Could I simply ask ChatGPT to select a personalised present for my cousin Johnny and arrange for its timely delivery? Unfortunately, the happy chatbot was unable to assist, stating that it “can’t make purchases or handle shipping directly.”
The Exponential Rise
Since ChatGPT’s inception two years ago, we have witnessed a flurry of AI features and solutions that promise to relieve us of routine work.
However, save from the occasional built-in chatbot on a shopping website or app, they are still unable to purchase gifts. But by Christmas of next year, everything might have changed. The emergence of AI agents, or bots that can act on your behalf in the real world, is something that many experts predict will happen in 2025.
Agents Are Here
An bot is capable of more than just recommending a Santa suit store. It can purchase it on your behalf and deliver it right to your home. Additionally, the idea behind “agentic AI” is that groups of agents will collaborate.
You would issue a prompt to your group of agents: “This year, I’m preparing Christmas dinner. Make sure one of my closest Facebook friends is a chef, invite them, and ask them to bring the turkey. Without you ever having to do a thing, the agents would take care of everything. Most importantly, bots should be able to collaborate across different websites.
Limited agents are actually already available. According to a survey conducted by developer Langchain, 51% of participants already employ agents in their work. An estimated US$1.8 billion was invested in agent initiatives by venture funds in 2024.
According to Deloitte’s most recent Global Predictions Report, 25% of generative AI-using businesses will begin agentic initiatives by 2025. By 2028, 15% of daily job choices will be made by agents, according to research firm Gartner.
Agent Apocalypse ?
Additionally, we are witnessing the emergence of consumer agents. These systems have the ability to automate a variety of browser-based actions, including shopping. The company that created the well-known Claude generative bot, Anthropic, unveiled a “computer use” function in October that enables the AI to use a user’s keyboard and mouse to navigate and interact with any website.
Leon Furze, a specialist in education, developed a demonstration that uses a computer to automatically navigate to a learning management system, open the assignment page, compose the assignment’s text, and press the submit button. From a single text prompt, everything is completed automatically.
Google Deepmind’s equivalent, Project Mariner, was published more recently and similarly enables an AI to explore and perform tasks in the Chrome browser on its own. Both of these technologies are still in their early stages, and only a select group of reliable testers have access to Project Mariner. However, they foreshadow future developments.
Currently, it is not possible to automate your Christmas shopping with either of these tools—at least not readily. What would be required, then, to create an agent that is actually helpful for Christmas shopping?
Tech Is Available
A shopping agent’s technological aspect is really simple. I might wish to provide a prompt like “Send photo gifts to my family in England” as a user. Choose some amusing family pictures from my phone, look for a website that sells photo presents, order suitable presents for every member of the family, and mail them using my address book.”
Is Moose Like Robot Feet, The Future Of Robotics Exploration Or Dangerous ? Here are 9 Points To Ponder !
https://webnewsforus.com/moose-like-robot-feet/
Multiple agents would be needed to carry out this task: one to locate the images, another to locate the shopping websites, a third to customise the presents, a credit card agent to purchase them, and an address-finding agent.
There is no technological reason why this cannot be accomplished today, whether it be through Project Mariner, computers, or another agent platform.
Trust Issues
Nonetheless, there are two major obstacles to the usefulness of agents. Trust is the first and most obvious. Would you give your credit card information to an agent?
Hallucinations, in which the AI simply makes things up because it doesn’t know the answer, are still an issue even after two years of advancements since ChatGPT.
According to a recent study, 52% of AI-generated responses to coding questions contained errors, even in AI programming, one of the most common and beneficial applications l. Uncle Joe receives Aunty Molly’s gift after the AI makes just one mistake. And let’s hope it’s a non-threatening mistake like a poorly matched present or a breach of your bank account information.
What agents must understand
The second, less evident obstacle is that bots must comprehend context in order to be helpful. Context is crucial, even for something as basic as purchasing gifts. I’ve known my mother’s preferences for years.
I will do far better than a generic response, but I won’t always get it right. This information is typically implicit, and ChatGPT cannot possibly retrieve the extensive record of human encounters that result in that ideal gift.
That being said,bots are already gathering user data. Simply ask ChatGPT, “What do you know about me?” to demonstrate this.
The answer may surprise you, depending on your settings. Maybe the programs we use on a daily basis may eventually know enough about our family and ourselves to enable completely automated Christmas shopping.
