


Restated: What is generative artificial intelligence, and how is it evolving?
Generative artificial intelligence is the general term for any automated process that uses algorithms to generate, manipulate, or synthesize data, typically in the form of images or human-readable text. It’s called generative because AI creates something that didn’t exist before. This is how it differs from discriminative AI, which differentiates between different types of input. In other words, discriminative AI tries to answer questions like: “Is this picture a rabbit or a lion?” while generative AI responds “Picture me a picture of a lion and a rabbit sitting together. picture" prompt.
Mainly introduces generative AI and its use with popular models such as ChatGPT and DALL-E. We'll also consider the limitations of the technology, including why "too many fingers" has become the dead giveaway of generative art.
The emergence of generative artificial intelligence
Generative artificial intelligence has been around for many years, ever since ELIZA, a chatbot that simulated talking to a therapist, was developed in 1966 at the Massachusetts Institute of Technology (MIT). But years of work in artificial intelligence and machine learning have recently come to fruition with the release of new generative AI systems. People have definitely heard of ChatGPT, a text-based AI chatbot that produces very human-like prose. DALL-E and StableDiffusion also attracted attention for their ability to create vibrant and realistic images based on text cues. We often refer to these systems and others like them as models because they represent attempts to simulate or model some aspect of the real world based on a subset (sometimes a very large subset) of information.
The output of these systems is so incredible that many are raising philosophical questions about the nature of consciousness and worrying about the economic impact of generative AI on human work. But while all of these AI creations are undeniably big news, there may be less going on beneath the surface than some people think. We'll discuss these big questions later. First, let's look at what's going on under models like ChatGPT and DALL-E.
How does generative AI work?
Generative AI uses machine learning to process large amounts of visual or textual data, much of it scraped from the internet, and then determine Which things are most likely to appear near other things. Much of the programming effort in generative AI goes into creating algorithms that can differentiate among the “things” that interest the AI’s creators—words and sentences for a chatbot like ChatGPT, or the visual elements of DALL-E. But fundamentally, generative AI creates its output by evaluating a large corpus of data, and then responds to prompts with something within a probability range determined by the corpus.
Autocomplete—when your phone or Gmail prompts you with what the rest of the word or sentence you’re typing might be—is a low-level form of generative artificial intelligence. Models like ChatGPT and DALL-E simply take this idea to more advanced heights
Training Generative AI Models
The process of developing a model to fit all this data is called training. For different types of models, some basic techniques are used here. ChatGPT uses so-called transformers (that's what the T means). The converter derives meaning from long text sequences to understand the relationship between different words or semantic components and then determines the likelihood that they occur close to each other. These deformers are run unsupervised on a large corpus of natural language text in a process called pre-training (PinChatGPT), and then fine-tuned by humans interacting with the model.
Another technique used to train models is called a generative adversarial network (GAN). In this technique, two algorithms compete with each other. One is to generate text or images based on probabilities obtained from large data sets; the other is discriminative AI, which is trained by humans to evaluate whether the output is real or AI-generated. Generative AI will repeatedly try to "trick" the discriminating AI, automatically adapting to successful results. Once generative AI consistently “wins” this competition, discriminative AI is fine-tuned by humans, and the process begins all over again.
One of the most important things to remember here is that despite the presence of human intervention during training, most learning and adaptation occurs automatically. In order for the model to produce interesting results, many iterations are required, so automation is essential. This process requires a lot of calculations.
Is generative AI sentient?
The mathematics and coding used to create and train generative AI models are quite complex and well beyond the scope of this article. But if you interact with the end result model of this process, the experience is certainly incredible. You can have the Dell-e produce something that looks like a real work of art. You can have a conversation with ChatGPT just like you would with another human being. Did Researchers Really Create a Thinking Machine?
Крис Фиппс — бывший директор отдела обработки естественного языка в IBM, участвовавший в разработке продуктов Watson для искусственного интеллекта. Он назвал ChatGPT «очень хорошей машиной прогнозирования».
Он очень хорош в предсказании того, что люди сочтут последовательным. Он не всегда последователен (в большинстве случаев так и есть), но это не потому, что ChatGPT его «понимает». Верно обратное: люди, потребляющие выходные данные, действительно хорошо умеют делать любые неявные предположения, которые нам нужны, чтобы сделать выходные данные значимыми.
Фиппс, который также является комиком, сравнил это с обычной импровизационной игрой под названием MindMeld.
Два человека думают о слове, а затем одновременно произносят его вслух - вы можете сказать «ботинок», я говорю «дерево». Слова мы придумали совершенно независимо, и изначально они не имели никакого отношения друг к другу. Следующие два участника берут два слова и пытаются найти, что между ними общего, произнося их вслух. Игра продолжается до тех пор, пока оба участника не произнесут одно и то же слово.
Может быть, оба человека сказали «лесоруб». Это может показаться волшебным, но на самом деле мы используем человеческий мозг, чтобы рассуждать о вводимых данных («ботинок» и «дереве») и находить связи. Мы занимаемся пониманием, а не машинами. В ChatGPT и DALL-E происходит гораздо больше, чем люди признают. ChatGPT может написать историю, но нам, людям, нужно приложить немало усилий, чтобы придать ей смысл.
Тестирование пределов компьютерного интеллекта
Люди могут дать этим моделям ИИ некоторые подсказки, которые сделают точку зрения Фиппса совершенно очевидной. Например, рассмотрим следующую загадку: «Что тяжелее: фунт свинца или фунт перьев?» Ответ, конечно, заключается в том, что они весят одинаково (один фунт), хотя наш инстинкт или здравый смысл могут подсказать нам, что они весят одинаково (один фунт). что перья светлее.
ChatGPT правильно ответит на эту загадку, и вы можете подумать, что он делает это потому, что это холодный, логичный компьютер, в котором нет «здравого смысла», который его сбивает. Но это не то, что происходит за кулисами. ChatGPT не логически обдумывает ответ; он просто выдает результат на основе прогнозов на вопрос о фунте перьев и фунте свинца. Поскольку его обучающий набор содержит кучу текста, объясняющего головоломку, он собирает версию правильного ответа. Однако, если вы спросите ChatGPT, весят ли два фунта перьев больше, чем один фунт свинца, он с уверенностью ответит вам, что они весят одинаково, потому что, исходя из его обучающего набора, это по-прежнему наиболее вероятный результат для вывода в подсказку о перья и свинец.
The above is the detailed content of Restated: What is generative artificial intelligence, and how is it evolving?. For more information, please follow other related articles on the PHP Chinese website!

Upheaval Games: Revolutionizing Game Development with AI Agents Upheaval, a game development studio comprised of veterans from industry giants like Blizzard and Obsidian, is poised to revolutionize game creation with its innovative AI-powered platfor

Uber's RoboTaxi Strategy: A Ride-Hail Ecosystem for Autonomous Vehicles At the recent Curbivore conference, Uber's Richard Willder unveiled their strategy to become the ride-hail platform for robotaxi providers. Leveraging their dominant position in

Video games are proving to be invaluable testing grounds for cutting-edge AI research, particularly in the development of autonomous agents and real-world robots, even potentially contributing to the quest for Artificial General Intelligence (AGI). A

The impact of the evolving venture capital landscape is evident in the media, financial reports, and everyday conversations. However, the specific consequences for investors, startups, and funds are often overlooked. Venture Capital 3.0: A Paradigm

Adobe MAX London 2025 delivered significant updates to Creative Cloud and Firefly, reflecting a strategic shift towards accessibility and generative AI. This analysis incorporates insights from pre-event briefings with Adobe leadership. (Note: Adob

Meta's LlamaCon announcements showcase a comprehensive AI strategy designed to compete directly with closed AI systems like OpenAI's, while simultaneously creating new revenue streams for its open-source models. This multifaceted approach targets bo

There are serious differences in the field of artificial intelligence on this conclusion. Some insist that it is time to expose the "emperor's new clothes", while others strongly oppose the idea that artificial intelligence is just ordinary technology. Let's discuss it. An analysis of this innovative AI breakthrough is part of my ongoing Forbes column that covers the latest advancements in the field of AI, including identifying and explaining a variety of influential AI complexities (click here to view the link). Artificial intelligence as a common technology First, some basic knowledge is needed to lay the foundation for this important discussion. There is currently a large amount of research dedicated to further developing artificial intelligence. The overall goal is to achieve artificial general intelligence (AGI) and even possible artificial super intelligence (AS)

The effectiveness of a company's AI model is now a key performance indicator. Since the AI boom, generative AI has been used for everything from composing birthday invitations to writing software code. This has led to a proliferation of language mod


Hot AI Tools

Undresser.AI Undress
AI-powered app for creating realistic nude photos

AI Clothes Remover
Online AI tool for removing clothes from photos.

Undress AI Tool
Undress images for free

Clothoff.io
AI clothes remover

Video Face Swap
Swap faces in any video effortlessly with our completely free AI face swap tool!

Hot Article

Hot Tools

SecLists
SecLists is the ultimate security tester's companion. It is a collection of various types of lists that are frequently used during security assessments, all in one place. SecLists helps make security testing more efficient and productive by conveniently providing all the lists a security tester might need. List types include usernames, passwords, URLs, fuzzing payloads, sensitive data patterns, web shells, and more. The tester can simply pull this repository onto a new test machine and he will have access to every type of list he needs.

SublimeText3 English version
Recommended: Win version, supports code prompts!

Zend Studio 13.0.1
Powerful PHP integrated development environment

SublimeText3 Mac version
God-level code editing software (SublimeText3)

Notepad++7.3.1
Easy-to-use and free code editor
