New Delhi: Google CEO Sundar Pichai has pushed back against the idea that the company missed the early wave of AI chatbots, arguing instead that Google chose not to release its technology first. In an episode of a podcast with John Collison, Pichai acknowledged that the company had already developed systems like modern chatbots long before they would see the light of day, but that they were withheld due to quality and safety issues.
In reference to the 2022 scandal of former Google engineer Blake Lemoine, Pichai claimed that the internal prototype of a Google chatbot known as LaMDA was effectively a primitive version of what users today would call such tools as ChatGPT. However, he emphasised that the choice not to launch it was not spontaneous but an indication of technological backwardness.
Google had the tech but chose caution
Pichai said that Google had already created a conversational AI that worked internally but was not at the stage of being used publicly. The system, he remarked, still gave answers which might be dangerous or inconsistent. He mentioned that the model did not have enough reinforcement learning with human feedback (RLHF), which is one of the processes that are utilised to make AI outputs safer and more precise.
Pichai stressed that Google has much higher internal quality standards on products, in particular on search-related products, and could not possibly have released it at that time. His remarks imply that the company was putting more emphasis on speed in the sense that it was more about trust and safety as its competitors were accelerating.
Lemoine episode and LaMDA debate
The comments also touched once again on the controversy of Lemoine, who publicly stated that the LaMDA chatbot of Google had become sentient. He released transcripts of his conversations with the system, which he claimed to have feelings and fears about. Google denied such assertions, and the majority of AI scientists concurred that LaMDA was not a conscious being but merely a very sophisticated language model.
Pichai did not take sides on the sentience debate. Rather, he recontextualised the episode as a demonstration of the level of development of Google’s internal AI systems at the time.
ChatGPT’s launch triggered ‘Code Red’
The product rapidly took off when OpenAI released ChatGPT in late 2022. The New York Times reported that the viral success led to a so-called code red within Google. Pichai allegedly restructured teams in the areas of AI research, product, and safety to expedite the reaction of the firm.
The creation of ChatGPT also took off at an astonishing pace due to a significant collaboration between OpenAI and Microsoft that contributed to the expansion of the product’s reach and visibility.
From Bard to Gemini: Google’s AI reset
In early 2023, Google replied with the release of Bard, which was, however, criticised due to its inaccuracies and rushed implementation. The company then rebranded Bard to Gemini, making it its flagship AI platform.
In retrospect, Pichai did not view the delayed entry of Google as a failure but rather as a move to be strategic. He referenced the case to previous instances in the history of the company, like when Google ventured into the video arena when the rivals – who would later acquire YouTube and have the market cornered – were already in the market.
Three years later, Pichai makes it clear that Google is not lagging behind in AI innovation. It just decided to wait until it felt that the technology was available.