DailyPost 2411
AUTOGPT ARCHITECTURE
While we have hardly become familiar the ChatGPT architecture; what we can derive out of it and which manner, there is another super upgrade, which can make you thinking; the AutoGPT architecture. The pace of tech and its applications have reached a level, that weeks are turning out to be centuries, if you miss out on it, you can literally be thrown out of the race. The proverbial pace of technology, which we have talking for decades, has become a reality now. This AI revolution would sure be insane, if we go by the trends of the last few months. The nimbleness of change of the whole architecture of the software and emanating applications can throw the world in a disarray.
The next evolution in the AI revolution is AutoGPT, which can broadly be described as ChatGPT on steroids. All you have to is to use the GPT-4 API and lo and behold it carries out the tasks without inputs. You start with the task, you may call it a prompt and then the prompts keep feeding back into itself and it creates an infinite loop till the time your desired task is done. If this were to work the way it is being demonstrated, which would most likely be the case, them a profession like prompt engineering would come to an end even before it has actually started.
AutoGPT would give you what has been in your wish list since the time you would have got into any profession. One of the first insane examples are; AutoGPT was tasked with the objective to create a website using React. The restriction was to make a form, add a header “made by AutoGPT” and then change the background to being blue. The AutoGPT starts working and step by step you see the prompts, output and the rationale. The method of working and reasoning and the look and feel of the work is so human. It doesn’t need human intervention any more.
We are going to see the rise of AI agents. One such one is called AgentGPT. Someone is creating a HustleGPT tasked with creating a StartUp with only a hundred dollars funding. Response has the details of all tasks neatly planned and reasoning thereof. Most of them cite the references, so that the authenticity is well proven. You can have your own AgentGPT, for the specialized task you have on mind, say YouTubeGPT. Research will fall prey in no time. The quantum of work in some areas are already proven to be cut down by 80%. We might not be able to carry the baggage of mediocrity and drudgery any further. It’s the rise of these AI tools that are really going to wipe out certain industries and certain parts of the workforce because it would do things much more effective and much quicker compared to humans.
HOW MUCH CAN WE ADAPT TO THIS CHANGE WOULD DECIDE AI’S AND OUR FATE?
Sanjay Sahay