Auto-GPT

 

Recently, artificial intelligence (AI), which is called 'the starting point of general-purpose artificial intelligence (AGI) that can replace all human work', has emerged in Silicon Valley. It is a program called 'AutoGPT'. AutoGPT is an open-source Python (one of the programming languages) program created by a developer called 'Significant Gravitas' based on OpenAI's language model GPT-4. Came to the hub.

​

​Generative AI service 'AutoGPT' appeared

AI start-up Significant Gravitas unveiled a generative AI service called AutoGPT on the 30th of last month.

After the release of AutoGPT, the tag '#AutoGPT' ranked first in Twitter trending, and in star acquisition, an indicator of user interest on Github, it exceeded 30,000 in 10 days of publication and has reached 95,000 so far. received.

AutoGPT is a program that is evaluated as closest to general-purpose artificial intelligence (AGI). When users set goals, it automatically learns how to deliver results. AI achieves its goals by changing the way it solves problems on its own without human intervention.

For example, if you present a goal of 'creating the most viewed videos on YouTube', AutoGPT automatically searches for the most popular videos on YouTube and categorizes them by subject. It then explains in detail how to shoot.

Examples of using AutoGPT like this have recently been uploaded on YouTube. Hundreds of videos have been uploaded in the past weekend alone.

What is AutoGPT?

AutoGPT is an open source Python (one of the programming languages) program based on OpenAI's language model GPT-4, and was released on GitHub, a code sharing platform for developers, on the 30th of last month.

Here, Auto is an abbreviation of Autonomous, which means autonomous. The name came from the meaning that when the user gives a task, the intermediate tasks are automatically performed. AutoGPT consists of several language model agents such as ChatGPT. Each agent performs a number of secondary tasks required to carry out a command. For example, one agent separates tasks to be done, another agent performs a set task, and so on. In addition, some agents evaluate the work performed by other agents. It is a structure that repeats this autonomous process by itself and advances step by step toward the goal suggested by the user.

AutoGPT (AutoGPT) is when a user sets a goal, AI automatically searches the Internet, explores ways, and presents results. AI achieves its goal by changing the way it solves problems on its own without human intervention. Along with the evaluation that it is the early appearance of AGI that can coexist with humans, some people say, “I am afraid of the rapid development of AI.”

To summarize, AutoGPT is an AI chatbot that generates results using the 'autonomous iterations' function that corrects mistakes on its own. It is based on 'GPT-4'. All you need is a separate memory space for storing information and the versatility to search, store and edit data.

​

AutoGPT Developer

Auto GPT was not developed by Open AI, which created 'Chat GPT'. A developer called 'Significant Gravitas' made it based on OpenAI's language model GPT-4, and on the 30th of last month, it was uploaded to GitHub, a code sharing platform for developers.

Since then, numerous developers have shared their cases of using AutoGPT on Twitter, etc., and rumors have spread. For example, if you set a goal of 'making the most viewed videos on YouTube', AutoGPT will automatically search for the most popular videos on YouTube and categorize them by topic. It is also an expression that explains in detail how to shoot. One developer wrote on his Twitter, “I instructed AutoGPT to create an app for me, and AutoGPT figured out that there was no development-related program on my computer, searched for the program, installed it, and created the app I instructed.” .

 

Auto GPT VS Chat GPT

The basic work performance of AutoGPT is similar to that of OpenAI ChatGPT. AutoGPT can also do most of the things that ChatGPT can do, with the biggest difference being that AutoGPT generates prompts autonomously. The application itself creates the prompts it needs to complete the task, and carries out them.

AutoGPT is unique in that it includes an AI agent to make autonomous decisions. Like personal assistants, these agents are programmed to make decisions and perform tasks according to predetermined rules and goals.

AutoGPT on the GitHub repository   features Internet access for searching and information gathering, long- and short-term memory management, GPT-4 instances for text creation, access to popular websites and platforms, and GPT-3.5-based file storage and summarization. described as having

All you have to do is to request chatGPT, such as debugging code or writing emails. Instead, AutoGPT can ask you to complete much higher-level tasks with fewer prompts than ChatGPT. According to the published demo, you just need to enter sample goal prompts, such as increasing your net worth, growing your Twitter account, or developing and managing multiple businesses. .

AutoGPT is unique in its ability to link together LLM “thoughts” to achieve goals autonomously. It is one of the first examples of GPT-4 to run completely autonomously and sets it apart from other AI tools.

In terms of ability itself, AutoGPT seems to be superior to ChatGPT. However, while ChatGPT is easy to use, AutoGPT requires some computer knowledge to use. AI experts point out that ChatGPT's superiority lies in its easy user experience rather than its ability to perform.

What can you do with Auto GPT?

As mentioned earlier, AutoGPT is a new artificial intelligence tool with the ability to “prompt itself” to achieve a given goal while retaining the ability to manage internet, memory and code writing and execution. It is still in its infancy, but AI enthusiasts People are already flocking to AutoGPT to explore its various noteworthy applications.

Simply put, AutoGPT can generate high-quality human text, summarize large amounts of text, answer questions, and even translate between languages. AutoGPT can also be used for text classification, sentiment analysis, and named entity recognition. It shows great potential in natural language processing applications such as chatbots, virtual assistants, and language learning tools. AutoGPT can also be used in the marketing research, content creation and software development industries.

​

2. Build the app

The second thing you can do is build your app.

A computer engineer named Varun Maya recently shared his process using AutoGPT on Twitter, which became popular. After that, I actually created a node to create a server,” he wrote. A node is a software platform used to develop network applications.

3. Marketing Research

AutoGPT is an ideal tool for key market research as it allows you to access the internet, remember details and focus on your work.

Twitter user '@sairahul1' instructed AutoGPT to 'research the headphone product market'. We searched for headphone products on the market and made a list of the top five products, summarizing the features and price of each product. Afterwards, AutoGPT performed tasks sequentially while generating prompts on its own, and created the desired report.

4. See the future prospects

A Twitter user named 'Daniel Kraft' suggested that AutoGPT, which has such technology, see the future of medicine and pharmaceuticals. AutoGPT then analyzed emerging technologies and trends in the healthcare industry and selected machine learning algorithms capable of analyzing large healthcare datasets to create task prompts for patient care and decision making.

5. Thinking about the philosophical questions of life

AutoGPT ponders deep philosophical questions.

AutoGPT was recently challenged with the question “What is life?” The results were shared on Twitter by Loopuleasa. About an hour later, for $1, AutoGPT delivered a response. It searched 660,000 characters of posts and wiki topics. After referring to the human concept of various schools of thought, such as , physics, etc., we came to the conclusion that each field has its own unique point of view.

​

How to use Auto GPT?

How to use AutoGPT is not simple. You can download it from Github.

Once VS Code is installed, the user's computer must have version 3.8 or higher of the programming language 'Python' installed, and have a paid account of OpenAI to have a GPT-4 or GPT-3.5 access API key.

API keys must be obtained by applying for the API of Pinecone, an open AI and artificial intelligence distribution platform.

You can use text-to-speech (TTS) and run searches with Google API keys. It is also possible to attach a separate Python plugin. Local cache (LoaclCache) is used as default, but it can be changed to Redis or Pinecon.

AutoGPT provides 'continuous mode'. Running in continuous mode, it performs all operations 100% automatically without user approval. The developer emphasizes that it does not recommend using continuous mode.

Also, since GPT-4 is used, if the amount of work increases, you have to pay API usage fees to OpenAI.

Looking at AutoGPT's task performance, it shows the similarity of how people process business and how they process information. It uses self-reasoning to think of the next task based on the results of the previous task, requiring no human supervision.

Is AutoGPT Free?

AutoGPT offers impressive features, but its high cost is a significant barrier to practical use in production environments. The GPT-4 model that Auto-GPT relies on can be expensive as each step of the operation requires a model call. GPT models are represented as sequences of 'tokens' (language units understood by computers) in the model. The GPT-4 model is priced at $0.03 per 1,000 tokens for presenting messages and $0.06 per 1,000 tokens for results. For example, it costs $14.4 to take 50 steps, each step maximizing an 8K context window.

These costs add up quickly, making the current implementation of AutoGPT unaffordable for many organizations and users. AutoGPT shows great potential, but cost is a significant barrier that needs to be addressed before widespread adoption.

Auto GPT (AutoGPT) limits?

AutoGPT is powerful AI, but it has one major hurdle. Adoption in a production environment is difficult due to high cost. Each step requires an expensive call to the GPT-4 model that maximizes tokens for better inference. The limited set of features provided by AutoGPT, such as web browsing and code execution, narrows its troubleshooting capabilities. Additionally, the inference capabilities of GPT-4 are still limited, further limiting the potential of autoGPT.

Therefore, not all results are 100% reliable. Also, researchers are now targeting Artificial General Intelligence (AGI), which means that AI systems can solve problems for which they are not specially trained, and from this moment on, we need to put in proper safeguards against AI. do.

Evaluation of Auto GPT

Some have evaluated that this auto-GPT paradigm has presented a new direction toward general artificial intelligence (AGI).

AGI refers to an AI level that can completely replace humans in all areas that humans can do beyond the current level of AI that has limited performance and handles independent tasks.

In Silicon Valley, AutoGPT is seen as an early version of super-intelligent AI like AI assistant Jarvis in the movie 'Iron Man'. ZDNet, a technology media outlet, said, “AI that can operate well without human supervision can replace numerous roles currently performed by humans.” Fortune magazine said, “AutoGPT shows that new risks are emerging with the astonishingly rapid pace of innovation.”

One developer said, “If you thought ChatGPT was scary, wait until you hear AutoGPT.” said. “AutoGPT uses GPT-4 and GPT-3.5 to iterate, review, build, and create an entire project at its own prompt.”

​

So far, we have learned about the introduction, functions, and usage of Auto GPT~!!!

​

See below for more details on Auto GPT

https://www.btcc.com/en-US/academy/crypto-basics/about-autogpt

​

In the future, if there are additional features, we will introduce them.

​

thank you!!

​