HyperWrite, a startup well-known for its generative AI writing extension, has introduced an experimental AI that can do all your online research for you. Matt Shumer, the Co-Founder of HyperWrite, has developed an AI agent that can browse the internet and interact with websites much like a human user. This represents a significant advancement in the field of artificial intelligence since it moves the technology closer to being able to carry out difficult tasks more independently.
Source: LinkedIn
CEO Matt Shumer gave VentureBeat a closed presentation through Zoom of the new AI agent’s skills. The agent accessed the Domino’s Pizza website and began making an order thanks to HyperWrite’s Chrome addon. It even looked for an address and zip code to finish the deal. Although the demo ended before a credit card number would have been entered, Mr. Shumer confirmed that the capability would only be enabled for select test users at first. The agent is designed to act as a personal assistant by performing simple web tasks automatically when requested.
Also Read: Meet AgentGPT, an AI That Can Create Chatbots, Automate Things, and More!
Current AI Applications’ Restrictions
Most GPT API-based programs now only allow for single-session usage, preventing the model from remembering data from earlier interactions. This restriction results from the quantity of information, or the number of “tokens,” needed for each query and GPT’s propensity for hallucinations. As token counts increase, this issue becomes more obvious.
Also Read: OpenAI Co-Founder & Chief Data Scientist On the Potential of AGI
Personal Assistant Agent for HyperWrite
Shumer refers to the personal assistant agent from HyperWrite as “the first user-friendly AI agent that can operate a browser like a human.” Meanwhile, experts issued a warning that software displaying human-level web capabilities can also be vulnerable to well-known human flaws like phishing, hacking, or fraud.
Also Read: GPT-4 Capable of Doing Autonomous Scientific Research
Fears about the Abuse of AI Agents
According to Shumer, we understand how crucial it is for generative AI to act and carry out activities independently. However, potential issues could arise when such an AI is abused to hack into business websites, distort communications, or disseminate spam on social media platforms. Although these things already occur with earlier technologies, it is vital to avoid facilitating other people’s nefarious acts.
The AI Agent Explosion Trend
The popularity of artificially intelligent agents is skyrocketing! If you follow AI on Twitter, you probably see a lot of mentions like AutoGPT and BabyAGI in your news feed. These AI agents are driven by OpenAI’s GPT-4 and other models. They can generate prompts automatically, a capability known as “self-prompting” or “auto-prompting,”. They can also develop and execute prompts depending on an initial input.
A Focus on Safe AI Agent Deployment
Shumer emphasized that his team’s attention is on safety-related concerns. We’re taking our time to do this properly because we want to determine the best course of action. How can you use it in a way that genuinely protects society as a whole? What about the particular user? You wouldn’t want the AI bot to just randomly order stuff without consulting you, right? You need the ability for it to contact you.
Also Read: Virtual World Created by AI Develops Human-Like Features
Our Say
The ground-breaking AI agent developed by HyperWrite is a significant advancement in the field of artificial intelligence. Their next-gen generative AI agent can browse the web and communicate with websites just like a human. Through this project, HyperWrite is paving the path for more sophisticated and autonomous AI applications. As AI agents proliferate, it will be crucial to concentrate on their responsible and safe deployment to ensure that they contribute to society while reducing any hazards.
