Google’s Chatbot Bard VS ChatGPT: The Battle for AI Supremacy 

In an official statement published on February 6th, Google announced that its long-awaited chatbot Bard would finally see the light of day in the upcoming weeks. The competition is tight, with ChatGPT and Bing’s AI-powered search engine, so all eyes are on Google’s candidate.
Unfortunately, it has already experienced a debacle.
Will it be able to rise above it and exceed ChatGPT’s success?
Read on to find out.

Reading Time: 4 minutes

google chatbot bard

Illustration: Milica Mijajlovic

Google’s background in AI 

Google’s approach to AI could be described as in-depth but cautious. They’ve been working on AI solutions for years now but decided not to release most of them due to various concerns. 

For example, their innovative AI music generator MusicLM has no competition whatsoever, but Google’s team stated they have no plans of releasing it at this point. It’s mostly due to the risk of misappropriation of creative content and concerns over inherent biases present in training. 

Anyhow, you’ll often hear AI experts stating that Google in fact has much more capacity to develop an advanced chatbot in comparison to its rival OpenAI. As a matter of fact, the letter “T” in GPT technology refers to Google’s famous Transformer. And this seems to be the most cited argument for their superiority compared to competitors. 

  1. BERT was the first Transformer model, that offered an innovative way of understanding the complexity of human communication; 
  2. MUM was introduced two years ago, and is 1,000 times more powerful than BERT. It has a multi-lingual understanding of information which can pick out key moments in videos and provide critical information, including crisis support; 
  3. Their newest AI technologies (LaMDA, PaLM, Imagen and MusicLM) are aimed at creating new approaches to addressing information, from language and images to video and audio. 

When you think about it, maybe we shouldn’t think of Bard as a direct response to ChatGPT but rather as an independent solution that came naturally as artificial intelligence progressed over the years. 

Either way, now that the ChatGPT is so widely represented, Google’s solution is expected to offer even better quality. 

To do that, Bard needs to address all of ChatGPT’s imperfections. 

But how will it manage to succeed at that? 

Let’s find out. 

What can we expect from Bard? 

On February 6th, Google published an official message from their CEO Sundar Pichai, titled “An important next step on our AI journey”, referring to the upcoming release of Bard. 

These are the most important takeaways from his letter: 

  • They’ll use AI technology to continue their mission of organizing the world’s information. 
  • He underlined that Transformer (2017) is the basis of generative AI applications we’re seeing today. 
  • Bard will be powered by LaMDA (Language Model for Dialogue Applications), which Google unveiled two years ago, known for its conversational capability. 
  • Currently, Bard is entrusted to what they call “Trusted Testers”. 
  • It will draw information from the web (if we might add, this will be the most important distinction from ChatGPT). 
  • Initially, it’ll be released with the light version of LaMDA, which doesn’t require as much computing power. 
  • After being released to the public, Google will combine external feedback from users and internal feedback from testers to mold the final version. 
  • Their first stop for implementing AI advancements will be Google Search. 
  • For queries to which there’s no one right answer, Search will use the help of AI to “distill complex information and multiple perspectives into easy-to-digest formats”. 

As you may have noticed, these points mostly refer to their mission and ambition, but do not answer the most common questions that users have regarding the AI solution. Most importantly, it doesn’t say what users can expect from Bard that the competition doesn’t already offer. 

On the other hand, as we pointed out, the main difference (that we know of so far) compared to ChatGPT lies in the fact that Bard will draw information directly from the web, meaning that its knowledge won’t end in 2021. 

Comment
byu/ChocolateTsar from discussion
intechnology

Now, this may come as a relief to those who have been frustrated with ChatGPT’s limited knowledge but keep in mind that OpenAI went with this decision for a reason. 

Namely, if the chatbot had unlimited knowledge, and could be self-trained to explore the web independently, there’s a huge risk of providing offensive, biased, discriminatory answers or sharing sensitive information. That’s the main reason why some of the previous versions were withdrawn or unreleased. 

In other words, releasing the almighty, omniscient chatbot to the world could have serious consequences. 

On the other side, knowing Google’s approach explained at the beginning of this article, they wouldn’t make such a drastic move at this point. That said, maybe we should trust that they’ve come up with a unique, safe solution that will upgrade what we’ve seen so far not only from competitors but from Google itself. 

Bard’s pre-launch debacle   

At this point, we can only speculate about Bard’s future functionalities, which may or may not be true. 

However, based on the promotional video, we can see that Google’s long-awaited solution has already experienced a debacle. 

Namely, in a video aimed to demonstrate Bard’s capability, the public has already pointed out major flaws. 

When asked about the James Webb Space Telescope (JWST), Bard delivered false information that didn’t go unnoticed. 

More precisely, it stated that the telescope “took the very first pictures of a planet outside of our own solar system” when, in fact, the first image of an exoplanet dating from 2004 was taken by Chauvin et al. with the VLT/NACO using adaptive optics. 

According to global media, this mistake has cost Google $100 bn, with Alphabet’s shares being plummeted. 

Not only that, but the online community is convinced that Google is losing ground in its key area to Microsoft, a key backer of OpenAI. 

But how did this mistake even happen? 

Well, these AI models use sources based on open-source datasets that are incorporated with biases and flaws, which are then inherited by chatbots. Obviously, this problem hasn’t been solved yet by either ChatGPT or Bard’s current version. 

image-1

The networks don’t have any concept of what is ‘true’ or ‘false’. They simply produce the likeliest text they can in response to the questions or prompts they are given. As a consequence, large language models often get things wrong.

Source: The Guardian 

Being very well aware of this, Google’s officials stated that this mistake has yet again highlighted the importance of a rigorous testing process and that Bard won’t be released to the public until they’re satisfied with the feedback of their Trusted Testers. 

A journalist by day and a podcaster by night. She's not writing to impress but to be understood.