Since the launch of OpenAI’s ChatGPT, many people have been questioning the future of Google and the potential threat to its search engine. After what seemed a never-ending debate of ChatGPT Vs Google, now there is a new flame of Google’s LaMDA Vs. ChatGPT.

LaMDA is built on the Transformer neural network architecture developed by Google Research in 2017, having training on human dialogue and stories, resulting in engaging and open-ended conversations.

Let us tell you that Google’s LaMDA (Language Model for Dialogue and Analytics) was released way before OpenAI’s ChatGPT with apparently “sentient” features by its own engineer.

According to the engineer behind Google’s LaMDA, Blake Lemoine, in the debate of LaMDA Vs ChatGPT, the latter is a big step in the right direction, but they’re still a few years behind Google.

So who really wins the battle? Google or OpenAI? Who is more powerful and innovative when it comes to conversational artificial intelligence? Let’s discover possible answers in the following sections.

ChatGPT vs LaMDA: What Makes Each One Special? 

To understand the difference between ChatGPT and LaMDA, we really need to know the basic nature of both models, how they work and what makes them unique from other AI chatbots.

So, Let’s Understand Google’s LaMDA Vs. ChatGPT:

What’s LaMDA? 

Source: inaa.org

Simply put, Google’s Language Model for Dialog Applications (LaMDA) is a transformer-based neural language model built on up to 137B parameters and pre-trained on 1.56T words of publicly available dialogue data along with web documents. Moreover, the model is fine-tuned on three metrics: Quality, Safety, and Groundedness.

To talk about the working method and process of LaMDA, we can say that its progress is quantified by collecting responses from the pre-trained, fine-tuned, and human-generated responses leading to multi-turn two-author dialogues. Going further, these responses are then evaluated by a different set of human raters on a series of questions against the pre-defined metrics.  

What’s ChatGPT?

Source: cnbcfm.com

On the other hand, ChatGPT by OpenAI is based on the GPT-3.5 architecture, containing 175B parameters. GPT-3.5 mainly has three models: code-davinci-002, text-davinci-002, and text-davinci-003. 

The first is the base model for code completion tasks; the middle is trained on supervised fine-tuning on human-written demonstration while having samples rated 7/7 by human labelers on overall quality scores. The latest model includes reinforcement learning with human feedback (RLHF), a reward-based model trained on comparisons by humans, having text and code limited to Q4 2021.

Similar to Google’s LaMDA, with ChatGPT, human AI trainers have access to give suggestions to craft responses and train models, supporting both users and AI assistants. Based on the response received from the trainers and users, the chatbot prioritizes the responses accordingly while generating alternative responses. 

Who’s Ahead in the Race: LaMDA Vs. ChatGPT?

Source: pressablecdn.com

To go further in the race of Google’s LaMDA Vs. ChatGPT, have a look at the responses from both platforms. Responses from ChatGPT reads more like Q&A, dry and shallow, whereas the responses from Google’s LaMDA are more conversational, friendly, and sensible.

The reason behind the different responses from both platforms is because of the training basis. LaMDA is trained on dialogues contradictory to ChatGPT, mainly on web texts. Additionally, ChatGPT has faced a lot of sarcasm and bashing as it sometimes produces incorrect information, fake quotes, and non-existing references.

On the other hand, Google’s LaMDA has an edge because it considers various metrics to generate contextual, non-generic, insightful, sensible, specific, interesting, and witty responses. 

Does that Mean ChatGPT is Over?

Do cutting-edge and extraordinary pre-defined metrics of Google’s LaMDA mean that ChatGPT is over?

Not really! What empowers ChatGPT is its reinforcement learning model (RLHF), which works on a reward-based mechanism based on human feedback resulting in better and better responses.

But if we look at LaMDA’s model, it doesn’t have RLHF, which may slow down its growth. ChatGPT’s model continuously learns from users’ behaviors and generates different, better, and improved responses than before – all thanks to ChatGPT’s neural network architecture, known as a transformer designed to process and analyze large amounts of data.

There is definitely a scope for improvement when we talk about getting accurate, contextual, and relative answers from ChatGPT despite its RLHF model.

If we believe the ongoing say and rumors, Google may soon integrate the latest version of the model into its search engine, keeping the increasing popularity of ChatGPT in mind. Undoubtedly, if Google embeds LaDAM into its search engine, it will definitely change how people use it.

To make things more interesting, let’s also note that OpenAI is partnered with Microsoft leading to similar scenarios on Bing with their access to the search engine.

It will be interesting to see who wins the battle of conversational AI chatbots, and only time can tell who will dominate the field! 

Also Read: ChatGPT-3: Get to Know about AI Chatbot Tool (Updated)

It’s Not Only About Google & OpenAI! There’s More in the Room:

Source: seekingalpha.com

Did you know that when it comes to AI chatbots, Google and OpenAI are not alone in the race? Meta, too, has its own chatbot known as Blenderbot, and its third iteration was released a few months ago. Meta’s conversational AI prototype is based on 175B parameters fused with long-term memory.

The interesting part is that Meta’s model uses dialogue history, the internet, and memory to generate output. Apparently, Google and Meta have been keeping mum about their chatbots, but OpenAI’s release of ChatGPT seems to be a game changer.

Want to get much more interesting updates about the techno world? Don’t forget to visit Techconcord.