Get the most recent info and news about Transtutors on HackerNoon, where 10k+ technologists publish stories for 4M+ monthly readers.

Transtutors caters to students pursuing different courses from different universities across Globe. We have a team of 100 + Qualified Academicians, Experts and Tutors who come together to help students with their problems to provide with Better Education Services and Leaning Opportunities Through its focused modules on Homework Help and preparation for examinations, our team has helped thousands of students across globe to achieve stupendous success in their school and college assignments course curriculum and exams. Transtutors has an added advantage of inducting Tutors who are in-demand worldwide along with access to the latest web-based technologies for online tutoring. We have trained Academicians who are well versed with the curriculum s of schools and colleges in US, UK & Australia. Transtutors is one of the fastest growing online tutoring companies. It is a unit of Transweb Global Inc. - a market leader comprising of a wide spectrum of educational services. Virtual Learning and Online Tutoring has been on the rise among students because of its multiple advantages. It is user-friendly and uncomplicated, cost effective and the most important part is that it can provide an eager student ready to learn or seeking help with instant connection to a tutor. It is the best podium for students and tutors to connect and solve problems related to assignments, questions, projects- anything and everything related to education!

Transtutors

transtutors

Meet the Winners of Startups in North and Central America

One of the hottest areas in technology today is machine learning. The rate of development from month to month is astonishing. Architectures are rapidly evolving, and can no longer keep pace. But what's even worse, Moore's Law itself is about to hit a wall. Moore's Law The essence of Moore's Law is that the number of transistors on integrated circuits approximately doubles every eighteen months. This trend has held for a long time, as technological advancements have enabled the production of ever-smaller and more efficient transistors. However, we've reached a physical limit where the laws of physics give us a reality check. Beyond a certain point, making smaller transistors is no longer possible. They don't work because electrons leak out of them. The only solution is to look for alternative architectures. One direction is the creation of specialized hardware, such as (Application-Specific Integrated Circuits). Recently, there was headline news about , a specialized processor designed specifically for large language models (LLMs). It is capable of running LLMs much faster and with much less energy consumption than traditional GPUs. Although this circuit is made with conventional technology, it is optimized for the task, which accounts for its significant speedup. Such specialized hardware can exploit the reserves in current technology, but the collision of Moore's Law with its limits also poses a constraint here. ASICs Groq's LPU The company Extropic came forward with a bolder idea. Their solution would banish transistors and replace them with something else. Like Google, the company was founded by two guys: and . Both worked in the field of quantum computing before founding the company, and their chip lies somewhere halfway between traditional integrated circuits and quantum computers. Guillaume Verdon Trevor McCourt As I mentioned before, the biggest limitation of traditional integrated circuits is that transistors cannot be miniaturized indefinitely. In very small domains, the rules of quantum mechanics significantly impact particles, making transistor operation impossible. In contrast, quantum computers leverage the effects of quantum mechanics. The greatest challenge with quantum computers is that qubits must be very well isolated from the environment. Even minimal noise can disrupt the system's operation. This is why quantum computers need to be cooled to absolute zero degrees to prevent the surrounding heat from interfering with their function. Extropic's brilliant idea is to, instead of fighting against environmental noise and heat, make use of them. This is where the term " " comes from. Thermodynamic computing In the field of machine learning, random noise is used extensively. One example of this is , which is also used by the Midjourney image generator. Here, a bit of noise is added to the image at every step until the entire image becomes noise. The neural network is then trained to reverse this process, that is, to generate an image from noise based on the given text. If the system is trained with enough images and text, it will be capable of generating images from random noise based on text. This is how Midjourney operates. normal distribution stable diffusion In fact, the use of noise is even more widespread. In any neural network, the initial weights are usually initialized with random noise. During training, shapes the final weights from this random noise. Thus, a similar process occurs in every neural network as in stable diffusion: the final model emerges from noise through the modification of weights during training. backpropagation On current computer architectures, this random distribution can only be emulated with circuits made of deterministic transistors. In contrast, Extropic's circuit does this in an analog manner. The starting state is completely random, normally distributed thermal noise. Through programming the circuit, this noise can be modified within each component. Instead of transistors, analog weights take their place, which are noisy, but the outcome can be determined through statistical analysis of the output. The guys call this . probabilistic computing These analog circuits are much faster and consume much less energy, and since the thermal noise is not only non-disruptive but actually an essential component of the operation, they do not require the special conditions needed by quantum computers. The chips can be manufactured with existing production technology, so they could enter the commercial market within a few years. As we have seen from the above, Extropic's technology is very promising. However, what personally piqued my interest is that it is more biologically plausible. Of course, I don't think that the neurons in artificial neural networks have anything to do with human brain neurons. These are two very different systems. However, it is clear that the human brain does not learn through . Biological learning is something entirely different, and randomness certainly plays a significant role in it. gradient descent In fact, in biology and nature, everything operates randomly. What we see as deterministic at a high level is actually just what statistically stands out from many random events. This is how, for example, many living beings (including us humans) came to be through completely random evolution yet are built with almost engineering precision. I suspect that the human brain operates in a similar way to evolution. A multitude of random events within a suitably directed system, which we perceive from the outside as consistent thinking. This is why , and now I see the same principle in Extropic's chip. genetic algorithms were so intriguing to me If you are interested, check the or with the founder guys. company homepage this interview

Thermodynamic Computing: How to Save Machine Learning by Replacing Transistors

Psychology of the Unconscious by C. G. Jung, is part of the HackerNoon Books Series. You can jump to any chapter in this book . TRANSLATOR’S NOTE here TRANSLATOR’S NOTE That humanity is seeking a new message, a new light upon the meaning of life, and something tangible, as it were, with which it can work towards a larger understanding of itself and its relation to the universe, is a fact I think none will gainsay. Therefore, it has seemed to me particularly timely to introduce to the English-speaking world Dr. Jung’s remarkable book, “Wandlungen und Symbole der Libido.” In this work he has plunged boldly into the treacherous sea of mythology and folklore, the productions of the ancient mind and that of the common people, and turned upon this vast material the same scientific and painstaking method of psychologic analysis that is applied to the modern mind, in order to reveal the common bond of desire and longing which unites all humanity, and thus bridge the gaps presumed to exist between ancient and widely separated peoples and those of our modern time. The discovery of this undercurrent affecting and influencing ancient peoples as well as modern serves as a foundation or platform from which he proceeds to hold aloft a new ideal, a new goal of attainment possible of achievement and which can be intellectually satisfying, as well as emotionally appealing: the goal of . moral autonomy This book, remarkable for its erudition and the tremendous labor expended upon it, as well as for the new vilight which it sheds upon human life, its motives, its needs and its possibilities, is not one for desultory reading or superficial examination. Such an approach will prevent the reader from gaining anything of its real value; but for those who can bring a serious interest and willingness to give a careful study to it the work will prove to be a veritable mine capable of yielding the greatest riches. The difficulties in translating a book such as this are almost insuperable, but I have tried faithfully to express Dr. Jung’s thought, keeping as close to the original text as possible and, at the same time, rendering the difficult material and complicated German phrasing as simply and clearly as the subject-matter would allow. In all this work I owe much to Miss Helen I. Brayton, without whose faithful assistance the work would never have been completed. I wish to acknowledge my gratitude to Mr. Louis Untermeyer, whose help in rendering the poetic quotations into English verse has been invaluable, and to express as well my gratitude to other friends who have assisted me in various ways from time to time. B. M. H. New York, 1915. About HackerNoon Book Series: We bring you the most important technical, scientific, and insightful public domain books. This book is part of the public domain. C. G. Jung (2021). Psychology of the Unconscious. Urbana, Illinois: Project Gutenberg. Retrieved https://www.gutenberg.org/cache/epub/65903/pg65903-images.html This eBook is for the use of anyone anywhere at no cost and with almost no restrictions whatsoever. You may copy it, give it away or re-use it under the terms of the Project Gutenberg License included with this eBook or online at , located at . www.gutenberg.org https://www.gutenberg.org/policy/license.html

TRANSLATOR’S NOTE

is part of HackerNoon’s Book Blog Post series. You can jump to any chapter in this book here: . Translator's Preface. Crime and Punishment, by Fyodor Dostoyevsky [LINK TO TABLE OF LINK] TRANSLATOR’S PREFACE About HackerNoon Book Series: We bring you the most important technical, scientific, and insightful public domain books. A few words about Dostoevsky himself may help the English reader to understand his work. Dostoevsky was the son of a doctor. His parents were very hard-working and deeply religious people, but so poor that they lived with their five children in only two rooms. The father and mother spent their evenings in reading aloud to their children, generally from books of a serious character. Though always sickly and delicate Dostoevsky came out third in the final examination of the Petersburg school of Engineering. There he had already begun his first work, “Poor Folk.” This story was published by the poet Nekrassov in his review and was received with acclamations. The shy, unknown youth found himself instantly something of a celebrity. A brilliant and successful career seemed to open before him, but those hopes were soon dashed. In 1849 he was arrested. Though neither by temperament nor conviction a revolutionist, Dostoevsky was one of a little group of young men who met together to read Fourier and Proudhon. He was accused of “taking part in conversations against the censorship, of reading a letter from Byelinsky to Gogol, and of knowing of the intention to set up a printing press.” Under Nicholas I. (that “stern and just man,” as Maurice Baring calls him) this was enough, and he was condemned to death. After eight months’ imprisonment he was with twenty-one others taken out to the Semyonovsky Square to be shot. Writing to his brother Mihail, Dostoevsky says: “They snapped words over our heads, and they made us put on the white shirts worn by persons condemned to death. Thereupon we were bound in threes to stakes, to suffer execution. Being the third in the row, I concluded I had only a few minutes of life before me. I thought of you and your dear ones and I contrived to kiss Plestcheiev and Dourov, who were next to me, and to bid them farewell. Suddenly the troops beat a tattoo, we were unbound, brought back upon the scaffold, and informed that his Majesty had spared us our lives.” The sentence was commuted to hard labour. One of the prisoners, Grigoryev, went mad as soon as he was untied, and never regained his sanity. The intense suffering of this experience left a lasting stamp on Dostoevsky’s mind. Though his religious temper led him in the end to accept every suffering with resignation and to regard it as a blessing in his own case, he constantly recurs to the subject in his writings. He describes the awful agony of the condemned man and insists on the cruelty of inflicting such torture. Then followed four years of penal servitude, spent in the company of common criminals in Siberia, where he began the “Dead House,” and some years of service in a disciplinary battalion. He had shown signs of some obscure nervous disease before his arrest and this now developed into violent attacks of epilepsy, from which he suffered for the rest of his life. The fits occurred three or four times a year and were more frequent in periods of great strain. In 1859 he was allowed to return to Russia. He started a journal—“Vremya,” which was forbidden by the Censorship through a misunderstanding. In 1864 he lost his first wife and his brother Mihail. He was in terrible poverty, yet he took upon himself the payment of his brother’s debts. He started another journal—“The Epoch,” which within a few months was also prohibited. He was weighed down by debt, his brother’s family was dependent on him, he was forced to write at heart-breaking speed, and is said never to have corrected his work. The later years of his life were much softened by the tenderness and devotion of his second wife. In June 1880 he made his famous speech at the unveiling of the monument to Pushkin in Moscow and he was received with extraordinary demonstrations of love and honour. A few months later Dostoevsky died. He was followed to the grave by a vast multitude of mourners, who “gave the hapless man the funeral of a king.” He is still probably the most widely read writer in Russia. In the words of a Russian critic, who seeks to explain the feeling inspired by Dostoevsky: “He was one of ourselves, a man of our blood and our bone, but one who has suffered and has seen so much more deeply than we have his insight impresses us as wisdom... that wisdom of the heart which we seek that we may learn from it how to live. All his other gifts came to him from nature, this he won for himself and through it he became great.” About HackerNoon Book Series: We bring you the most important technical, scientific, and insightful public domain books. This book is part of the public domain. Urbana, Illinois: Project Gutenberg. Retrieved April 2022, from Fyodor Dostoyevsky. 2001. CRIME AND PUNISHMENT. https://www.gutenberg.org/files/2554/2554-h/2554-h.htm#link2HCH0016 This eBook is for the use of anyone anywhere at no cost and with almost no restrictions whatsoever.  You may copy it, give it away or re-use it under the terms of the Project Gutenberg License included with this eBook or online at , located at . www.gutenberg.org https://www.gutenberg.org/policy/license.html

Crime and Punishment: Translator's Preface

The Interpretation of Dreams by Sigmund Freud, is part of the HackerNoon Books Series. You can jump to any chapter in this book here . TRANSLATOR’S PREFACE TRANSLATOR’S PREFACE Since the appearance of the author’s Selected Papers on Hysteria and other Psychoneuroses, and Three Contributions to the Sexual Theory, much has been said and written about Freud’s works. Some of our readers have made an honest endeavour to test and utilise the author’s theories, but they have been handicapped by their inability to read fluently very difficult German, for only two of Freud’s works have hitherto been accessible to English readers. For them this work will be of invaluable assistance. To be sure, numerous articles on the Freudian psychology have of late made their appearance in our literature; but these scattered papers, read by those unacquainted with the original work, often serve to confuse rather than enlighten. For Freud cannot be mastered from the reading of a few pamphlets, or even one or two of his original works. Let me repeat what I have so often said: No one is really qualified to use or to judge Freud’s psychoanalytic method who has not thoroughly mastered his theory of the neuroses—The Interpretation of Dreams, Three Contributions to the Sexual Theory, The Psychopathology of Everyday Life, and Wit and its Relation to the Unconscious, and who has not had considerable experience in analysing the dreams and psychopathological actions of himself and others. That there is required also a thorough training in normal and abnormal psychology goes without saying. The Interpretation of Dreams is the author’s greatest and most important work; it is here that he develops his psychoanalytic technique, a thorough knowledge of which is absolutely indispensable for every worker in this field. The difficult task of making a translation of this work has, therefore, been xiiundertaken primarily for the purpose of assisting those who are actively engaged in treating patients by Freud’s psychoanalytic method. Considered apart from its practical aim, the book presents much that is of interest to the psychologist and the general reader. For, notwithstanding the fact that dreams have of late years been the subject of investigation at the hands of many competent observers, only few have contributed anything tangible towards their solution; it was Freud who divested the dream of its mystery, and solved its riddles. He not only showed us that the dream is full of meaning, but amply demonstrated that it is intimately connected with normal and abnormal mental life. It is in the treatment of the abnormal mental states that we must recognise the most important value of dream interpretation. The dream does not only reveal to us the cryptic mechanisms of hallucinations, delusions, phobias, obsessions, and other psychopathological conditions, but it is also the most potent instrument in the removal of these. I take this opportunity of expressing my indebtedness to Professor F. C. Prescott for reading the manuscript and for helping me overcome the almost insurmountable difficulties in the translation. A. A. BRILL. New York City. About HackerNoon Book Series: We bring you the most important technical, scientific, and insightful public domain books. This book is part of the public domain. Sigmund Freud (2021). The Interpretation of Dreams. Urbana, Illinois: Project Gutenberg. Retrieved October 2022 https://www.gutenberg.org/cache/epub/66048/pg66048-images.html This eBook is for the use of anyone anywhere at no cost and with almost no restrictions whatsoever. You may copy it, give it away or re-use it under the terms of the Project Gutenberg License included with this eBook or online at www.gutenberg.org , located at https://www.gutenberg.org/policy/license.html.

TRANSLATOR’S PREFACE

Totem and Taboo by Sigmund Freud, is part of the HackerNoon Books Series. You can jump to any chapter in this book here . TRANSLATOR’S INTRODUCTION TRANSLATOR’S INTRODUCTION When one reviews the history of psychoanalysis one finds that it had its inception in the study of morbid mental states. Beginning with the observation of hysteria and the other neuroses Professor Freud gradually extended his investigations to normal psychology and evolved new concepts and new methods of study. The neurotic symptoms were no longer imaginary troubles the nature of which one could not grasp, but were conceived as mental and emotional maladjustments to one’s environment. The stamp of degeneracy impressed upon neurotics by other schools of medicine was altogether eradicated. Deeper investigation showed conclusively that a person might become neurotic if subjected to certain environments, and that there was no definite dividing line between normal and abnormal. The hysterical symptoms, obsessions, doubts, phobias, as well as hallucinations of the insane, show the same mechanisms as those similar psychic structures which one constantly encounters in normal persons in the form of mistakes in talking, reading, writing, forgettin, dreams and wit. The dream, always highly valued by the populace, and as much despised by the educated classes, has a definite structure and meaning when subjected to analysis. Professor Freud’s monumental work, The Interpretation of Dreams , marked a new epoch in the history of mental science. One might use the same words in reference to his profound analysis of wit Faulty psychic actions, dreams and wit are products of the unconscious mental activity, and like neurotic or psychotic manifestations represent efforts at adjustment to one’s environment. The slip of the tongue shows that on account of unconscious inhibitions the individual concerned is unable to express his true thoughts; the dream is a distorted or plain expression of those wishes which are prohibited in the waking states, and the witticism, owing to its veiled or indirect way of expression, enables the individual to obtain pleasure from forbidden sources. But whereas dreams, witticisms, and faulty actions give evidences of inner conflicts which the individual overcomes, the neurotic or psychotic symptom is the result of a failure and represents a morbid adjustment. The aforementioned psychic formations are therefore nothing but manifestations of the struggle with reality, the constant effort to adjust one’s primitive feelings to the demands of civilization. In spite of all later development the individual retains all his infantile psychic structures. Nothing is lost; the infantile wishes and primitive impulses can always be demonstrated in the grown-up and on occasion can be brought back to the surface. In his dreams the normal person is constantly reviving his childhood, and the neurotic or psychotic individual merges back into a sort of psychic infantilism through his morbid productions. The unconscious mental activity which is made up of repressed infantile material for ever tries to express itself. Whenever the individual finds it impossible to dominate the difficulties of the world of reality there is a regression to the infantile, and psychic disturbances ensue which are conceived as peculiar thoughts and acts. Thus the civilized adult is the result of his childhood or the sum total of his early impressions; psychoanalysis thus confirms the old saying: The child is father to the man. It is at this point in the development of psychoanalysis that the paths gradually broadened until they finally culminated in this work. There were many indications that the childhood of the individual showed a marked resemblance to the primitive history or the childhood of races. The knowledge gained from dream analysis and phantasies , when applied to the productions of racial phantasies, like myths and fairy tales, seemed to indicate that the first impulse to form myths was due to the same emotional strivings which produced dreams, fancies and symptoms . Further study in this direction has thrown much light on our great cultural institutions, such as religion, morality, law and philosophy, all of which Professor Freud has modestly formulated in this volume and thus initiated a new epoch in the study of racial psychology. [7] [8] I take great pleasure in acknowledging my indebtedness to Mr Alfred B. Kuttner for the invaluable assistance he rendered in the translation of this work. A. A. BRILL. About HackerNoon Book Series: We bring you the most important technical, scientific, and insightful public domain books. This book is part of the public domain. Sigmund Freud (2021). Totem and Taboo. Urbana, Illinois: Project Gutenberg. Retrieved October 2022 https://www.gutenberg.org/cache/epub/41214/pg41214-images.html This eBook is for the use of anyone anywhere at no cost and with almost no restrictions whatsoever. You may copy it, give it away or re-use it under the terms of the Project Gutenberg License included with this eBook or online at www.gutenberg.org , located at https://www.gutenberg.org/policy/license.html.

TRANSLATOR’S INTRODUCTION

Low Resource Languages (LRLs) in Large Language Models (LLMs) In recent years, the emergence of Large Language Models (LLMs) has brought about significant shifts in the daily routines of consumers. Individuals can now undertake a diverse range of tasks, such as retrieving information, composing text, and refining documents through these powerful language tools. This integration of LLMs into daily life has resulted in notable boosts in productivity, both at work and in personal endeavors. However, it’s important to recognize that not all consumers have experienced these benefits equally. Indeed, a considerable number of people around the world who speak less common languages aren’t able to interact with LLMs, primarily due to the inadequacy of language models designed for these specific languages. With 7,000 languages currently spoken in the world, the largest multilingual LLMs have been trained using only fewer than a hundred languages, thus leaving many languages and people completely behind. Supporting non-English languages requires high-quality, abundant data sources, which can be difficult to find and access. And not only do those models perform worse but it also has been reported by that they are more likely to give unethical responses thus making them more vulnerable to malicious attacks. Brown University Why do we have underrepresented languages in LLMs? The performance of LLMs tailored for Low Resource Languages (LRL) is hindered by several key challenges. Firstly, the foundation models for many LLMs rely on data scraped from the internet, which often lacks comprehensive coverage of LRLs. The below graph shows a distribution of data across the internet divided into language groups. While more common languages have hundreds of GB of data potentially available for training models, the languages in the tail of the graph only have data available in the range of hundreds of megabytes. This limitation is further magnified by the absence of fine-tuned instruction datasets for many LRLs. An instruction dataset consists of a question set paired with ideal answers and is a crucial part of LLM training - in this case, in specific languages. This is how the model learns to follow instructions, and without this asset, models are only capable of predicting the next word in the sequence instead of assisting humans with complex questions and problem-solving tasks. The above is caused by the fact that LLMs are trained in sequential steps. The first step is to learn the language by reading a large amount of unannotated text which gives the model the ability to predict the next world in the sequence. The second step is tailoring this predictive behavior to follow specific instructions, such as answering questions, writing summaries, or extracting data. This is why fine-tuning datasets is of such importance, as their quality will further determine the ability of LLM to assist users with required tasks. In the following section, we will present a method to create a high-quality dataset for Swahili that can be used to fine-tune the LLM for this language. The method can be applied to any low-resource language. Innovative pipeline to gather data for LRLs Swahili is a language spoken by over 200 million people across 14 different African countries and is the official national language in Tanzania, Kenya, Uganda, and the Democratic Republic of the Congo. It belongs to the group of low-resource languages and is an example of a language that does not have an out-of-the-box instruction dataset for LLM fine-tuning. In general, three approaches exist to create a fine-tuning dataset for a language. The first one is the direct generation of a dataset by assessors, in this case, language experts, which requires developing both questions and ideal answers in the desired language. This can be challenging for the Swahili language because assessors need to be high-level experts and the process is generally expensive. Another potential solution is taking an existing instruction dataset in English and translating it into Swahili. This could be done by translators who speak both Swahili and English but this can also be time and resource intensive. An automatic translator could be used, however, this typically results in insufficient or poor-quality results. Another solution combines automated translation with human validation, offering a cost-efficient and scalable approach, which is critical to ensure LRL models are accurate, reflect local customs and norms, and are useful to the communities that will be using them. This method utilizes the best available automatic translator from Swahili to English and then asks native Swahili speakers to filter out examples that do not meet quality standards. recently undertook a development project, where they created an 11,000 fine-tuning dataset for Swahili from the 15,000 original . Each data point consisting of a prompt and an answer was translated from English to Swahili using automatic translation resulting initially in 15,000 question answers pairs in Swahili. This dataset was further reduced by asking native speakers to remove pairs with low quality thus leaving a fine-tuned Swahili dataset with 11,000 instances. Toloka Dolly dataset The dataset was then used to improve , one of the top-performing multilingual language models for Swahili, which demonstrated significant performance enhancements for this language. The fine-tuned dataset boosted accuracy and f-score (a measure of predictive performance) for classification tasks, but more importantly, it significantly increased Recall-Oriented Understudy for Gisting Evaluation, which is a set of metrics used for evaluating automatic summarization and machine translation software in NLP, and in generative tasks where the model must respond to open questions. This experiment shows the potential for improving LLM performance in LRLs and therefore opens up a path for building truly multilingual models. mT5 ROUGE , or chrF++, Character n-gram F-score (chrF), Creating a More Inclusive AI Ecosystem As developers and organizations strive to create a more inclusive AI ecosystem, evaluation becomes even more critical, as does human involvement in training LLMs. Cohere’s recent launch of , a language model that supports over one hundred languages, including Swahili and other LRLs, exemplifies this commitment. Addressing data scarcity and enhancing model performance for LRLs is an important step to building more inclusive and responsible AI systems that serve diverse linguistic communities worldwide. Aya

Why Integrating Low Resource Languages Into LLMs Is Essential for Responsible AI

দায়ী এআই-এর জন্য কেন LLM-এ নিম্ন সম্পদের ভাষাগুলিকে একীভূত করা অপরিহার্য

Warum die Integration ressourcenarmer Sprachen in LL.M. für eine verantwortungsvolle KI unerlässlich ist

Por qué la integración de idiomas de bajos recursos en los LLM es esencial para una IA responsable

Pourquoi l'intégration de langages à faibles ressources dans les LLM est essentielle pour une IA responsable

जिम्मेदार AI के लिए LLM में कम संसाधन वाली भाषाओं को एकीकृत करना क्यों आवश्यक है

責任ある AI を実現するために、リソースの少ない言語を LLM に統合することが不可欠な理由

책임 있는 AI를 위해 저자원 언어를 LLM에 통합하는 것이 필수적인 이유

Por que a integração de linguagens de poucos recursos em LLMs é essencial para uma IA responsável

Почему интеграция малоресурсных языков в программы LLM важна для ответственного ИИ

Sorumlu Yapay Zeka için Düşük Kaynaklı Dilleri Yüksek Lisans Programlarına Entegre Etmek Neden Önemlidir?

Tại sao việc tích hợp các ngôn ngữ tài nguyên thấp vào LLM là điều cần thiết cho AI có trách nhiệm

为什么将低资源语言集成到法学硕士 (LLM) 课程对于负责任的 AI 至关重要

A Digestible High-Level Overview of CPU & GPU Cores

CPU এবং GPU ক্ষমতাগুলির একটি হজমযোগ্য উচ্চ-স্তরের ওভারভিউ

Ein leicht verständlicher Überblick über die CPU- und GPU-Funktionen

Una descripción general digerible de alto nivel de las capacidades de CPU y GPU

Un aperçu digeste de haut niveau des capacités du CPU et du GPU

CPU और GPU क्षमताओं का एक सुपाच्य उच्च-स्तरीय अवलोकन

CPU と GPU の機能について、わかりやすい概要を説明します。

CPU 및 GPU 기능에 대한 소화 가능한 상위 수준 개요

Uma visão geral resumida de alto nível dos recursos de CPU e GPU

Удобный общий обзор возможностей процессора и графического процессора

CPU ve GPU Yeteneklerine Sindirilebilir Üst Düzey Genel Bakış

Tổng quan cấp cao dễ hiểu về khả năng của CPU và GPU

CPU 和 GPU 功能的易懂高级概述

Nvidia's annual developer conference in 2024, known as GTC (GPU Technology Conference), was a pivotal event that unveiled groundbreaking advancements in artificial intelligence (AI) chips, software tools, and strategic partnerships. The conference, led by Nvidia's CEO Jensen Huang, showcased the company's commitment to innovation and leadership in the tech industry. And if you want to understand the latest trends from the world of AI - subscribe to my newsletter ‘ ‘. And the best part - it’s completely free! AI Hunters Now, onto the key takeaways from Nvidia's GTC 2024: Flagship AI Chip - B200 "Blackwell" The Flagship AI Chip - B200 "Blackwell" unveiled by Nvidia at GTC 2024 represents a groundbreaking advancement in AI chip technology. Named after mathematician David Blackwell, this new chip is a significant leap forward in performance, being 30 times faster than its predecessor, the Hopper series, particularly excelling at tasks like supporting advanced chatbots and large language models. With an impressive 208 billion transistors, more than double the previous chip, the B200 chip offers exceptional processing power. Despite its remarkable performance, the B200 chip is designed with efficiency in mind, using 25 times less energy and money than the previous Hopper series. This focus on efficiency not only enhances its cost-effectiveness but also positions it as a sustainable solution for AI applications. The innovative design of the B200 chip, which combines two squares of silicon into a single chip, allows all transistors to access memory simultaneously, further boosting productivity and performance. Nvidia's dominance in the data center AI chip market, with a market share of roughly 80%, underscores the significance of the B200 chip in shaping the future of AI technology. Major tech giants like Amazon, Google, Meta Platforms, Microsoft, OpenAI, and Tesla are expected to adopt this new chip, highlighting its importance and impact in the industry. Expected to hit the market later in 2024, the B200 chip offers advanced capabilities for AI applications. With its powerful performance, energy efficiency, and anticipated industry-wide adoption, the Blackwell B200 chip is poised to redefine AI processing and drive innovation across various sectors. Points to Reflect Upon: Performance Leap: Consider the implications of the B200 chip's 30x performance increase for AI applications and industries. Assess the feasibility of integrating the B200 chip into existing AI infrastructure for enhanced performance. Efficiency Benefits: Conduct a cost-benefit analysis to determine the potential savings from transitioning to the B200 chip. Software Tools for AI Integration Nvidia Inference Microservices (NIM) introduced at GTC 2024 represent a significant advancement in AI deployment technology. NIM is a cloud-native microservices solution that simplifies the development and deployment of generative AI applications across various infrastructures, including cloud, data centers, and GPU-accelerated workstations. It offers a streamlined path for developing AI-powered enterprise applications and deploying AI models in production, aiming to shorten time-to-market and simplify deployment processes for organizations transitioning to full-scale production deployments. NIM is designed to bridge the gap between the complexities of AI model development and the operational needs of enterprises, enabling a broader pool of developers to contribute to AI transformations within their companies. Some core benefits of NIM include its portability and control, allowing model deployment across various infrastructures, from local workstations to cloud environments. It provides prebuilt containers and Helm charts with optimized models validated across different NVIDIA hardware platforms, cloud service providers, and Kubernetes distributions, ensuring support across all NVIDIA-powered environments. Developers can access AI models through industry-standard APIs, simplifying the development of AI applications and enabling swift updates within the ecosystem. NIM also addresses the need for domain-specific solutions by packaging domain-specific NVIDIA CUDA libraries and specialized code tailored to various domains like language, speech, video processing, healthcare, and more. Leveraging optimized inference engines for each model and hardware setup, NIM ensures the best possible latency and throughput on accelerated infrastructure, reducing the cost of running inference workloads as they scale and enhancing the end-user experience. Part of the NVIDIA AI Enterprise suite, NIM is built with an enterprise-grade base container, providing a solid foundation for enterprise AI software through rigorous validation, enterprise support, and regular security updates. With support for various AI models, including community models, NVIDIA AI Foundation models, and custom AI models, NIM simplifies the AI model deployment process by packaging algorithmic, system, and runtime optimizations and adding industry-standard APIs. This comprehensive solution accelerates the deployment of scalable and customized AI applications in production, making AI more accessible and customizable across industries. Points to Reflect Upon: Business Impact: Consider how NIM can empower businesses to harness AI capabilities and drive innovation. Operational Efficiency: Evaluate how enhanced system efficiency can optimize business processes and decision-making. Strategic Partnerships and Collaborations Nvidia's expanded partnerships with industry leaders and software makers announced at GTC 2024 mark a significant milestone in the development of AI technologies and their integration into various sectors. These collaborations highlight Nvidia's strategic efforts to advance AI infrastructure, enhance AI capabilities, and drive innovation across industries. One key partnership involves Microsoft, where Nvidia and Microsoft are working together to bring the power of Nvidia Grace Blackwell GB200 and advanced Nvidia technologies to Azure. This collaboration aims to deliver trillion-parameter foundation models for tasks like natural language processing, computer vision, speech recognition, and more. Additionally, Microsoft is introducing the Azure NC H100 v5 VM virtual machine based on the Nvidia H100 NVL platform to cater to midrange training and inferencing needs. Furthermore, Nvidia's inference microservices are being integrated into Azure, enhancing AI capabilities for users by providing cloud-native microservices for optimized inference on various foundation models. This integration streamlines AI deployments and accelerates the development of performance-optimized production AI applications. In the automotive sector, Nvidia has deepened its ties with leading Chinese automakers like BYD, Xpeng, GAC Aion's Hyper brand, Zeekr, and Li Auto. These partnerships focus on integrating Nvidia's cutting-edge technologies, such as the Drive Thor in-vehicle chips, to enhance autonomous driving capabilities and infotainment systems in vehicles. Collaborations with US-based software company Cerence aim to adapt large language model AI systems for in-car applications, promising advanced voice command systems and improved user experiences. Overall, these partnerships unveiled at GTC 2024 demonstrate Nvidia's commitment to driving innovation across various industries by leveraging cutting-edge technologies and fostering collaborations with industry leaders and software makers. Points to Reflect Upon: Industry Impact: Consider how partnerships with key players can influence the adoption and integration of Nvidia's technology in various sectors. Innovation Potential: Reflect on the potential for collaborative efforts to drive innovation and create new opportunities in AI and technology. Focus on Sustainability and Sovereign AI Nvidia's emphasis on sustainability and sovereign AI initiatives at GTC 2024 underscores the company's commitment to responsible technology development. The focus on addressing environmental concerns related to AI training and data center energy consumption demonstrates Nvidia's proactive stance on sustainability issues. By prioritizing energy efficiency and sustainable practices in AI development, Nvidia is setting a standard for responsible innovation in the tech industry. Furthermore, Nvidia's dedication to sovereign AI initiatives, which enable nations to produce AI using their infrastructure and workforce, highlights the company's commitment to promoting ethical and secure AI practices globally. By empowering countries to develop AI capabilities locally, Nvidia is fostering digital sovereignty and supporting economic growth while ensuring data protection and security. Overall, Nvidia's dual focus on sustainability and sovereign AI at GTC 2024 showcases the company's holistic approach to technology development, balancing innovation with ethical considerations and environmental responsibility. This strategic emphasis on sustainability and sovereignty sets Nvidia apart as a leader in responsible AI practices and underscores its commitment to shaping a more sustainable and secure future for AI technologies. Points to Reflect Upon: Environmental Responsibility: Consider the implications of Nvidia's sustainability efforts on the tech industry and environmental conservation. Develop a sustainability strategy to align with Nvidia's initiatives. Ethical AI Practices: Reflect on the importance of promoting ethical AI practices and ensuring transparency in AI development. Global Engagement: Explore opportunities for global engagement to promote sovereign AI practices and support technological independence. Market Position and Future Outlook Nvidia's strong market position and future outlook were prominently highlighted at GTC 2024, emphasizing the company's leadership in GPU technology and AI innovation amidst competition from rivals like AMD, Intel, and emerging startups. Announcements at GTC 2024 regarding new products such as the Blackwell GPU series, DGX GB200 System, and DGX SuperPOD underscore Nvidia's focus on advancing AI infrastructure and enabling next-generation AI models. This strategic vision positions Nvidia at the forefront of technological advancements that will shape the future of AI computing and high-performance computing (HPC). Overall, Nvidia's market position and future outlook at GTC 2024 highlight its resilience, innovation drive, and strategic vision for continued growth and leadership in the evolving landscape of AI technologies. To sum up, Nvidia's annual developer conference in 2024 was a platform for unveiling cutting-edge AI technology, software tools, and strategic partnerships that reinforce the company's position as a leader in the tech industry. The key takeaways from GTC 2024 underscore Nvidia's commitment to innovation, sustainability, and collaboration, setting the stage for continued advancements in AI, GPU technology, and high-performance computing. P.S. Check out some of my previous articles at HackerNoon: Google Introduces Gemma: Revolutionary Access to Pre-Trained AI Models Sora's AI Innovation Set to Disrupt Video Content Creation Breaking Down the Next Big Thing in NFTs: ERC-404 Spotlight Gemini 1.5 Unleashes Unprecedented Context for AI Applications The AI Revolution in Journalism: A New Era of Enhanced Reporting DeFi Community Building: A Step-by-Step Guide for Crypto Startups

Nvidia GPU Technology Conference In 2024: A Deep-Dive 

Artificial intelligence gives humanity tremendous opportunities. With that, great risks. We celebrate online translators, amicable chatbots, well-drawing neural networks, and robot vacuum cleaners. We jubilate everything that makes life easier — not to mention scientific achievements. That said, we seem to be afraid of AI. But why? Is it even justified? Let's find out. First Steps in AI The first research on AI began in the 1930s. At that time, AI was already deemed a computer program able to perform tasks and learn. In 1927, an engineer of the Westinghouse Electric and Manufacturing Company, engineered the first humanoid robot, Herbert Televox. Roy Wensley It was programmed to answer calls and execute simple commands. Imagine you call home and ask Televox whether the lights are turned off! Or you can request to preheat dinner in the oven, and the robot will turn it on remotely. All that in 1927! Rise of AI: Dartmouth Conference The term artificial intelligence was first used in 1956, at the , New Hampshire. At the conference, scientists discussed many areas of AI application, from vision, learning, and research to language, gaming, and human interactions with robots. Dartmouth Conference in Hanover Since then, the golden age of AI has started. Artificial intelligence has become the official field of science. AI robots could solve specific tasks. For more accomplished versions, a bigger computing capacity was necessary. Not So Positive AI: Intelligence Explosion The number of robots — from a vacuum cleaner to a smart kettle — used in daily life was growing. Researchers tried to predict the further flow of things. In 1966, British mathematician Good in his work ‘Speculations Concerning the First Ultraintelligent Machine’ formulated the principle of technological singularity, or intelligence explosion. Irving John According to this phenomenon, “ultraintelligent machine could design even better machines; there would then unquestionably be the ‘intelligence explosion,’ and the intelligence of man would be left far behind. Thus the first ultraintelligent machine is the last invention that man need ever make.” This work caused a great discussion in society and led to ethical concerns: Is it safe to interact with machines? Where are the boundaries of interactions? Who controls robots? Trade-Offs: Asimov’s Laws of Robotics One of the most famous solutions to the problem was proposed by the American writer and professor of biochemistry in 1942. In Asimov’s story ‘Runaround,’ he outlined the three Laws of Robotics: Isaac Asimov A robot may not injure a human being or, through inaction, allow a human being to come to harm. A robot must obey the orders given to it by human beings, except where such orders would conflict with the First Law. A robot must protect its own existence as long as such protection does not conflict with the First or Second Laws. But, many years later, these postulates remained fiction and became the ground for future doubts, fears, and frightening scenarios. Where Does the Fear Come From? In the 1990s, neural networks became a real breakthrough. They could collect data, form experiences, analyze the received information, and draw conclusions at much higher speeds. Since there were more precise and less provocative definitions for modern technologies, the phrase artificial intelligence lost its relevance. John McCarthy, an American computer and cognitive scientist, said, ‘As soon as it works, no one calls it AI anymore." It is human nature to be scared of the new — the law of survival in action. Artificial intelligence provokes a defensive reaction. But it’s not the AI itself that frightens us. It’s the fear of the unknown. Many do not understand artificial intelligence, its processes and its consequences. People are also scared lest artificial intelligence replace them at work. Here also comes media. In movies, books, and media, AI is often an evil genius who wants to destroy humanity. At the same time, there are many stories of human interaction with technology: Self-realization through a robot (the movie ‘Real Steel’); Complete suppression of human will by a machine (the picture ‘Upgrade’); Use of technology for good (Iron Man's virtual assistant Jarvis); Capture of the world by intelligent machines (the movie ‘Terminator’ or the ‘Matrix’ trilogy). Smart machines appeared in fantasy long before they did in real. The word ‘robot’ was first used in the science-fiction play R.U.R. by in 1920. Čapek’s robots were very close to the modern perception of androids. Karel Čapek People view AI as a living organism capable of evoking feelings, if not to replace a person. We can see that in the picture ‘Her.’ Moreover, we no longer perceive technologies like Siri, Mercedes, or chatbots as AI. However advanced, machines do not have a nervous system. Even a perfectly accurate emulation of the human brain in an electronic version cannot make a robot feel and empathize. Mikalai Dzisko*, Senior Editor at , says:* Ocean Power ‘Journalists also distort the image of neural networks and artificial intelligence. Instead of educating their audience and showing them how to use the neural network, they mouse around for intolerance and unethical behavior in ChatGPT. Will it write the f-word? When will it offend a user? Is it possible to bring it into conflict? How many genders does the neural network admit? These questions blur the reputation and essence of the technology. Will the Fear Of AI Disappear? Despite widespread fear, some movies, books, and people support a positive reputation of neural networks: In the 1990’s sci-fi picture ‘Total Recall,’ AI helps the main character to investigate crimes on Mars; In the Mass Effect computer games, you can’t win without AI; In the science fiction series ‘Black Mirror,’ artificial intelligence is sometimes represented as a positive character: in one of the episodes, it helps people find the perfect match — the utopian Tinder of the future. Artificial intelligence is often mentioned in gender studies. One of the most famous feminists of the last century, Donna Haraway, believed that cyborgization would help overcome binary, with all the ensuing problems and stereotypes. Today, human rights defenders are moving on. The non-profit organization raises the questions of people who can already be called cyborgs. Cyborg Foundation Its co-founder, artist Neil Harbisson, calls himself the first cyborg in the world because he implanted a special antenna in his skull that converts light waves into sound. ‘Should they have a license to use the technology? Their international law? What should cyborg insurance look like? Who will protect the civil liberties and inviolability of cyborg bodies?’ asks Eliezer Yudkowsky, co-founder and research fellow at the Machine Intelligence Research  Institute Stephen Hawking made a huge contribution to the popularization of the thesis ‘Technology is man's best friend.’ The famous physicist and cosmologist publicly spoke in favor of research and the development of artificial intelligence. Hawking believed that, with proper use, AI can help people solve many global problems. By personal example, he showed that a serious illness is no longer an obstacle to active scientific and public activity the moment you let technology into your life and body. Transhumanism VS. Posthumanism Transhumanists preach that the main task of a person is to improve the quality of their own life and overcome illness and death. Technology should be the best helper in this process. Cyborgization, regenerative medicine, various digital practices — virtual reality, metaverses — all this should make a person happy. Sounds defiant, doesn’t it? On the other hand, technology has been simplifying our lives for a long time. We just do not notice it: cars, microwaves, phones, the Internet, search engines, and so on. There is also a more radical approach, posthumanism. This philosophy is based on the consideration that man has reached the peak of their evolution in the Darwinian understanding and, to reach the next level; it is necessary not just to apply the product of his intellectual activity, but to directly implement it into life: to implant a chip in the brain to improve vision or hearing. This can also include biohacking, improving health indicators with the help of medicines, a proper lifestyle, and, of course, technologies, such as chips that read blood pressure, temperature, and sugar levels. Not the treatment of diseases, but the upgrade of the body and its own capabilities. Three Whales of Transhumanism Both trends are mostly theoretical, but the three main ideas of transhumanism seem intriguing. The first is super longevity. The belief that life is finite, and even more so limited by a conditional century, is a cultural stereotype that forces a person to distribute events within the time allotted to a human: to get married before the age of 25, to give birth before 30, to reach success before 45. What if you imagine that life is 200, or even 250 years? How will our attitude towards the time change? To your own health and biological age? What if we don't rush, try to do more, and adjust our lifestyle to this idea? A second duty of a transhuman is to raise the standard of living by reducing stress. Stress has become the main disease of the 21st century, the primary source of other problems and illnesses. A high standard of living is not only material goods, it is primarily harmony with oneself. This is what transhumanism adherers suggest to focus on. The third thesis of transhumanism is as follows: a person must combine the abilities of the brain and technology. Disturbing, right? In fact, we have been delegating some brain functions to neural networks for a long time: memory, sensory organs, and biological clocks, to name a few. We get up on the alarm clock, we do not waste time trying to remember the actor's name, we trust the route suggested by the navigator — and we are unlikely to follow it on our own. The neural network helps eliminate some processes, change our needs, reduce the time for completing tasks, and manage to do the important ones. Is It Even Possible? In theory, it is possible to achieve superintelligence by combining the capabilities of the human brain and a computer. Special implants could stimulate brain function, allow you to access the Internet, and interact with computers or other implant carriers. Today, such technologies are used only for medical purposes. For example, scientists at have developed implants that can stimulate individual parts of the brain with the help of electrical impulses and thus help address depression or Parkinson's disease. Stanford University All these utopian developments can inspire or make you think about improving your life. In the fight against AI anxiety, you just need to start. The first piece of advice unanimously given by those who work with neural networks is to stop calling this technology artificial intelligence. Intelligence — the defining characteristic between a human and an animal — remains man's main strength. There are no competitors on the horizon so far.

AI: Friend or Foe? What's Behind Our Fear of Artificial Intelligence?

The future of AI chips is about more than NVIDIA: AMD, Intel, chiplets, upstarts, analog AI, optical computing, and AI chips designed by AI. The interest and investment in AI is skyrocketing, and generative AI is fueling it. Over one-third of CxOs have already embraced GenAI in their operations, with nearly half preparing to invest in it. reportedly What’s powering it all – AI chips – used to receive less attention. Up to the moment, OpenAI’s for a “wildly-ambitious” tech project to boost the world’s chip capacity. Sam Altman claimed he wants to raise up to $7 trillion and sensationalism aside, however, keeping an eye on AI chips means being aware of today’s blockers and tomorrow’s opportunities. Geopolitics According to a , the global AI chip market is expected to reach $89.6 Billion by 2029. The demand for AI chips has increased substantially over time. Growth in AI technology, rising demand for AI chips in consumer electronics, and AI chip innovation all contribute to this forecast. recent study by IMARC Few people have more insights to share on AI hardware than . In an extensive conversation, Pialis shared his insider’s perspective on the AI chip landscape, the transformative rise of chiplets, specialized hardware for training and inference, the future of AI chips and emerging directions like analog and optical computing, and much more. Tony Pialis, CEO and co-founder of Alphawave An AI Leader Behind the Scenes Pialis is a semiconductor serial entrepreneur. He has previously founded and successfully sold two startups: Snowbush Microelectronics and V Semiconductor Inc. V Semiconductor was acquired by Intel in 2012, where Pialis spent a few years as the Vice President of Analog and Mixed-Signal IP. Pialis co-founded AlphaWave in 2017 with the goal of becoming “the next great semiconductor company.” Alphawave went public in 2021 with an impressive market cap of $4.5 billion. The company’s core products include silicon IP, chips, custom silicon, and high-speed interconnect technology tailored for major hyper-scaler customers like Google, Microsoft, Amazon, and Meta. Alphawave is an AI leader behind the scenes, powering hyper scalers and founded on a couple of key premises that Pialis laid out. https://pod.co/orchestrate-all-the-things-podcast-connecting-the-dots-with-george-anadiotis/the-future-of-ai-hardware-leaders-dark-horses-and-rising-stars-featuring-tony-pialis-alphawave-ceo-co-founder?embedable=true , the major players in terms of scaling out data centers and computing these days are no longer the Ciscos of the world. It’s transitioned to Hyperscalers – the Googles, Microsofts, Amazons, and Metas. Hyperscalers have internal design capability, and they also build their own servers, networks, data centers, and campuses. First , the major challenge facing AI moving forward is not computing. Pialis’ view is that the ability to design and implement compute is there. The challenge is the connectivity needed to connect the compute in order to process all of the data. This is what AlphaWave is focusing on. Second The Explosion of Specialized AI Hardware While consumer apps like ChatGPT drove the initial hype cycle in 2023, enterprise adoption reports are mixed. However, according to Pialis, the AI semiconductor sector has seen an enormous explosion of investment and new in the second half of 2023 across industries and geographies. design starts Pialis cited major national initiatives around building domestic AI chip capabilities in the US, Canada, UK, France, Germany, Korea, Taiwan and Japan. After years of relying largely on NVIDIA, governments want to foster their own chip industries to reduce strategic dependence on any one vendor. While NVIDIA CEO Jensen Huang also says that , it does not seem like this also includes the hardware layer. every country needs Sovereign AI Pialis believes this surging demand is spurring not just startups but also tech giants to develop specialized training and inference hardware. In his view, not every organization could or should develop its own AI model, but that’s bound to change. “Over time, AI will inevitably progress to be something like a utility, where hyperscalers will provide access to all of the compute power, like electricity. I think it will be reasonably priced. Anyone will be able to use this utility to train, develop, optimize and deploy their own models. But there’s going to be a lot of profit taking between now and then, and who knows how long it’ll take to eventually reach that state”, said Pialis. Leaders, Dark Horses, and Rising Stars in the AI Chip Race Few people would disagree that NVIDIA is clearly the dominant force in AI accelerators today, and Pialis shares this view. He sees AMD as a strong second place and praised CEO Lisa Su for her leadership. Others such as Ben Lorica also see . favorable winds for AMD in the GenAI chip market But Pialis cautioned not to underestimate Intel, calling it a dark horse that could still prove highly formidable in this market. More specifically, Pialis spoke highly of . Intel’s Habana acqui-hired division led by David Dahan Having worked at Intel, Pialis has an insider view. Having , and kept track of their , we tend to agree. Dahan helped architect new Intel processors that have showcased performance exceeding NVIDIA’s latest GPUs on key benchmarks. covered Habana, met Dahan MLPerf results But performance is not everything. As Pialis also noted, NVIDIA’s software platform including CUDA gives it a huge advantage in the AI chip segment. The ecosystem effects are strong, as engineers and researchers develop optimized frameworks and models for NVIDIA’s architecture. However, there is potential for alternatives; for example, . Pialis opined that a rule of thumb for AI hardware companies is that for every hardware engineer, they need 2-3 software engineers. So even though there’s a lot of talk about NVIDIA and hardware, the vast majority of the investment actually goes into the software. building upon AMD’s hardware and software stack https://www.youtube.com/watch?v=VYx1rGieDmE&embedable=true This was confirmed by NVIDIA’s Dave Salvator. At the , Salvator said that NVIDIA has twice as many software engineers as hardware engineers. That’s not at all by accident, as he emphasized. last MLPerf results briefing for 2023 However, Pialis sees far more potential for challengers in . The reason is that standards are still emerging in that area. As OctoML’s Luis Ceze , there’s for LLM serving, for portable deployment, and and for multiplexed fined-tuned model inference. Inference, is in fact, a bigger market than training as Pialis noted. the inference accelerator market shared vLLM MLC-LLM LoRAX Punica “There’s so much focus on training, large models and the cost of training. But it’s deployment where we all benefit from, on the inference side. That needs significant scale. That needs a different solution. More chips will inevitably be sold on the inference side than the training side. I’m sure business plans also improve with that increase in volume”, Pialis said. Startups like and are attracting significant funding, and so are companies from countries like the UK, Korea, and China trying to reduce reliance on US firms. As far as hyperscalers go, Pialis sees Amazon and Google as being neck to neck, with Microsoft showing strong momentum, and Meta being behind and rumored to acquire one of the smaller startups. Groq Tenstorrent Join the Orchestrate All the Things Newsletter Stories about how Technology, Data, AI, and Media flow into each other shaping our lives. Analysis, Essays, Interviews, and News. Mid-to-long form, 1-3 times per month. The Chiplet Technological Revolution One of the biggest transformations underway in semiconductors, according to Pialis, is the shift to . Previously, progress meant integrating more and more functionality into a single monolithic chip. But transistor dimensions are now down to ~5 atoms wide, and even tiny defects can ruin a whole chip. chiplets Pialis shared an anecdote from his own experience to highlight this. As he related, when he visited OpenAI at some point, he saw a bunch of engineers on their knees praying in front of servers. Not because they were “ ”, but rather because they didn’t want the model they were training to crash. feeling the AGI Some of us may have first heard about chiplets in the context of the China–US trade war that spills over to technology. Chiplets are a key element in the strategy of both and the . China US For Pialis, “chiplets are another technological revolution to deal with the problem that God has thrown at us.” That may sound like a wild appeal to the supernatural, but it has more to do with the challenges of physics at nanoscale. “When we build transistors, which are the basic building blocks of any integrated circuit, we’re stacking atoms. When you’re stacking atoms, the laws of probability, the laws of averaging, fall apart, because now it’s two atoms rather than hundreds of atoms and electrons. So what you get is defects”, Pialis explained. Chiplets are an innovative solution to this nanoscale physics challenge. Rather than one giant chip, designs are split into smaller Lego-like chiplets that can be connected through advanced packaging. This modularity lets chipmakers avoid throwing out entire devices due to one small flaw. The benefits, Pialis said, are important for both makers and buyers. “Silicon is no longer the foundation of leading edge semiconductors; it’s the package. Silicon is now something that is a component on this packaging. There’s been a lot of talk about semiconductor supply chains. There’s a lot of silicon capacity now. Where there’s still very little to no capacity is on the packaging, specifically for designs built using chiplets”, Pialis noted. Chiplets as Lego Building Blocks However, there is a standout example of an AI hardware company that seems to be successful, despite going the opposite way: Cerebras. Cerebras is known for its wafer-scale hardware. Pialis was positive that Cerebras was going to encounter the same defects as everyone else. The difference with Cerebras is they have redundancy. A wafer is like a panel, Pialis noted. Cerebras doesn’t even cut the panel. What they do is they have software that works around the defects. Fundamentally, they’re using silicon as the foundation rather than packaging. They have multiple chips that are arrayed, but they’re not cut. They just connect on the wafer. It definitely is a different approach. However, Pialis added, there are advantages to cutting things up. Imagine a vendor like Intel is trying to build different units such as a CPU, a GPU, a DPU, or a networking device. By breaking things up into smaller pieces, they become like Lego building blocks. So you can have a processor core chiplet, a PCI Express connectivity chiplet, an Ethernet networking chiplet, a DDR memory I/O chiplet, and a memory I/O chiplet. These chiplets can be mixed and matched in a package to build out an entire product portfolio. Pialis thinks that from a design complexity and upfront investment perspective, that’s a winning formula. Pialis sees massive benefits in chiplets, reducing costs by over 60% and power consumption by 40%. This is a major incentive for hyperscale data centers. While he sees Apple, AMD, and Intel leading in chiplets today, Pialis believes chiplets will soon become essential for any company focused on leading-edge hardware. Join the Orchestrate All the Things Newsletter Stories about how Technology, Data, AI, and Media flow into each other shaping our lives.Analysis, Essays, Interviews, and News. Mid-to-long form, 1-3 times per month. Software and Chiplet Modularity, Composability, Programmability Drawing parallels from the world of software engineering, one can wonder how come the modularity that chiplets bring to the table did not prevail earlier. But in hardware, up to now, the winners were the ones that could integrate into a monolithic device the most. The reason is that this drove down cost, which Pialis noted has created “a maniacal focus on integration.” However, as we now approach the dimensions of a single atom, the manufacturing cost eclipses the cost of integration. But even in software, . excessive modularity can also incur excessive overhead Pialis expects that some hardware vendors may take the chiplet approach too far. If functionality is broken into too many tiny pieces, the cost of integrating all of these becomes limiting. So ultimately, it’ll be a hybrid approach that wins. There are two ways of splitting things up with chiplets. The first way of building chiplets is building a single Lego block, mirroring it over and over with the same functionality, and they just talk to each other. That seems more akin to the “traditional” approach. However, one may wonder how these identical chiplet building blocks can be synthesized. The answer is, again, software. Different packages can be composed for different needs, based on the same chiplet, using the software. For example, multiples of 1, 2, 4, or 8 chiplets. The same silicon, just packaged differently, priced differently, and with different software to leverage the incremental compute and memory bandwidth associated with these devices. The other way of building chiplets is by splitting and cutting, creating Lego blocks for different types of functionality. This creates chiplets like a compute Lego block, a training I/O Lego block, a network I/O Lego block, and so on. Pialis sees more momentum behind this approach, as reusable Lego blocks can accelerate other products too. Analog AI, Optical Computing, and AI-Assisted Hardware Design Chiplets may be the immediate way forward. But at a time when power-hungry digital accelerators like GPUs dominate, Pialis claims there’s more to the future of AI chips. And that involves a radical break with the current paradigm. AI needs massively parallelized arithmetic processing, and binary is the dominant approach for this. Using binary compute, numbers are represented as ones and zeros, and floating point arithmetic is done defining precision and range. In analog arithmetic processing, floating point can be represented via voltages or currents. In theory, that may have infinite precision. In the real world where noise exists, that may yield less precision. But that may well work for applications. Using very small currents keeps things operating at low power. Edge AI https://pod.co/orchestrate-all-the-things-podcast-connecting-the-dots-with-george-anadiotis/the-future-of-ai-hardware-leaders-dark-horses-and-rising-stars-featuring-tony-pialis-alphawave-ceo-co-founder?embedable=true There’s another form of compute that some companies are investing in: optical compute for arithmetic. Optical compute uses the properties of optics to implement what are called (multiply & accumulate functionality) which is the core of any arithmetic unit. This results in even lower power demands. MACs Pialis noted that analog and optical approaches are attracting billions in investment to meet specialized needs around computational scale, energy efficiency, and precision. However, it remains unclear if analog techniques can scale effectively to match digital compute for cutting-edge AI models. There is active debate on this question in the hardware community. Even though these techniques are often dismissed as impractical, Pialis believes the investment will eventually pay off. For example, recently, it was reported that capable of processing computer vision tasks with remarkable speed and energy efficiency. researchers at Tsinghua University have developed the world’s first super all-analog photoelectric chip Finally, there’s also the question of . Pialis said that the most effective hardware designers these days are the best software developers. But if AI models can be trained by leveraging the wealth of their experience, that may bring about a sea change. using AI to help design hardware to power AI While the future of AI chips is unknown, Pialis believes the fundamental principles of engineering are timeless. Hopefully, that translates to something that does not end up draining the world’s energy and resources on the promise of future gains to be made.

What Is the Future of AI Chips? Leaders, Dark Horses, and Rising Stars

189 Stories To Learn About Reading Books

280 Stories To Learn About Future Of Ai

84 Stories To Learn About Chatbot

*** The History of Science Funding The scientist has always been a risky employee in terms of ROI.  They are trying to uncover truths about the universe, and oftentimes, the universe gives them the bird. Because experimenting means risking a lot of money and time on a possibly worthless idea, a scientist is a forever intern, until they make a breakthrough, then they become senior associates. That said, the scientist has been trying to make his own money. With the Alchemists, this was literal. While they invented Chemistry and catalyzed the birth of worldwide gold rushes, turning lead into gold was, alas, a wrong idea believed for centuries. The search consumed the lifetime of one person after another, with only dreams of wealth to get them through. But their tenacity was a testament to their resilience and patience. Patience meant, for the time being, remaining in the employ of someone who could afford the risk of backing science in the past. And that meant a King because they had gold/money collected from taxes. Wizards and fortune tellers didn’t make much sense, and they were comedians come to give the King a jolly time before he went to war. Though, while throwing guts on a wall may be stupid, it was mysterious and if they were right, doubly impressive. And here was this (broke) scientist, insisting that they knew a REAL strategy to help the King win. What was their credibility? Circa 200BC, the scientist is Archimedes and the King is the political head of Syracuse. Being a leader in those times largely meant you needed to win wars and as history had taught Kings the utter uselessness of charlatans, the scientist was given a chance to make his case. Archimedes made his case alright. It that he sunk the ships of the Romans with ingenious cranes and catapults that threw heavy stones. has been recounted Since his time, patronage from Kings, Princes, Bankers and Presidents to scientists has been immense. The scientist is respected, and all great scientists, even Leonardo da Vinci who was mostly fawned over for his impressive artwork, took off some time to design weapons. For weapons help the King win and hence, keep the scientist with a job. Scientists soon became ever more pivotal in war efforts. Especially Physicists and Chemists, who have designed countless forms of military hardware to date. However, the emergence of scientific methods in Economics, Sociology, Psychology, Biology, all riding on the power of statistical analysis and logical inference, quickly made the scientist important in every field of human endeavour. The problem however remained. To get a loan, the scientist usually had to, and still has to, rely on government funding because the government remains the biggest fish able to take on sufficient risk in the financial markets. Hate it all you want, but without credit from the government, be it debt or real money, a lot of scientific work would have stalled. Best example - SpaceX. However, governments and their ability to debase money and finance things on debt, made the world wars last longer to destroy the world much harder. It also led to economic depressions and after the 2008 financial crisis, Satoshi Nakamoto, a genius computer scientist, decided to create Bitcoin – a monetary tool outside the purview of government currency debasement. Birth of a financial model divorced from the King Bitcoin has been controversial from its inception. Of course, Kings and governments hate it. It is cutting into their business. A big part of the private sector also loathes its nature of priding in technological exclusivity, so that six years later, a hard fork was implemented to create Ethereum, a cryptocurrency built to provide extra blockchain services such as smart contracts. Many other hard forks quickly followed from the faithless. However, Bitcoin is still mainstream and has recently become interesting to Big Finance. The Bitcoin ETF proves as much. After 15 years of laughing at a thing that keeps building bigger and better, even the harshest critic will shut up and start listening more intimately. Because they now know they are missing something. Energy scientists are also starting to pay attention, with some claiming that Bitcoin mining may be our . are doing well selling cheaper electricity packages that have been subsidized by Bitcoin mining on the side. fast track to an energy transition Energy engineers But these are late to the party. Computer scientists have been earning well in Bitcoin, Ethereum, Doge for decades. And if any scientist has been comfortably living outside the need to get big budget funding from governments, that has been the computer scientist, from way back in the 20th century. While the field of Computer Science is much younger than Physics, Astronomy, Biology, many computer scientists quickly learned to earn big from private enterprise and individualized work, without begging the Kings and their Big Bankers for a loan. Futuristic Science Funding itself through Bitcoin Bitcoin is money for the people, and science has evolved to also be something done as a public good, on an open source plan, by dedicated people who put in their own time and money to make something good happen. As I mentioned, Computer Scientists now exist on the internet who are eagerly improving the Bitcoin ecosystem using the power of collaboration and their earnest efforts. These mostly come from the land of software development and after getting orange-pilled, decide to stay at Bitcoiner-land and build the Bitcoin ecosystem. However, there has not been much of an on-ramp for other types of scientists and this is the opportunity existing, wide open, for anybody to take. Currently, many a scientist in the Bitcoin world is a computer scientist of the type who builds zero-knowledge proofs, rollups, and more layers of the Bitcoin blockchain. But many more scientific people are needed. To re-iterate, I feel more electrical and power engineering scientists should be at the forefront of how Bitcoin mining equipment can provide electricity access to those without, making investments in clean, stranded energy sources much easier to manage. What better job than to be the electric engineer who designs their own job by bringing electricity to your people? Every university student of electrical engineering, whatever the developing country, should learn about this. Because developing countries in Africa, Asia and Latin America are struggling with scaling electricity usage from otherwise cheap energy sources, because of the high premiums that locals just cannot pay. The knowledge of using Bitcoin to subsidize electricity access from stranded energy sources should be spread as far and as widely as possible. Now, as an aspiring quantum information scientist, I feel there is also a place for me in the Bitcoin ecosystem. While the public might know how quantum information science threatens to undo the good that Bitcoin has done by potentially hacking it, I find that building quantum-resistant algorithms is far from scratching the surface of quantum capabilities. For example, quantum technologies could provide for Bitcoin nodes, allowing even weak signals to be successfully transmitted and ensure blocks are mined. better signal sensors Quantum computing could help guide better on how Bitcoin infrastructure could be better set up across nations and continents for optimum usage. solutions in logistics Quantum networks could provide better ways to securely. transmit Bitcoin information Some problems in the Bitcoin ecosystem could be (Bounded-error Quantum Polynomial) problems, which are the purview of quantum information science and these should be discovered and worked on using quantum computers. BQP Most importantly, quantum computers will be orders of magnitude better at mining proof-of-work cryptocurrencies than classical computers. When Bitcoin befriends Quantum Mechanics Moore’s law is coming to an end. IBM has already built transistors with only 5 atoms. To get smaller, it becomes harder to manage thermal noise and other nanoscale problems. To manage thermal noise, even more cooling will be done, the way it is done for qubits. Also, a 1-atom transistor is effectively a qubit. This is the quantum realm, and like it or not, more people will study quantum mechanics in the not-too-distant future, and by extension, quantum computing. The classical computing world can only scale by a multiplicity of the same technology, which will consume even more energy. The efficiency gains of quantum technologies like quantum computers will be needed. While , the science of building qubits is ramping upwards and systems are getting more powerful. There might not be an exponential growth in qubits with their own Moore’s law, but a linear growth in qubits leads to exponential computational power. Which kind of is a Moore’s law! I have said quantum computing is very expensive You remember this graph; If approximately 14 billion dollars in 2023 can bring in (logical qubits are the real qubits as manipulated in mathematical theory), by linear extrapolation, $125 billion by 2030 as seen above will bring in 428 logical qubits. 48 logical qubits But 7 years from now, Bitcoin itself might have ballooned to $200,000 per bitcoin which is a $4.2 trillion market cap. By linear extrapolation, $3 trillion is needed in global investments to take us to 10,000 logical qubits. As we can see, Bitcoin alone can pay for it over a 7-year span. Why 10,000 qubits? My Research Project Over the past three months, I have been studying QCourse551-1 which was offered in a collaboration between QWorld and Classiq Technologies. My research project was Project 11 - Bitcoin Mining using Grover’s Search Algorithm. The motivation for my project came from that shows that 10,000 logical qubits alone can do Bitcoin mining faster than any classical system using the same energy consumption. a research paper This is because 10,000 qubits can encode the full double SHA256 hashing process that goes into mining Bitcoin, but since they use Grover’s search algorithm, they effectively achieve quadratic speed-up by square_root(N) where N is the number of elements to check in an unstructured database. ASICs mining for bitcoins also check an unstructured database of random hashes for the nonce value. The question is, which one is faster? Most ASIC miners today are checking terahashes or For example checks 120TH/s. That is 120 million hashes every second. 1 million hashes every second. this miner Can a quantum computer beat that? Of course. square_root( 120 million) = only 10954 checks per second. A 1,095,400 % efficiency increase. And the more the competition, the more the quantum computer tears ahead in the race. The is rocketing upwards of 600 million million hashes per second. This is outputting enough heat to make climate scientists worried sick. But what if we replaced all those clunky ASICs with state-of-the-art QPUs from the year 2032, how many comparative hashes per second would they clock? square_root( 600 million million) = 24 million checks per second. A . current hash rate 2.4 billion % increase in efficiency If the above is not an attractive proposition, I don’t know what is. Bitcoin should fund quantum computing research. It will take a while, but when it happens, it will be glorious. It will be a more canonical event than the Moon landing. In our research, we were checking a much-simplified system with only 8-bit inputs and outputs, modeled on a Grover’s operator working with under 25 qubits. Given our really low hashing power, the advantages of Grover’s search could not be gleamed first hand, and could only be inferred with a relevant complexity analysis of the runs/depth of the Classical computer (my laptop) vs the depth of our Grover’s circuit executed on the Classiq Aer simulator. While it would have been good to model the energetics and the complexity graph of what this means, unfortunately, we ran out of time. However, we had some deliverables to show for our work. Below is what we had including the depth of each circuit. a) The simplified MD5 classical algorithm We modeled a simplified version of the MD5 (a most primitive cousin of SHA256) hashing 95 printable ASCII characters to find the nonce for hex values with difficulty = 1. Note: Our hex values were only 2-digit because we are working with 8-bit values. So our possible difficulty 1 values are only in 16 possibilities i.e. 00, 01, 02, 03, 04, 05, 06, 07, 08, 09, 0a, 0b, 0c, 0d, 0e, 0f. 00 is also a difficulty 2 element. Hence the searching/hashing function is called simplified_MD5_8bit_difficulty_1or2. def simplified_MD5_8bit_difficulty_1or2():
    printables_list = '0123456789abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ!"#$%&\'()*+,-./:;<=>?@[\\]^_`{|}~'
    for i in printables_list:
        x2 = ord(i)

        x3 = bin(x2)[2:]

        if len(str(x3))==7:
            x4 = x3+'0'
        if len(str(x3))==6:
            x4 = x3+'00'

        x5 = list(x4)

        a0 = 2*int((x5[0]))
        a1 = int(x5[1])

        b0 = 2*int((x5[2]))
        b1 = int(x5[3])

        c0 = 2*int((x5[4]))
        c1 = int(x5[5])

        d0 = 2*int((x5[6]))
        d1 = int(x5[7])

        a = a0 + a1

        b = b0 + b1

        c = c0 + c1

        d = d0 + d1
    
        a = ((d ^ (b & (c ^ d))) + (c ^ (d & (b ^ c))) + (b ^ c ^ d) + (c ^ (b | ~(d))) + b)
        
        print()

        d1 = d//2
        d0 = d%2
        d = d1*(2**7) + d0*(2**6)

        a1 = a//2
        a0 = a%2
        a = a1*(2**5) + a0*(2**4)

        b1 = b//2
        b0 = b%2
        b = b1*(2**3) + b0*(2**2)

        c1 = c//2
        c0 = c%2
        c = c1*(2**1) + c0*(2**0)

        decimal_result = d + a + b + c
        binary_result = bin(decimal_result)[2:]
        print(binary_result)
        hex_result = hex(decimal_result)
        print(hex_result)

        if hex_result == '0x0':
            difficulty_2.append(hex_result)
            collisions_difficulty_2.append(i)
        if hex_result == ('0x1' or '0x2' or '0x3' or '0x4' or '0x5' or '0x6' or '0x7' or '0x8' or '0x9' or '0xa' or '0xb' or '0xc' or '0xd' or '0xe' or '0xf'):
            difficulty_1.append(hex_result)
            collisions_difficulty_1.append(i)

    print('Difficulty_1 list = ',difficulty_1)
    print('Difficulty_2 list = ',difficulty_2)
    print('Colliding inputs for difficulty 1 = ',collisions_difficulty_1)
    print('Colliding inputs for difficulty 2 = ',collisions_difficulty_2)
    
difficulty_2 = []
difficulty_1 = []     
collisions_difficulty_1 = []
collisions_difficulty_2 = []

import numpy as np
simplified_MD5_8bit_difficulty_1or2() As can be seen, to find the difficulty 1 (and by extension the difficulty 2) elements, we have to reiterate through 40 lines of code (ignoring spaces and printing in the for loop). There are 96 printable ASCII characters to hash and check hence 96 * 40 which gives us a circuit depth of 3840. However, this is run through fast by my laptop. Less than a few seconds. b) The simplified MD5 quantum algorithm We modeled a quantum version of our simplified MD5 algorithm to simulate finding the difficulty 1 nonce values no matter what the input ASCII character to the circuit was. This is possible because all we need is a zero at the beginning of our hex value. Since our hash set was small (only 96 ASCII characters), we tuned the system to ignore the input ASCII value so it is redundant, and to aim to capture all possible zeros given the Arithmetic expression; a == ((d ^ (b & (c ^ d))) + (c ^ (d & (b ^ c))) + (b ^ c ^ d) + (c ^ (b | ~(d))) + b) which is also in the classical Python algorithm. Notice that the 4 functions being added together are derived from the . original MD5 algorithm Since we want difficulty 1 values, we add and to the above function and then synthesize it in the Classiq platform. d = 0 a = 0 (Note: This is a shortcut thanks to the tireless work of the Classiq team. The Python code was also built in the Classiq Python SDK but it had a last-minute glitch. Hence we only have working Python SDK code for the Grover’s operator, but not for the Grover’s operator with a difficulty-1 hash/search, which is what you had to see here ) Hence our input Arithmetic expression is; a == ((d ^ (b & (c ^ d))) + (c ^ (d & (b ^ c))) + (b ^ c ^ d) + (c ^ (b | ~(d))) + b) and d == 0 and    a == 0 The resulting circuit has a depth of only 913 The quantum circuit wins! Endnote While the project time ended, our work is still a work in progress, to continue until we build a functional 10,000 qubit quantum bitcoin miner. Voluntary efforts to support us technically are most welcome. Some of our code is still glitchy and we have only scratched the surface of the Classiq platform’s capabilities. An ingenious software tool that takes away the cumbersome tinkering with gates, as would have been needed for anybody building the same project using Qiskit. Special thanks to Classiq and QWorld for the opportunity. And our mentor - Dr. Eyal Cornfeld, for the guidance. ***

Hash Rate Analysis Projections: Why Bitcoin Should Embrace Quantum Information Science

হ্যাশ রেট বিশ্লেষণ অনুমান: কেন বিটকয়েন কোয়ান্টাম তথ্য বিজ্ঞানকে আলিঙ্গন করা উচিত

Prognosen zur Hash-Rate-Analyse: Warum Bitcoin die Quanteninformationswissenschaft nutzen sollte

Proyecciones del análisis de la tasa de hash: por qué Bitcoin debería adoptar la ciencia de la información cuántica

Projections d'analyse du taux de hachage : pourquoi Bitcoin devrait adopter la science de l'information quantique

हैश दर विश्लेषण अनुमान: बिटकॉइन को क्वांटम सूचना विज्ञान को क्यों अपनाना चाहिए

ハッシュレート分析の予測: ビットコインが量子情報科学を採用すべき理由

해시레이트 분석 예측: 비트코인이 양자 정보 과학을 수용해야 하는 이유

Projeções de análise de taxa de hash: por que o Bitcoin deveria adotar a ciência da informação quântica

Прогнозы анализа скорости хэширования: почему Биткойн должен использовать квантовую информатику

Hash Rate Analizi Projeksiyonları: Bitcoin Neden Kuantum Bilgi Bilimini Kucaklamalı?

Dự đoán phân tích tỷ lệ băm: Tại sao Bitcoin nên áp dụng khoa học thông tin lượng tử

哈希率分析预测：为什么比特币应该拥抱量子信息科学

229 Stories To Learn About Data Analytics

Quantum Computers are the most expensive technology on Earth. I suffice, that their success is a direct measure of how rich a technological civilization is. With the riches measured in usable, carbon-neutralized energy. (If we don’t count the carbon now we shall still have to count it later). Moore’s law will not work for quantum computers This idea came to me after watching a clip from with the quantum physicist, applied mathematician and entrepreneur - Guillaume Verdon. Lex Fridman’s podcast Classical computers were first invented in ancient times before Christ, to be honest. But for practical purposes they truly emerged on the global scene with the invention of the transistor in 1947 at Bell Labs. With the invention of the transistor emerged Moore’s law. “ ” – Wikipedia. This law was posited by Gordon Moore in 1965 and has been accurate since 1975 to date. Moore's law is the observation that the number of transistors in an integrated circuit (IC) doubles about every two years It is a law indeed. But for classical computers. Why? Just like artificial intelligence, a classical computer with a human user is a system trying to encode key details of entropic processes in the universe with as small an entropic model as possible. To compute is to represent states in the universe (be they naturally occurring or occurring in abstract mathematical worlds or wherever) and run one of their process changes in a simulation. The less energy we can use to do this, the better. Smaller transistors use less energy and thankfully, Moore’s law has seen these basic units of computation, transistors, shrink tremendously from the light-bulb-sized setup below. To this highly magnified image of IBM’s 2-nanometer chip technology. It looks like an X-ray of some strange animal’s dental formula. Each transistor in the above image is roughly the size of 5 atoms and the 50 billion transistors in the chip can fit on a fingernail. All this miniaturization has been possible because we did not need a lot of energy or material to represent a bit of information, at the fundamental level. We did not know it 49 years ago when Gordon Moore postulated his law. We do now. However, we need a lot of energy and materials to represent a logical qubit of information. And that is precisely why quantum computers will not scale according to Moore’s law. As the number of qubits we need increases, the energy needed hence the cost of running these quantum computations will likely increase to match the growth of real, carbon-neutralized GDP for the entire planet. But qubits are cold. What do you mean they need a lot of energy? Theoretically, a quantum computation consumes less energy than a classical computation. Because the energy changes are reversible, we could, theoretically, run quantum computations on zero energy! Because theoretically still, any energy we input is never output. For example, the fundamental logical gate in classical computations, called a NAND gate, looks as below. You can see that 2 bits of information, A and B, go in, but only one bit of information goes out. Since that information is energy, this shows that classical computers are wasteful of energy. Landauer showed The quantum version of the NAND gate is called the Tofolli gate and is seen below. 3 inputs of energy to give 3 outputs of energy. No energy wasted. But wait, you need 3 qubit inputs while in the classical case, you only needed 2 bits. Already, we can see we need more than is usual. Simulating nature is materially and energetically very expensive Needing more energy pulses for our inputs is not even a scratch on the surface of our energy budget when using quantum computers. Thing is, we are simulating nature using artificial nature, millions of times bigger than what we are simulating. Take superconducting qubits. While natural qubits like “spinning” electrons are extremely tiny things small enough to survive their quantum states from being hit by most cosmic rays, superconducting qubits behave like artificial spinning electrons we can easily control. The downside is that millions of cosmic rays see them very clearly. And they hit them. That is just one form of external noise. It costs us and it is going to keep costing us. We need and shall keep needing a lot of energy to keep energetic noise like cosmic rays out of our quantum systems. As Mr. Guillaume explains, the trick up our sleeve is to build error-correcting codes of codes of codes of codes of the noisy system, which eventually gives us a noise-less system (in the average case). But this makes the system bigger hence a bigger budget for the cool down. Heat is the second form of external noise. Should have been the first one we talk about. Taking the analogy of the spinning electron again, sitted comfortably in its orbit, infrared signals carrying heat rarely bump the electron enough to interrupt its spin. The electron is electrostatically bound to the nucleus and if that electron is sharing the orbit with another electron, the two are essentially entangled and again, harder to bother with a spin-out of their stable configurations. Thirdly, why we suffer so much in energetic costs is because it takes a very huge amount of energy to keep track of quantum states with our measurements. Nature doesn’t do measurements, we do. We are thus adding information (a measurement) to our simulation of nature. Hence more energy (remember Landauer). Measurement also introduces unnecessary energy into a quantum system which affects the position and momentum of our particles a la Heisenberg’s uncertainty principle. It also doesn’t help that collapsing the wave function of a superposed quantum state into 1 of many states, albeit randomly, gives us exactly zero information about the hidden variables of that state. This is true randomness, not what classical computers mimic which is pseudo randomness. Our classical randomization algorithms are all pseudorandom. Einstein . complained about this Therefore we usually have to make multiple repeated measurements to figure out the probability distribution of our quantum states, be it to learn how to nudge them or mitigate errors. These repeated initializations plus measurements are called “shots” and we normally need them in the thousands. They all consume a lot of energy. Hence money. We need a bigger budget Here’s a law for you, “If we check every 2 years for the next 2 decades, qubits will keep being more expensive to run for computation than transistors” The funding is going exponential but the best we can show for it so far is . However, we cannot afford to slow down, we are close to a great use-case. 433 qubits For now, progress in quantum computers shall keep growing to match increasing funding to quantum computation but with zero profits, and it will not break even in the same way classical computing has broken even. You know, with some sort of Moore’s law for qubits. Certainly, there are learning curves that take us from poor qubits to better qubits and this curve reduces energy costs hence the budgetary constraints. But at the lowest limit, running perfect simulations of natural models is going to cost us big time and will never be cheaper than classical computation en masse. Classical computers might be approximations of nature, but they show us that we can go very far with approximating nature. However, since we need to sometimes take a critical look at reality if only to break out of approximations that are hitting their limit, then we need to be able to pay for it. Big quantum computational systems like LIGO cost $1.1 billion but it has to be paid if you want to look at the clear night sky and see gravitational waves. No classical computational systems, no matter how cleverly built, will be able to encode everything that is going on in LIGO. The E/acc movement is right, humanity needs to reach for more rungs up the ladder of the Kardashev scale. We need to be able to consume hundreds of times more usable, carbon-free energy than we use today. This energy is what truly pays for everything including quantum computations and while the technology may not be good enough for making profitable desktop-quantum-PCs, quantum computers will, once every while, give us a view of reality we did not expect. A view that re-ignites our interest in the world. Quantum Computers help nudge humanity and its classical computational approximators ever closer to truer greatness than if we had never discovered the power that is simulating nature with a quantum computational model. Even though it might be the most expensive technology on earth. *** P.S. >> Because running quantum computations is so expensive our fears of futuristic quantum computers hacking our email are unfounded. Social hacking will remain a cheaper way to hack your email, quantum computers, not so much. And if your password is PASSWORD please do not blame quantum computers.

The Most Expensive Technology on Earth 

পৃথিবীর সবচেয়ে ব্যয়বহুল প্রযুক্তি

Stories, News & Mentions about TRANSTUTORS

Transtutors’s Stories on HackerNoon