Google, a subsidiary of Alphabet, is developing artificial intelligence technology that can control browsers to complete tasks such as research and shopping. It plans to showcase this new AI product in December, similar to a product announced by Anthropic last week.
According to The Information yesterday, Google will showcase a product codenamed "Project Jarvis" in December. This product, also known as a "computer usage agent," aims to take over users' browsers to help consumers complete various daily tasks such as collecting research, purchasing products, or booking flights. Google will also release its next-generation flagship product Gemini Large Language Model in December to power Jarvis.
The next competitive direction in the AI industry: AI Agents
Previously, it was reported that OpenAI, supported by Microsoft, also hopes that its model can autonomously browse web pages with the help of "CUA" (using computer proxies) and take action based on research results.
Anthropic and Google are attempting to further advance this proxy concept through software that directly interacts with personal computers or browsers.
There are also key differences in the agency products of the two companies. Anthropic stated that its product can be installed on different applications on a computer, while Jarvis can currently only operate a browser and has been "customized" for Google's Chrome browser.
Jarvis' release schedule indicates that although Google has accumulated some basic research in AI technology, it is clearly still catching up with its competitors. At present, Google is still developing AI with so-called "reasoning ability", and OpenAI had already launched this feature as early as September.
Google's Gemini chatbot is seriously lagging behind OpenAI's ChatGPT competition, leading companies to turn to OpenAI's large language model LLMs, which also makes it difficult for Google's Gemini model to catch up. In order to improve the efficiency of AI development, Google merged the team responsible for Gemini chatbot into its main AI team DeepMind last week.
Currently, AI developers have regarded "agents," which are AI systems capable of completing complex tasks without human supervision, as the next stage in the industry. Salesforce, Microsoft, and Workday have all purchased LLMs from OpenAI and other companies and are competing to use this technology to develop AI agents.
The largest artificial intelligence company is researching models that can complete the tasks described by The Information. Microsoft will allow users to discuss the webpage they are browsing with it. It is expected that next year Apple Smart will understand the content on the user's screen and perform certain operations in various applications.
Insiders say that at least for now, Jarvis' target users are those who want to automate daily tasks on their web pages. At Google's developer conference this spring, CEO Sundar Pichai hinted that future versions of Gemini will be able to perform multiple actions autonomously, such as helping users return a pair of shoes.
Jarvis is currently running at a relatively slow speed
Google may first release the product to a small number of early testers to help identify and fix its shortcomings. The agent is currently running relatively slowly because the model needs to think for a few seconds before taking each action.
In addition, Google also needs to access customers' privacy information such as login passwords and credit card information in order to access different websites to complete tasks or make purchases according to customers' requirements.
Analysts point out that Google needs to convince people that its AI agents can securely process their personal data, which is necessary for it to perform its tasks.
In addition, LLMs also have some common vulnerabilities, such as the possibility of generating incorrect answers. Previously, Google used LLMs driven conversational answers in its search engine, which resulted in many obvious errors.
The name Jarvis comes from a character in Marvel Comics' Iron Man ', symbolizing' Just A Rather Very Intelligent System '(J.A.R.V.I.S.), meaning' just a pretty smart system '.
In the Marvel Cinematic Universe, Jarvis was introduced as a sophisticated AI and appeared in the 2008 film Iron Man, 2010 film Iron Man 2, 2012 film The Avengers, and 2013 film Iron Man 3. Jarvis, as the assistant to Tony Stark played by Robert Downey Jr., is responsible for connecting to any computer terminal: controlling Stark's house and the internal systems of the Iron Man suit. Jarvis is able to have quite cultured conversations with Stark and often jokes about the recklessness and arrogance of his creator.