Google and Anthropic build AI to work in programs for humans
An AI that can use software designed for humans can sound like something out of science fiction, especially considering the current trends in the development of large language models, which lead them up the path of smart assistants wielding their own sets of tools. Yet, we have Google and Anthropic, developers of Gemini and Claude LLMs, respectively, breaking ice in this direction, with the former just known to work on the project, and the latter having already released a public beta of Claude 3.5 Sonnet, that can do the tricks in question.
Google’s AI capable of human-style interaction with software
So far, there aren’t many details known about Google’s AI that can work in programs designed for human use. What’s known is that:
- the name of the LLM is Jarvis;
- it will be able to act like people do in the Chrome browser environment.
With lack of information from Alphabet, the part of the web populated by AI enthusiasts swells with ruminations about why the giant is making Jarvis, and to what purpose. The obvious take is automation of repetitive tasks, the less obvious – two cents from us – is another layer in the foundation of a brand new AI-driven Chrome OS.
Anthropic’s Claude 3.5 Sonnet and its Computer Use function
Anthropic, an LLM developer that was incorporated in 2021 as a Public Benefit Company by former OpenAI employees concerned about the harm AIs can do, released Claude 3.5 Sonnet with a Computer Use function on October 22, 2024. The company has been more open about how the system works and what it can do:
- Claude interprets what’s on screen (snapshots), understands the context;
- and moves the mouse pointer or inputs text in the fields accordingly, following the commands received.
What are the implications of AIs using programs for humans?
Anthropic, being ahead in this race, claims that the goal of building an AI that can use software designed for people is to automate repetitive tasks like entry of data into numerous fields and multiplying lines of code (with minor adjustments). On the one hand, it makes sense to assume that AI, operating on the software level, can connect thereto through interfaces (APIs) reserved for the purpose of enabling direct program-to-program communication. On the other hand, a lot of applications that people use on a daily basis simply don’t have those interfaces, so the new capabilities of AIs make sense. Time will tell where this is going, and we’ll keep tabs on the developments, so stay tuned.
As you can see, the world of software is changing, and to stay current with those changes, you may want to keep all your programs fresh and updated. After all, no one knows when your favorite application will be enhanced with an interface enabling direct communication with an AI. Download and install Informer, a lightweight and totally free software updater, to have all programs on your PC updated automatically:
