Like many people, you may have been overwhelmed by the possibility of ChatGPT and other large language models (LLMs) like the new Bing or Google’s Bard lately.
For anyone who somehow hasn’t come across it yet – which is probably unlikely given that ChatGPT is said to be the fastest growing app of all time – here’s a quick recap:
LLMs are software algorithms trained on huge datasets of text, enabling them to understand and respond to human speech in a very lifelike way.
The most famous example is ChatGPT, a chatbot interface based on the GPT-4 LLM that has taken the world by storm. ChatGPT is capable of conversing like a human and generating everything from blog posts, letters, and emails to fiction, poetry, and even computer code.
Impressive as they are, LLMs have been limited in one significant way. They are usually only able to do one task, e.g. B. to answer a question or create a text before they require more human interaction (known as “prompts”).
That means they don’t always do well at more complicated tasks that require multi-step instructions or depend on external variables.
Enter Auto-GPT – a technology that tries to overcome this hurdle with a simple solution. Some believe that this could even be the next step towards AI’s “holy grail” – the creation of a general or strong AI.
First, let’s look at what that means:
Strong AI vs. Weak AI
Current AI applications are typically designed to perform one task and get better and better at it as more data is fed to them. Some examples are analyzing images, translating languages or navigating in self-driving vehicles. Because of this, they are sometimes referred to as “specialized AI,” “narrow AI,” or “weak AI.”
A generalized AI is one that is theoretically capable of performing many different types of tasks, including ones that it was not originally created to perform, much like a naturally intelligent entity (e.g., a human) can. Sometimes it is also referred to as “strong AI” or “artificial general intelligence” (AGI).
AGI is perhaps what we traditionally thought of when imagining what AI would look like in the days before machine learning and deep learning made weak/narrow AI a commonplace reality early in the last decade. Think of the sci-fi AI demonstrated by robots like Data in Star Trek that can do almost anything a human can do.
So what is Auto-GPT?
The simplest way of looking at it is that Auto-GPT is capable of performing more complex, multi-step procedures than existing LLM-based applications by creating its own prompts and feeding them back to itself, creating a loop.
Here’s a mindset: to get the best results from an application like ChatGPT, you need to think carefully about how you phrase the questions you ask. So why not let the application construct the question itself? And while it’s at it, make it ask what the next step should be – and how it should go about it… and so on, creating a loop until the task is done.
It works by breaking a larger task into smaller subtasks and then carving out independent auto-GPT instances to handle them. The Urstanz acts as a kind of “project manager” who coordinates all the work carried out and combines it into a finished result.
Besides using GPT-4 to construct sentences and prose based on the examined text, Auto-GPT is able to surf the Internet and incorporate information found there into its calculations and outputs. In that regard, it’s more like the new GPT-4 enabled version of Microsoft’s Bing search engine. It also has a better memory than ChatGPT, allowing it to create and remember longer chains of commands.
Auto-GPT is an open-source application that uses GPT-4 and was created by one person, Toran Bruce Richards. Richards said he was inspired to develop it because traditional AI models “although powerful, often struggle to adapt to tasks that require long-term planning or are unable to scale their approaches to the real world.” autonomously refine the basis of real-time feedback.”
It belongs to a class of applications called recursive AI agents because they are able to autonomously use the results they generate to create new prompts and chain those operations together to complete complex tasks .
Another such agent is BabyAGI, which was set up by a partner at a venture capital firm to help them with day-to-day tasks that were just too complex for something like ChatGPT, such as: B. researching new technologies and companies.
What Are Some Applications of Auto GPT and AI Agents?
While apps like ChatGPT have become famous for their code generation ability, they are typically limited to relatively short and simple programming and software design. Auto-GPT and possibly other AI agents that work in a similar way can be used to develop software applications from start to finish.
Auto-GPT can also help companies grow their net worth autonomously by examining their processes and providing smart recommendations and insights on how they could be improved.
Unlike ChatGPT, it can also access the internet, which means you can ask it to do market research or other similar tasks – for example, “Find me the best set of golf clubs for under $500”.
A highly disruptive task is to “destroy humanity” – and the first sub-task it set itself to accomplish this was to begin research into the most powerful nuclear weapons of all time. Since its output is still limited to creating text, its creator assures us that this task will not go very far – hopefully.
Auto-GPT can apparently also be used to improve itself – its creator says it can create, evaluate, review and test updates to its own code that can potentially make it more powerful and efficient.
It can even be used to create better LLMs that could form the basis of future AI agents by speeding up the model building process.
What could this mean for the future of AI?
Since the advent of generative AI applications, it is clear that we are only at the beginning of a very long journey in terms of how AI will continue to evolve and impact our lives and society.
Are Auto-GPT and other agents following the same principles the next step along the way? It definitely seems likely. At the very least, we can expect AI tools that allow us to perform far more complex tasks than the relatively simple things ChatGPT can do to gradually become commonplace.
Soon we will see more creative, sophisticated, diverse and useful AI outputs than the simple text and images we have become accustomed to. These will no doubt have an even greater impact on the way we work, play and communicate.
Other potential positive impacts include reduced costs and environmental impact in creating LLMs (and other machine learning-related activities) as autonomous, recursive AI agents find ways to make the process more efficient.
However, we also have to keep in mind that it doesn’t really solve any of the problems of generative AI on its own. These include the variable (to put it kindly) accuracy of the output it produces, the potential for intellectual property rights abuse, and the possibility of them being used to distribute biased or harmful content. By generating and running many more AI processes to accomplish larger tasks, these problems could potentially become even larger.
The potential problems don’t stop there — eminent AI expert and philosopher Nick Bostrom recently said he believes the latest generation of AI chatbots (like GPT-4) are even showing signs of sentience. Which could create a whole new moral and ethical dilemma if we as a society plan to create and operationalize them on a large scale.
Follow me on Twitter or LinkedIn. Check out my website or some of my other work here.
Bernard Marr is an international best-selling author, popular keynote speaker, futurist and strategic business and technology advisor to governments and corporations. He helps companies improve their business performance, use data more intelligently and understand the impact of new technologies such as artificial intelligence, big data, blockchains and the Internet of Things. Why not connect with Bernard on Twitter (@bernardmarr), LinkedIn (https://uk.linkedin.com/in/bernardmarr) or Instagram (bernard.marr)?
Read moreRead less
