OpenAI had kicked off the AI revolution when it had released ChatGPT in late 2022, but another company could’ve beaten them to the punch by several years.
Marc Andreessen says that Google could’ve released a GPT-4 level LLM — which was released by OpenAI in 2023 — in 2019 itself. Andreessen says that Google chose not to pursue the technology because of concerns around attaching its brand to a product that often gave wrong answers, and because of concerns around user safety. He related this to the broader point of researchers being fa removed from product and marketing teams at traditional AI companies.
”I think the AI industry in particular has a very acute version of this challenge. In sort of a normal technology company, you’ve kind of got engineers who make products, and then you’ve got salespeople or marketing people who sell them,” Andreessen said on the TBPN podcast. “(But) in the AI companies, you have this third tier of the ‘researchers’. It has worked out incredibly well — the researchers have done amazing breakthroughs at these companies,” he added.
But he said that the researchers were often too far removed from the users who used their products. “There’s not necessarily clean handoff from the researchers to the market. And so it kind of raises this question of, are these companies therefore kind of three segment companies where they have research, and then they have product development, and then they have go to market,” he added.
“I think that’s a really open issue. Google’s kind of a case study of this. Google developed the transformer in 2017, and then they basically let it sit on the shelf, right? Because it was a research project, they didn’t productize it. They were very worried about — from people I’ve talked to — brand issues and safety issues. They had all these reasons to not productize it. I talked to somebody senior who was there at the time, and I asked them when could you have had ChatGPT with GPT-4 level output if you had just gone flat out starting in 2017. They said by 2019. They already knew how to do it and then, and they’ve (only) now caught up. But it took an extra five years to catch up. And I think a lot of these companies kind of have that challenge,” Andreessen said.
Andreessen was suggesting that Google was too conservative in capitalizing on a potentially revolutionary technology in spite of having developed it, while a little-known startup like OpenAI had no qualms about releasing it and testing it out among the public. That decision eventually proved momentous — ChatGPT’s release set off the current AI revolution, and OpenAI is now worth nearly $400 billion.
Andreessen said that Elon Musk is looking to avoid the mistakes that Google made. “Elon, as usual, is provoking. Within xAI, he’s now eliminated the distinction between research and product. He’s pushing this as hard as he can, and I think it’s a good question for a lot of the other companies — kind of how hard they wanna push on actually getting these things in fully productized for out to the market,” Andreessen said.
Google wasn’t the only company that could’ve released a ChatGPT-like interface before OpenAI, but it would’ve likely been the first. Meta had created a similar technology in late 2022, but it focused only on discussing research papers, and as such didn’t see broad adoption. Anthropic too had a version of a ChaatGPT-like bot ready for months before ChatGPT’s release, but let it remain as a Slack tool and never released it to the public. While Google and Anthropic have caught up and are now neck-and-neck with OpenAI’s models, OpenAI’s first mover advantage in the AI space has given it 700 million weekly active users, which other companies now might find it hard to pry away at.