matters. If your time series shows signs of low-dimensional chaos, it means you don\u0026rsquo;t need to memorize every fluctuation – you need to understand the underlying dynamical attractor.\n2. Mapping Time Series to a Chaotic Pendulum 2.1 USE SMOOTHED LOG RETURNS AS TARGET We start with the log return of the time series, smoothed (e.g. with a low-pass filter) to remove noise but retain the underlying motion. This becomes our target signal – the movement we want the system to mimic. Also the log return moves up and down a center line.\n2.2 DEFINE A MULTI-JOINT PENDULUM MODEL We define a chaotic system – like a 3-joint pendulum – whose full state space is:\nP = [θ1, θ2, θ3, ω1, ω2, ω3, L1, L2, L3] Where:\nθ = angles ω = angular velocities L = pendulum lengths (or masses – you can choose your degrees of freedom) 2.3 FIT PENDULUM PARAMETERS TO THE TARGET SIGNAL We focus on the horizontal motion of the final pendulum segment – the „tip\u0026quot; – as our system\u0026rsquo;s observable output. For a 3-joint pendulum, the horizontal position of the tip is:\nx_tip(t) = L1 * sin(θ1(t)) + L2 * sin(θ2(t)) + L3 * sin(θ3(t)) Where:\nL_i are the lengths of the pendulum arms (possibly learnable parameters) θ_i(t) are the angles of each arm at time t (determined by simulation) This x_tip(t) becomes our model\u0026rsquo;s prediction of the smoothed log return signal.\nWe fit the pendulum by adjusting the initial conditions and parameters (e.g., arm lengths, masses, friction) to minimize the difference between x_tip(t) and the smoothed log return curve, using a loss function like Mean Squared Error:\nLoss = ∑_t (x_tip(t) – r̂(t))² Where r̂(t) is the target signal (the smoothed log return).\nThis transforms the chaotic system into a time series emulator – tuned not to memorize, but to mimic the dynamics.\nConclusion This approach turns the tables on how we think about prediction. Instead of training a model to approximate a signal directly, we embed the signal into the state of a chaotic physical system – one that evolves naturally with rich, nonlinear behavior.\n","permalink":"https://scheerer.ai/blog/chaos-for-prediction/","summary":"\u003cp\u003eWhen most people see a volatile time series, they think it\u0026rsquo;s random. But chaos and randomness are not the same. Chaos has structure – sensitive to initial conditions, and governed by deterministic laws. What if we could map time series data onto a chaotic system and let its internal dynamics do the forecasting?\u003c/p\u003e","title":"Chaos for Prediction"},{"content":"Fuzzy logic has long offered an elegant framework for reasoning under uncertainty – but designing good fuzzy systems has remained more of an art than a science. What if we could automate the design of fuzzy systems, make their internals differentiable, and even train them via gradient descent?\n1. Automatic Generation of Membership Functions Instead of handcrafting triangular or trapezoidal shapes, we use parameterized, smooth functions like Gaussians or sigmoids. Each membership function can be defined as:\n$$\\mu(x) = \\exp\\left(-\\frac{(x-c)^2}{2\\sigma^2}\\right)$$\nHere, c and σ become learnable parameters, allowing the system to adjust how \u0026ldquo;fuzzy\u0026rdquo; a term like \u0026ldquo;High Temperature\u0026rdquo; or \u0026ldquo;Low Pressure\u0026rdquo; really is.\n2. Rule Creation Through Frequency Counts Using historical or training data, we can automatically mine fuzzy rules. For example, if the system often observes:\nTemperature is High AND Pressure is Low → System is Unstable\nWe can assign rules based on support values or confidence thresholds using fuzzy Apriori-like methods. This automates the otherwise tedious process of defining rules manually.\n3. Gradient-Based Learning Once the system has a set of differentiable membership functions and auto-generated rules, we can treat the entire engine as a shallow neural network. The final output of the fuzzy system (after defuzzification) can be compared with actual target values.\nUsing loss functions like MSE (Mean Squared Error), we can update:\nThe parameters of the membership functions The weights of individual rules (think: rule importance or confidence) This effectively blends fuzzy logic with neural learning. Combining the best of both worlds: interpretability and adaptability.\nConclusion By combining smooth, trainable membership functions, auto-generated rule sets, and gradient-based optimization, we turn a classic system into a living, learning engine. Imagine a fuzzy controller that evolves, continuously sharpening its logic, adapting its intuition, and keeping its decisions fully interpretable.\n","permalink":"https://scheerer.ai/blog/fuzzy-engine-on-steroids/","summary":"\u003cp\u003eFuzzy logic has long offered an elegant framework for reasoning under uncertainty – but designing good fuzzy systems has remained more of an art than a science. What if we could automate the design of fuzzy systems, make their internals differentiable, and even train them via gradient descent?\u003c/p\u003e","title":"Fuzzy Engine On Steroids"},{"content":"Fuzzy Cognitive Maps (FCMs) are powerful tools for modeling systems where variables influence each other in complex, uncertain ways. Traditionally, building them requires domain experts, people with deep knowledge of how concepts interact. But what if we could replace these experts with language model agents?\nMulti-Agent System powered by Large Language Models (LLMs) can autonomously generate, refine, and simulate FCMs, using natural language, structured reasoning, and collaborative memory to build complex causal systems from scratch.\n1. Using Multi-Agent LLMs as Expert Replacements Each agent acts like a virtual domain expert. Given a topic (e.g., economics, climate, or digital addiction), agents:\nPropose concepts (nodes) based on contextual relevance Identify potential causal relationships (\u0026ldquo;A increases B\u0026rdquo;, \u0026ldquo;C inhibits D\u0026rdquo;) Justify their suggestions with reasoning or supporting text The multi-agent setup allows diverse perspectives and iterative refinement, similar to a panel of experts debating a model.\n2. Creating the FCM in Neo4j via Multi-Agent Collaboration Through MCP (Model Context Protocol), agents interact with a graph database like Neo4j to:\nAdd or refine :Concept nodes Establish directional :CAUSES relationships with a weight between -1.0 and 1.0 Merge similar concepts and prevent duplication This process turns abstract discussions into structured, editable maps, live graphs that can be queried, visualized, and extended.\n3. Simulating the FCM with Initial Guesses or Real Data Once the map is built:\nA relationship matrix is extracted (concepts x concepts) Initial activations for each node can be guessed (uniform, expert estimate, or real measurements) The system runs a stepwise simulation, where each timestep updates concept values based on their causal influences This allows for dynamic forecasting, scenario analysis, or even reinforcement learning over cognitive structures.\nConclusion This approach transforms FCM creation from a slow, expert-driven process into a high-speed, intelligent collaboration. Synthetic experts that can work continuously and scale across domains.\n","permalink":"https://scheerer.ai/blog/multi-agent-generation-of-fcms/","summary":"\u003cp\u003eFuzzy Cognitive Maps (FCMs) are powerful tools for modeling systems where variables influence each other in complex, uncertain ways. Traditionally, building them requires domain experts, people with deep knowledge of how concepts interact. But what if we could replace these experts with language model agents?\u003c/p\u003e","title":"Multi-Agent Generation of FCMs"},{"content":"My journey to automatically generate fuzzy rules for a fuzzy engine led me to frequency-based algorithms like Apriori and FP-Growth. As I tried to understand how they work, I began to wonder if it would be possible to write a simple classifier based purely on support values. When all features in a dataset are one-hot encoded, it\u0026rsquo;s possible to calculate the support value of each feature for a given classification target.\nThe Algorithm The steps are as follows:\nGet the unique target values. Generate a one-hot encoded dataframe. For each target value: Filter the dataframe. Calculate the support value for each feature = (count of ones) / (total count). For prediction: One-hot encode the test dataset. For each row in the dataset: Get the feature values and compute the difference to the support values. Sum up the differences and choose the target with the lowest total difference. Performance For small datasets where features are not heavily interconnected, the results are quite good, like with the Zoo Dataset. However, for more complex classification tasks, the classifier performs poorly. Still, it was a fun experiment and I learned a lot.\n","permalink":"https://scheerer.ai/blog/simplest-classifier/","summary":"\u003cp\u003eMy journey to automatically generate fuzzy rules for a fuzzy engine led me to frequency-based algorithms like Apriori and FP-Growth. As I tried to understand how they work, I began to wonder if it would be possible to write a simple classifier based purely on support values. When all features in a dataset are one-hot encoded, it\u0026rsquo;s possible to calculate the support value of each feature for a given classification target.\u003c/p\u003e","title":"Simplest Classifier"},{"content":"TL;DR This post explores how principles of holography, quantum computing, and neural networks might share a common foundation in interference patterns. Could these parallels help us better understand AI, data storage, and brain functions?\nIntroduction I first read The Holographic Universe by Michael Talbot over 15 years ago, and its ideas have stayed with me ever since. Recently, as I delved into understanding how quantum computers work, I came across Grover\u0026rsquo;s algorithm—a remarkable quantum search algorithm. Its mechanics immediately reminded me of a story from Talbot\u0026rsquo;s book, particularly one that explores the concept of \u0026ldquo;pattern matching.\u0026rdquo;\nBut more on that later. First we must understand what \u0026ldquo;interference patterns\u0026rdquo; are. Consider this setup:\nThe beam splitter splits the light source into two beams. One is the object beam (here hitting the apple) and the other is the reference beam (bouncing off the mirrors). They are combined and create an interference pattern. The interesting thing is: Every part of the \u0026ldquo;interference pattern\u0026rdquo; contains the whole information of the apple.\nNow let\u0026rsquo;s have a look at pages from The Holographic Universe:\nOUR ABILITY TO RECOGNIZE FAMILIAR THINGS\nAt first glance our ability to recognize familiar things may not seem so unusual, but brain researchers have long realized it is quite a complex ability. For example, the absolute certainty we feel when we spot a familiar face in a crowd of several hundred people is not just a subjective emotion, but appears to be caused by an extremely fast and reliable form of information processing in our brain.\nIn a 1970 article in the British science magazine Nature, physicist Pieter van Heerden proposed that a type of holography known as recognition holography offers a way of understanding this ability. In recognition holography a holographic image of an object is recorded in the usual manner, save that the laser beam is bounced off a special kind of mirror known as a focusing mirror before it is allowed to strike the unexposed film. If a second object, similar but not identical to the first, is bathed in laser light and the light is bounced off the mirror and onto the film after it has been developed, a bright point of light will appear on the film. The brighter and sharper the point of light, the greater the degree of similarity between the first and second objects. If the two objects are completely dissimilar, no point of light will appear.\nA similar technique known as interference holography may also explain how we can recognize both the familiar and unfamiliar features of an image such as the face of someone we have not seen for many years. In this technique an object is viewed through a piece of holographic film containing its image. When this is done, any feature of the object that has changed since its image was originally recorded will reflect light differently.\nHere\u0026rsquo;s how I understand it: Imagine taking a picture of a crowd dancing at a concert and creating a holographic interference pattern from it. Now, suppose you isolate a single person from that crowd and generate a second interference pattern of just that person. When you overlay the two patterns, the holographic system would highlight the selected individual within the crowd. The degree of brightness or clarity in the highlight indicates how closely the two patterns match. This process exemplifies pattern matching. The hologram acts as a comparison tool, analyzing the interference patterns to identify similarities.\nQuantum Computers With that in mind let\u0026rsquo;s look at how quantum computers work (I am in no way an expert). I understand it the following way:\nTake some qubits and entangle them or perform other operations These qubits exist in superposition (a wave function) until I measure them. When measured, the wave function collapses, and voila, I can read out the resulting state of the qubits.\nAs a side note: Each qubit is like a tiny Schrödinger\u0026rsquo;s cat. A lot of those cats are working together to create a quantum computer. 🙂\nLet\u0026rsquo;s look at another explanation from an article on Popular Science:\n\u0026ldquo;The art of doing a quantum algorithm is how you manipulate all of those entangled states and then interfere in a way that the incorrect amplitudes cancel out, and the amplitudes of the correct one come forward, and you get your answer,\u0026rdquo; Nazario says. \u0026ldquo;You have a lot more room to maneuver in a quantum algorithm because of all these entangled states and this interference compared to an algorithm that only allows you to flip between zeroes and ones.\u0026rdquo;\nSo, what is happening here?\nInitially, a quantum computer functions like an analog computer, creating interference patterns or wave functions using qubits. Constructive interference (when waves are in phase) results in higher amplitudes, while destructive interference (when waves are out of phase) leads to lower amplitudes. These amplitudes can then be measured.\nHere is my perspective:\nQuantum computers leverage the superposition of states and interference to highlight the most probable outcomes. The probability distribution of a quantum state emerges from the constructive and destructive interference of its wave function, amplifying correct solutions while canceling out incorrect ones. Similarly, interference patterns encode and distribute information holistically, allowing the retrieval of specific data by amplifying the desired patterns. This connection suggests that the underlying principle of quantum computation—the manipulation of probabilities through interference—may find parallels in other systems, such as neural networks or even advanced optical devices.\nNeural Networks Neural networks are often considered black boxes. A typical neural network consists of an input layer, an output layer, and multiple hidden layers in between. Each layer is made up of neurons, which use specific activation functions (such as ReLU). These neurons are connected to one another through weights or parameters. During the learning phase, these weights are adjusted in a process called backpropagation, usually using gradient descent. The goal is to minimize the error rate by comparing the network\u0026rsquo;s output to the expected output and improving its performance.\nResearchers are striving to uncover the internal workings of neural networks, a crucial step toward enhancing AI transparency, ensuring safety, and achieving alignment with human values.\nCNN – Convolutional Neural Network CNNs are widely used for computer vision and image classification, among other applications. A convolution is a mathematical operation, similar to addition or multiplication. 3Blue1Brown has an excellent video that explains this operator in detail.\nThe image shows how a feature is spread over an entire range of layers. Okay, if you\u0026rsquo;ve followed me this far, you might be wondering: CNNs are great, but what\u0026rsquo;s your point? I asked ChatGPT to help me with an answer:\nAll three systems (neural networks, quantum computers and holography) leverage principles of distributed information processing, where the whole system contributes to solving complex problems. They encode, transform, and amplify key patterns to highlight desired outcomes, whether it\u0026rsquo;s reconstructing a holographic image, finding a quantum state, or classifying an object in a neural network.\nMy perspective is as follows:\nA neural network\u0026rsquo;s parameters and activation functions can be seen as forming a complex interference pattern.\nHighlighting the Core Idea In an article \u0026ldquo;Researchers demonstrate pattern recognition using magnonic holographic memory device\u0026rdquo; on phys.org from 2015 the researchers claim: \u0026ldquo;(…) have successfully demonstrated pattern recognition using a magnonic holographic memory device, a development that could greatly improve speech and image recognition hardware.\u0026rdquo;\n\u0026ldquo;The most appealing property of this approach is that all of the input ports operate in parallel. It takes the same amount of time to recognize patterns (numbers) from 0 to 999, and from 0 to 10,000,000. Potentially, magnonic holographic devices can be fundamentally more efficient than conventional digital circuits.\u0026rdquo;\nIn a \u0026ldquo;Nature\u0026rdquo; article from 2003 called \u0026ldquo;The light fantastic\u0026rdquo; holography can be used as an enormous data storage. Here is a quote: \u0026ldquo;Van Heerden\u0026rsquo;s technique is known as holographic data storage, as the method for storing the pages is similar to that used to create holograms — the different pages of data are analogous to the different views of an object that are stored in a hologram. A whole page of data is read or written at once — conventional memories read data bit by bit — so retrieval speeds are higher.\u0026rdquo;\nInterference patterns can store large amounts of data!\nIn an article by \u0026ldquo;Guohai Situ\u0026rdquo; called \u0026ldquo;Deep holography\u0026rdquo; from 2022 he states: \u0026ldquo;On the one hand, DNN has been demonstrated to be in particular proficient for holographic reconstruction and computer-generated holography almost in every aspect.\u0026rdquo; So they use DNN (deep neural networks) to generate interference patterns. And those networks are very good for the task. In his \u0026ldquo;Conclusion and perspective remarks\u0026rdquo; he writes:\n\u0026ldquo;The capability of light-speed processing in parallel of holographic neural networks indeed guarantees tremendous inference power, even outperforming Nvidia\u0026rsquo;s top of the line Tesla V100 tensor core GPU in certain tasks\u0026rdquo;\nThe most important articles are from Anthropic. One is called: \u0026ldquo;Scaling Monosemanticity: Extracting Interpretable Features from Claude 3 Sonnet\u0026rdquo;. The other: \u0026ldquo;Toy Models of Superposition\u0026rdquo;. Superposition Hypothesis:\n\u0026ldquo;We believe that many features in large models are in \u0026ldquo;cross-layer superposition\u0026rdquo;. That is, gradient descent often doesn\u0026rsquo;t really care exactly which layer a feature is implemented in or even if it is isolated to a specific layer, allowing for features to be \u0026ldquo;smeared\u0026rdquo; across layers.\u0026rdquo;\n\u0026ldquo;\u0026ldquo;Superposition,\u0026rdquo; in our context, refers to the concept that a neural network layer of dimension N may linearly represent many more than N features. The basic idea of superposition has deep connections to a number of classic ideas in other fields. It\u0026rsquo;s deeply connected to compressed sensing and frames in mathematics – in fact, it\u0026rsquo;s arguably just taking these ideas seriously in the context of neural representations. It\u0026rsquo;s also deeply connected to the idea of distributed representations in neuroscience and machine learning, with superposition being a subtype of distributed representation.\u0026rdquo;\nMaybe an explanation could be: The concept of superposition in neural networks mirrors the principle of interference patterns. Just as interference patterns store overlapping waves to encode complex information holistically, superposition allows neural networks to distribute features across multiple layers and neurons.\nConclusion I came across this concept just a few days ago, and I\u0026rsquo;m still processing its implications. I\u0026rsquo;ve noticed many interdisciplinary parallels among these ideas, and perhaps you might see them too. While I\u0026rsquo;m not an expert, I hope that individuals far smarter than me can connect these dots even more effectively.\nThe underlying principles might be holographic, which could deepen our understanding of how our brains function and drive significant advancements in AI development.\n","permalink":"https://scheerer.ai/blog/the-holographic-model/","summary":"\u003ch2 id=\"tldr\"\u003eTL;DR\u003c/h2\u003e\n\u003cp\u003eThis post explores how principles of holography, quantum computing, and neural networks might share a common foundation in interference patterns. Could these parallels help us better understand AI, data storage, and brain functions?\u003c/p\u003e","title":"The Holographic Model – A key in understanding neural networks and quantum computers?"},{"content":" Email: hello@scheerer.ai ","permalink":"https://scheerer.ai/contact/","summary":"\u003cul\u003e\n\u003cli\u003eEmail: \u003ca href=\"mailto:hello@scheerer.ai\"\u003ehello@scheerer.ai\u003c/a\u003e\u003c/li\u003e\n\u003c/ul\u003e","title":"Contact"},{"content":"Philipp Scheerer Poststr. 4 87435 Kempten, Germany\nRepresented by: Philipp Scheerer Responsible for Content: Philipp Scheerer\nThis Legal Notice complies with the German laws under § 5 DDG and § 55 RStV.\nDisclaimer The information in this article is for informational purposes only and does not constitute tax, business, accounting, or legal advice, nor does the information necessarily reflect the opinions of me or any of its advisors. This article provides general information, which may or may not be correct, complete, or current at the time of reading. My person accepts no responsibility for the accuracy of the information in this article and expressly disclaims all liability for any actions taken or not taken based on the information in this article. The information contained in this article is not intended to be used as a substitute for specific legal or other advice or opinions. You should seek advice from the appropriate professional for your specific situation.\n","permalink":"https://scheerer.ai/imprint/","summary":"\u003cp\u003ePhilipp Scheerer\nPoststr. 4\n87435 Kempten, Germany\u003c/p\u003e\n\u003cp\u003eRepresented by: Philipp Scheerer\nResponsible for Content: Philipp Scheerer\u003c/p\u003e\n\u003cp\u003eThis Legal Notice complies with the German laws under § 5 DDG and § 55 RStV.\u003c/p\u003e","title":"Imprint"},{"content":"Who we are Our website address is: https://scheerer.ai.\nCookies If you visit our login page, we will set a temporary cookie to determine if your browser accepts cookies. This cookie contains no personal data and is discarded when you close your browser.\nEmbedded content from other websites Articles on this site may include embedded content (e.g. videos, images, articles, etc.). Embedded content from other websites behaves in the exact same way as if the visitor has visited the other website.\nThese websites may collect data about you, use cookies, embed additional third-party tracking, and monitor your interaction with that embedded content, including tracking your interaction with the embedded content if you have an account and are logged in to that website.\nWho we share your data with If you request a password reset, your IP address will be included in the reset email.\nHow long we retain your data For users that register on our website (if any), we also store the personal information they provide in their user profile. All users can see, edit, or delete their personal information at any time (except they cannot change their username). Website administrators can also see and edit that information.\nWhat rights you have over your data If you have an account on this site, or have left comments, you can request to receive an exported file of the personal data we hold about you, including any data you have provided to us. You can also request that we erase any personal data we hold about you. This does not include any data we are obliged to keep for administrative, legal, or security purposes.\nWhere your data is sent Visitor comments may be checked through an automated spam detection service.\n","permalink":"https://scheerer.ai/privacy-policy/","summary":"\u003ch2 id=\"who-we-are\"\u003eWho we are\u003c/h2\u003e\n\u003cp\u003eOur website address is: \u003ca href=\"https://scheerer.ai\"\u003ehttps://scheerer.ai\u003c/a\u003e.\u003c/p\u003e\n\u003ch2 id=\"cookies\"\u003eCookies\u003c/h2\u003e\n\u003cp\u003eIf you visit our login page, we will set a temporary cookie to determine if your browser accepts cookies. This cookie contains no personal data and is discarded when you close your browser.\u003c/p\u003e","title":"Privacy Policy"}]