Did OpenAI Beat Google To The Punch Ahead Of Its Developers Conference?

The word AI was reportedly repeated 121 times during the 100-minute event, commonly referred to as Google I/O -- ‘input’ and ‘output’ concepts in computing or ‘1 and 0’ which denotes ‘true and false’ in computer language.

In a grand showcase of technological prowess, keynote speakers such as CEO Sundar Pichai and Google DeepMind head Demis Hassabis took centre stage to unveil an array of AI-powered marvels. 

The long line-up included Gemma - with new additions to its family of open models: Gem, which enables users to create custom chatbots, Veo -- an advanced video generation model, and Learn LM - a family of educational models.

A Showdown Of AI Assistants

Among the distinguished roster of innovations, one clearly stood out -- Google’s Project Astra. It reminded us of an uncannily similar offering by OpenAI -- GPT-4o, which it announced a day before the Google event.

Coincidence? We think not.

OpenAI's preemptive strike, just ahead of Google's grand unveiling, wasn't enough to catch the tech giant napping. No, Google was armed to the teeth. But more on that later. 

First off, Project Astra is an AI assistant (think Siri) which sees, listens and interprets things around it and gives answers to prompts. It can interpret codes written across a white board, scan the room and help identify a neighbourhood, find missing pieces to a computer equation, all the while reminding you where you forgot your spectacles. All through the device’s camera and microphone.

Project Astra by Google utilises an upgraded version of Gemini Ultra, an AI model designed to rival ChatGPT from March 2023. Similar to C, Gemini is "multimodal." Astra will hit the market sometime later this year.

Offering similar sophistication, in a twist of anticipation, OpenAI’s GPT-4o (‘o’ for ‘omni’ denoting omnipresence) is like ChatGPT on steroids and a lot more enthusiastic, even if it’s generated by a machine. And that’s what makes OpenAI achieve more human-like interaction between individuals and computers, said the company.

On social media, it is being likened to Scarlett Johannsan’s AI character in the movie ‘Her.’

GPT-4o is multimodal, can read people’s expressions and its voice feels more real with a wide dynamic range. During a demonstration, when an OpenAI researcher asked GPT-4o to narrate a dramatic bedtime story about robots and love, the bot literally turned into Mary Poppins.

When OpenAI’s Chief Technology Officer Mira Mureti interacted with GPT-4o in Italian, it translated and responded in English and vice versa. Leading EdTech Duolingo’s share price dropped five points after the announcement.

However, GPT-4o did hallucinate at one point when it said, “Wow that’s quite an outfit you have on” without being prompted. But OpenAI has time only till a couple of weeks to fix its issues till it opens up GPT-4o for public use.

Goodbye Chatbots?

With OpenAI and Google simultaneously unveiling faster conversational AI, are chatbots already outdated just two years after their market debut?

OpenAI's timing of announcing its advanced conversational bot on May 13 is very interesting. Google had set the date for the I/O conference event way back on March 15. This means that OpenAI had a two-month window to announce GPT-4o and other upgrades.

With so much going on in the AI space, where its biggest competitor Microsoft along with its partner firm OpenAI are ahead of the curve, Google is playing catch-up, one or two steps behind.

Is OpenAI's opportunism a testament to “fortune favouring the bold”? It has certainly proven advantageous in the past, particularly for the company. However, with Google now in the mix with its equally formidable tool, is the tech giant suggesting that "better late than never" holds true?