Former DeepMind Member: Google Had ChatGPT Ready a Year Ago but Didn't Release It Due to Concerns Over Model Hallucinations and Impact on Search Business
Thibault Sottiaux, a former member of Google DeepMind and now at OpenAI, revealed that he participated in an internal project called LMChat, which was essentially the same product as ChatGPT released a year later.
Google chose to keep the project internal and not release it due to concerns about model hallucinations, the potential impact on its search business, and the possibility that DeepMind products could disrupt its own operations. OpenAI subsequently launched ChatGPT, igniting the generative AI wave, forcing Google to catch up and reorganize DeepMind with Google Brain.
This event reinforces the market perception that "first-mover advantage determines the narrative." Funding and talent are rapidly flowing to players willing to release imperfect products; the beneficiaries are those like OpenAI who seize the window of opportunity, while the pressured are large tech companies whose internal innovation is stifled by existing business protection logic.
Source: Public Information
ABAB AI Insight
Google has an original advantage in the Transformer architecture but hesitated in the implementation of conversational products due to organizational and commercial protection instincts. The existence of LMChat indicates that technological readiness is not the bottleneck; the real constraints stem from the assessment of potential impacts on search advertising revenue and the isolation between research and product departments.
In terms of capital pathways, OpenAI rapidly captured user mindshare and developer ecosystems with a "release first, iterate later" strategy, forcing Google to shift from a leader to a follower, ultimately leading to the merger of DeepMind and Brain. The difference in motivation lies in the fact that one side views product release as the starting point for learning and data flywheels, while the other sees release as an irreversible risk event.
Similar cases can be seen in Kodak's internal suppression of digital photography and Microsoft's hesitation regarding early mobile systems. The current AI industry is still in a phase where "organizational courage and technical capability are equally important."
Essentially, this is about the transfer of pricing power. The mechanism is that when the speed of technological diffusion exceeds the decision-making speed of large organizations, the first movers gain the priority to define categories and attract resources, while later entrants, even with comparable technology, must incur higher costs to catch up.
ABAB News · Law of Cognition
Seeing the future but not daring to release it is equivalent to giving the future to others.
The instinct to protect core businesses often becomes the main reason for missing new categories.
Technological readiness has never been the bottleneck; organizational courage is.