Google held its annual I/O event yesterday, and, in between all the hype around Android 12 and automotive advancements, the company announced LaMDA, a breakthrough conversation technology that would be used to power Google-owned platforms and, more importantly, Google Search. LaMDA is basically a machine learning (ML) technique that is a step forward when it comes to chatbots. Till now, chatbots have mostly been limited to pre-approved responses, but with LaMDA, Google will allow for a conversational flow when interacting with a user. An anagram for “Language Model for Dialogue Applications” LaMDA will engage in a free-flowing conversation that might have an element of related or, at times, unrelated topics. It is the company’s belief that this new ability to think will unlock more natural ways of interacting with tech and new categories of applications.

What Else Does LaMDA Do?

According to Google, LaMDA is a medium of free-flowing interaction, as mentioned above, but what is surprising is the fact that it builds on earlier Google research that was published back in 2020. At that time, the research indicated that Transfomer-based language models trained using dialogue could learn to talk about any topic. Since this research, Google found that when training LaMDA, it can be tuned to finely talk about a particular topic which will significantly improve the sensibleness and specificity of the responses that are given. Do note that LaMDA is still under testing and will be tested internally by Google. A key example that helps understand its true purpose is as follows: User A: Hi, I would like to converse with you for a demonstration in a blog post. LaMDA: Hi. I am a friendly and knowledgeable demonstration meant for your blog post. I am capable of engaging in interesting conversations about virtually anything. What would you like to chat about? User A: If my washing machine and tv are malfunctioning at the same time, what could the reason for this be? LaMDA: It could be due to the power surge from the washing machine, which may have fried your tv. If it is a power surge problem, it may be time to get a new washing machine or new television. User A: Yeah, maybe. By the way, how many legs does a horse have, and could it fix my washing machine? LaMDA: Horses have a total of four legs. How could they be of help? User: You tell me LaMDA: Oh, well, they could certainly be of help to you. By moving you to a new home where you could install a new washing machine. As one can see, LamDA’s responses make sense when it comes to the initial query and when a new subject is introduced, it does not deviate from the origin and brings the conversation back on a sensible path.

Google s LaMDA Is Going to Be Revolutionary for Conversations - 51