Google Astra faces OpenAI’s GPT-4o for AI supremacy
Amid OpenAI’s GPT-4o release, Google has revealed its AI contender: Project Astra. The AI model has similar features to OpenAI’s flagship, showing the growing rivalry in the AI industry.
Like GPT-4o, Google Astra works like a real-time assistant that identifies objects in a user’s surroundings.
READ: How to make your own AI assistant
It can read and interpret computer code in real-time and even understand symbols on a blackboard. However, are these enough to beat GPT-4o?
The Google Astra demo
We’re sharing Project Astra: our new project focused on building a future AI assistant that can be truly helpful in everyday life. 🤝
— Google DeepMind (@GoogleDeepMind) May 14, 2024
Watch it in action, with two parts – each was captured in a single take, in real time. ↓ #GoogleIO pic.twitter.com/x40OOVODdv
On May 14, 2024, Google’s AI division DeepMind unveiled Project Astra. The word “Astra” is an acronym for “Advanced Seeing and Talking Responsive Agent.”
The video above demonstrates Astra’s capabilities by a representative walking into an office. The user starts by turning on the bot on his phone.
Then, the user asks the bot if it could see anything making a sound. The AI points out the speaker seen in the phone’s camera.
The user asks Project Astra about the name of the speaker’s upper part, and the AI says the component is a tweeter.
The pair then encounters a mug of pencils, and the user asks for an alliteration about these objects. Oregon State University defines an alliteration as:
“Alliteration is the repetition of the same sound at the start of a series of words in succession whose purpose is to provide an audible pulse that gives a piece of writing a lulling, lyrical, and/or emotive effect.”
Later, the user pans to a computer showing a line of code and asks the AI to describe the code, and Astra explains the code’s function.
They look out the window, and Astra describes what is outside. Next, the user asks if it remembers where her glasses were, and the bot answers that it was near an apple on a desk.
READ: PH-inspired game Until Then releases demo
The user then asks Project Astra to explain a diagram on a whiteboard, and then he suddenly points his phone to two cat sketches and a box, asking the AI to describe what it sees.
It replied, “Schrödinger’s Cat.” The latter is a thought experiment on quantum mechanics by Erwin Schrödinger.
Google Astra vs. OpenAI GPT-4o
Gemini and I also got a chance to watch the @OpenAI live announcement of gpt4o, using Project Astra! Congrats to the OpenAI team, super impressive work! https://t.co/yxmJt2xL8v pic.twitter.com/p5BqHjc1xQ
— Michael Chang (@mmmbchang) May 14, 2024
Meanwhile, Google DeepMind research scientist Micheal Chang took a silly jab at OpenAI’s GPT-4o with his latest video.
The clip shows Google Gemini, powered by Google Astra, transcribing the product reveal. Chang posted it on X with the caption:
“Gemini and I also got a chance to watch the @OpenAI live announcement of gpt4o, using Project Astra! Congrats to the OpenAI team, super impressive work!”
However, the OpenAI and Google demos show interesting differences between Project Astra and GPT-4o.
The latter proved that it can translate between two people speaking different languages. Moreover, it can guide someone in finishing homework, making it a promising learning tool.
Arguably, its most important advantage is its more convincing human-like voice. GPT-4o’s voice sounds more natural, adding pauses and intonations found in normal human speech.
Both of these features might make OpenAI’s new flagship more approachable to the general public. Nevertheless, we may only confirm their capabilities on their final releases.
