When OpenAI’s ChatGPT took the world by storm about a year ago, it sounded the alarm across the tech industry. Companies that weren’t already all-in on AI knew they needed to accelerate, and one of them was Google. Since then, the tech giant has launchedGoogle Bardand added countless AI tools to various services, all while working on its own all-encompassing AI project to rival OpenAI’s latest GPT advances, codenamed Gemini. While it was originally speculated that Gemini could go live before the end of 2023, it now seems like the project has hit a snag.
According toThe Information, sources close to the Gemini project now claim that Google CEO Sundar Pichai has put the brakes on upcoming launch events. While smaller editions of Gemini have already been tested, the large language model is still being developed, and it is supposedly facing challenges. Specifically, engineers believe multilingualism needs more improvement before Gemini can be launched — while it is performing well in English, its performance is lagging in other languages.
If a delay is imminent, this means that other Google products and services that were slated to benefit from Gemini (such as Bard) are going to be impacted as well. When it launches, Gemini is expected to come in different sizes, including a lightweight version for mobile. The idea would be to appeal to as many developers as possible.
When details on the Gemini project began to leak, rumors suggested that Google didn’t want it to size up alongside ChatGPT — insiders claimed that the goal wasto surpass the capabilitiesof the OpenAI product. Development has reportedly been focused on combining the text capabilities of large language models and AI image generation to create a more versatile tool. Google has supposedly been using YouTube video transcripts to train Gemini, and the company’s lawyers have been monitoring the process to ensure training materials do not violate copyright regulations.
To bolster its efforts,Google welcomed back its co-founder Sergey Brinto take a hands-on approach to the company’s AI efforts and help spearhead Gemini. Other notables who have hopped on board include Paul Barham, senior Google researcher, and Tom Hennigan of DeepMind. When Gemini is complete, it could be used to power Google’s suite of products, from Bard to Workspace’s Duet AI.
According to insiders, the earliest Gemini could debut is 2024, but nothing has been set in stone. Meanwhile, OpenAI is already working on surpassing its GPT-4 model and developing GPT-5. By the time Gemini is rolled out, it will be interesting to see if it can still help Google gain ground on OpenAI. While the potential is certainly there, the delay could be a costly setback if Google’s competitor uses the extra time to its advantage.
