As we know, OpenAI was the company that turned the public’s gaze to generative AI and LLMs (large language models). However, it wasn’t the first company to develop LLMs, as other companies were a few steps ahead of the game. Google is an example, but its co-founder revealed something about the company. Apparently, Google had LLMs ready before OpenAI, but it was too scared to release them.
Companies like Google, Meta, Microsoft, and others were working on advanced AI for years before ChatGPT hit the market. For example, two months before ChatGPT launched, Meta launched BlenderBot in August 2022. This was an interesting, and very under-powered chatbot compared to ChatGPT. So, this AI age was in production long before the public even heard of an LLM.
Google was too scared to launch its LLM before OpenAI
One of the driving forces behind Google is, and has been its co-founder Sergey Brin. Earlier this week, he spoke at the All-AI Summit in L.A., and he spoke about AI, Google, and the marriage of both. During an interview, he announced that he’d returned to Google after leaving in December 2019. He’s been working on AI technology pretty much every day at the company.
However, he also revealed that Google had the technology in place to release its own LLMs to the public before OpenAI. However, he remarked that the company was too timid to release them. We’re not sure what form Google would have released them as. It could have done like Bard/Gemini and released a standalone site.
According to Brin, Google had pretty much invented the LLM with its “Transformer” paper back in 2017. Imagine having knowledge of LLMs back in the year that Android 7 launched, LG made its LG V20, and flagship phones still had 4GB of RAM.
Just a few years later, Google could have probably put some sort of generative AI tool onto the market and still be years ahead of OpenAI. However, As Brin put it, the company was “too timid to deploy” the technology. It was for fear of making embarrassing mistakes.
Perfectionist
Yes, we know the irony… as many of Google’s AI releases seem to be plagued by mistakes. But, back in the day, the company was being a perfectionist about its AI. Brin wanted Google to release the models to the public so that people could get their hands on them for testing. Rather, the company held the tech “close to the chest and hidden until it’s like perfect.”
This is something that heavily contrasts the Google of today. We’re hearing stories of Google’s rushed and “fix it in post” mentality to AI releases. When ChatGPT hit the market, the company sounded the alarm bells and started to work on its own AI tools. The thing is that the company rushed tools to the market, and we’ve seen some major missteps. We can’t forget about the shaky Bard unveil, the Glue on Pizza debacle, or the inaccurate historical skin tones.
It’s tough to imagine after all of that, but Google had a cautious and controlled approach to AI.