As part of the LaMDA announcement in May of 2021, Google cited Assistant, Search, and Workspace as products where it hopes to “incorporat[e] better conversational features.” A new report today looks into Google’s reluctance to launch and incorporate LaMDA into products before OpenAI and Microsoft.
The Wall Street Journal today detailed Google’s history with chatbots. Before LaMDA, there was Meena, which was first publicized in early 2020.
The Google Brain team published a research paper that described a chatbot “that can chat about anything.” This end-to-end trained neural conversational model tries to understand the context of a conversation to provide a sensible reply.
At the time, the team wanted to release their tool in a limited manner, like OpenAI did with GPT-2, but “Google leadership rejected the proposal on the grounds that the chatbot didn’t meet the company’s AI principles around safety and fairness.”
A Google spokesman said the chatbot had been through many reviews and barred from wider releases for various reasons over the years.
Work continued, and Meena became LaMDA when it gained more data and computing power. At this time, in 2020, the team “looked for ways to integrate LaMDA into Google Assistant.”‘
The team overseeing Assistant began conducting experiments using LaMDA to answer user questions, said people familiar with the efforts. However, Google executives stopped short of making the chatbot available as a public demo, the people said.
Today’s report also has this tidbit about balancing the answers this technology can provide with the source material.
Inside Google, executives have said Google must deploy generative AI in results in a way that doesn’t upset website owners, in part by including source links, according to a person familiar with the matter.
Post GPT/Bing, Google has announced the conversational Bard and generates direct answers to your question in Search. The company also said to expect LLMs in Gmail and Google Docs. It has yet to really discuss Google Assistant’s future.
In many ways, Google Assistant is a more ideal candidate for LaMDA integration than Search. People have high expectations for the latter’s accuracy, while voice-first interactions with Assistant are important but often less critical.
Interactions are shorter and people often want an immediate answer instead of a preamble, which is what Bard looks to offer. Google could use the limitations of the format to its advantage, while still getting important training data from real-world usage.
FTC: We use income earning auto affiliate links. More.