A Closer Look at LaMDA, Google’s Super Realistic New Chat Bot
In the UK, Google just launched a limited trial of the notorious Skynet-esque AI that a former Google worker claimed was sentient. Their goal is to gain feedback from average people on their interactions with LaMDA (Language Model for Dialogue Applications).
As convincing as it is, the chat bot is not in fact sentient, of course. It does what it’s programmed to do, but so well that it can be shocking.
What is Google LaMDA and How Does it Work?
Google LaMDA is an AI technology that is designed to mimic human conversation to power chatbots. It works by gathering and utilizing information gathered from human interactions.
Google LaMDA mimics a human dialogue partner by generating responses to what a user has said. It uses an algorithm to generate relevant answers on the spot.
This algorithm is based on previous conversations that have been fed into LaMDA. When a user interacts with LaMDA, the system uses an algorithm to generate a response based on the conversation that has taken place so far.
When the user responds, it gets fed back into the algorithm and used to generate a new response. The system cycles through this process until the user exits the conversation.
What is Google Teaching it?
Google wants to know how well LaMDA performs different tasks such as answering factual questions or making suggestions. But it also wants to see what qualities users might assign to it.
To do this, Google has created a series of scenarios, based on real-life situations, for users to assign to Lamda. The AI’s online test kitchen has three assignments: On-demand recipes, restaurant reviews, and holiday gift ideas.
Users can choose which one they want to complete, and then they can select a conversation partner, choosing between a male or female voice.
How to Sign Up for Trial Run of Google LaMDA
Users can install and register to use LaMDA’s AI Test Kitchen App with their Google accounts and join the waiting list to try it. The majority of the AI’s functionality won’t be included, which makes us wonder just how powerful this AI already is in its full version.
After LaMDA launched in the States in August, thousands signed up each day. Especially after the fiasco involving a former Google employee, public interest in LaMDA has been high.
The Future of Google LaMDA
It's likely that Google will launch a more in-depth test kitchen app in the future. This would mean it’s found that its AI is capable of more advanced tasks and more varied responses with everyday people.
It’s likely that LaMDA will make suggestions based on what it knows about a user and their account. Google has stated that LaMDA is not capable of generating original sentences but can instead create variations on sentences and phrases it has encountered before based on existing info.
This would mean that LaMDA is still unable to produce completely original sentences to communicate with users and is therefore unable to have a truly meaningful conversation. But if it can use your data while talking with you, that would certainly give it an edge.
While LaMDA is impressive, and it may even be the most advanced AI to date, it is not close to being sentient. In order to be sentient, a system would need to have self-awareness and be able to think autonomously.
Surely we’ll see more advancements in AI in the coming years as companies like Google and Apple continue to invest heavily in the field. However, it is important to adopt a critical mindset when discussing the capabilities and future of AI since the algorithms used are still just algorithms!