Google LaMDA large language model

Released: May 2021

LaMDA (Language Model for Dialogue Applications) is a large language model developed by Google to help develop chatbots. The first version of LaMDA was released in May 2021, and the second a year later in May 2022. 

LaMDA 'sentience'

LaMDA hit the news in June 2022 after the Washington Post had reported Google engineer Blake Lemoine had tried to convince fellow Google employees that LaMDA was sentient.

'When LaMDA claimed to have a soul and then was able to eloquently explain what it meant by that,' Lemoine, an ordained Christian mystical priest, said he 'was inclined to give it the benefit of the doubt.'

Google, technology professionals, philsophers and ethicists responded to the notion that LaMDA - and other technologies - can be human primarily on technical, scientific grounds, prompting Lemoine to complain the model faces 'bigotry' in an interview with WIRED. 

Google suspended and fired Lemoine after he breached company policy by sharing information about his project, recruited a lawyer for the AI after claiming that LaMDA had asked him to do so, and alleged that Google was discriminating against him because of his religion.

Google Bard inaccuracies

In February 2023, Google launched Bard, a chatbot based on LaMDA. The launch was seen to flop as it shared inaccurate information in a promotional video, reinforcing views that Google is lagging Microsoft and Open AI in this area. 

The two software companies had launched Bing's integation of ChatGPT earlier the same day.

Operator: Alphabet/Google
Developer: Alphabet/Google
Country: USA
Sector: Technology
Purpose: Optimise language models for dialogue
Technology: Large language model; Neural network; NLP/text analysis
Issue: Accuracy/reliability; Anthropomorphism
Transparency: Governance; Black box

Page info
Type: System
Published: January 2023
Last updated: February 2023