AutoGPT tryout

AutoGPT has recently been trending on social media. It can obtain information from external networks and use this information to train and optimize GPT models. This feature has opened up people’s imagination, but how "auto" is it?

How to install AutoGPT:

To test AutoGPT’s true capabilities, I deployed the tool on an AWS server and tested it myself. I gave AutoGPT a task: to write a feasible landing plan for promoting the Our ChatGPT tool website (http://tenten.co/ai) to more users.

I spent a long time trying various methods but ultimately was not successful. I asked AutoGPT if it had any suggestions, but it did not give me a satisfactory answer. Although AutoGPT can obtain information from external networks, its performance in solving practical problems is unsatisfactory.

Finally, unable to wait any longer, I said:

please give me a result according to current info you’ve got

It respond:

I apologize, but I have not received any new information from you since my last response. Please let me know if there is anything specific you would like me to do or if you have any questions.

Overall, I believe there is still much room for exploration in developing this technology. We can obtain and analyze external information through predefined processes and actions that can be applied in many fields. However, to achieve commercial applications, we need customized development based on specific business scenarios to fully realize the value of AutoGPT.

AutoGPT has not yet reached the level of commercial application; however, with continuous technological advancements, we have reason to believe that it will become increasingly intelligent and "auto".

Below is the entire process, which involves going through various websites searching for content then analyzing them before storing them locally – repeating this cycle.