ChatGPT is just the “infancy” stage of AI. The startup’s ambition is for artificial intelligence to replace human thinking and complete all tasks without human intervention.
Automated AIs like Auto-GPT differ from ChatGPT in that they don't require human intervention for each response. Photo: Bloomberg .
Generative AI is a hot topic in the tech world recently. ChatGPT is a surprising success from a small AI startup like OpenAI. Basically, using the ChatGPT tool is very simple. Users just need to type in any question, the AI chatbot will quickly return useful answers and sound like a real person.
Of course, ChatGPT still has certain shortcomings such as the data set from 2021 is outdated, not updated to the present time, the information source lacks authenticity, there is no function to remember previous conversations...
The future of AI is to think like a real person
Therefore, AI experts are expecting a new development stage of AI chatbots, transforming into an "AI entity" (AI agent), a software capable of perceiving the environment around it through sensors that replace human eyes and ears, then reacting to the environment through parts that react similarly to hands, feet, etc.
The “AI entity” will be linked to large language models (LLMs) like the system behind ChatGPT to perform a given task set by a human. The software will then perform a series of different tasks to complete that task.
In the future, these AI agents will operate autonomously across the entire Internet environment, taking on many different tasks. Experts who are excited about this idea all believe that they are the closest achievement to achieving Artificial General Intelligence (AGI).
This is a computer system that has the ability to understand and learn all human thinking. Therefore, AGI can complete most of the intellectual tasks that humans can do, even better. The advanced point of AGI is that all tasks are performed completely automatically, without any human intervention.
After ChatGPT will be the era of AI automation. Photo: TechSmart
One notable AI agent is Auto-GPT, an open-source project founded by game publisher Bruce Richards and hailed as the latest breakthrough to change the AI field.
As an automated AI model, Auto-GPT uses GPT-3.5 or GPT-4 to complete tasks independently. This technology can respond automatically, performing tasks automatically to complete a goal given by the user. Auto-GPT differs from ChatGPT in that it does not require human intervention in each response.
Current AI is just a child's toy
Recently, many users have also loved sharing the responses given by Auto-GPT or AI models. One Twitter user took a screenshot of the results when asking the AI to conduct a market research and summarize the best headphone products on the market.
Sully users even imagined a company selling fake shoes and asked the AI to do market research for the product. The AI quickly listed five formidable competitors and analyzed their strengths and weaknesses. Recently, the CEO of the virtual assistant plugin company HyperWrite even posted a demo of their AI Personal Assistant that can order pizza online from Domino's Pizza.
But unlike ChatGPT, Auto-GPT is not for the masses. The generative AI model requires programming skills to set up basic settings and access OpenAI's application programming interface (API).
A Discord group called HustleGPT is made up of startup founders who are early adopters of AI, using it for everything from logo design to website development to marketing, says Dave Craige, who founded the group.
However, Bloomberg suggests that users should be cautious when using these tools. The news site tested AgentGPT's plan to invest $100 into $1 million in 10 years in the Total Stock Market Index Fund. The result was a plan that still had many flaws.
For example, AgentGPT claims that if you invest $500/month at a 7% annual return, you can get exactly $1,031,906 in 10 years. However, the online calculator calculator.net estimates that you will get about $86,000 in the worst case.
Software like Auto-GPT and BabyAGI are still in their infancy, according to Matt Schlicht, CEO of startup Octane AI. “So if you look at them as a single product, they’re like kids’ toys,” he said.
For AI tools to become more advanced, AI expert Henry Shevlin at the University of Cambridge believes they need to be tested in real-world situations, and laws governing the use of AI need to be quickly put in place.
According to Zing
Source
Comment (0)