What Went Wrong with Google’s Bard and Are ChatBots Overhyped?
News
Google’s announcement of revealing a competitor to ChatGPT suffered an expensively embarrassing blunder when it was revealed that during its promotional material, it was shown that the chatbot was responding incorrectly to a question.
A response in a video demo of the application, Bard, incorrectly suggested that Nasa’s James Webb space telescope was used to take the first images of a planet outside the Earth’s solar system or exoplanets.
Investors slashed the value of Alphabet, the parent firm of Google, by more than $100 billion (£82 billion). Google stated that the blunder, which was first reported by Reuters, highlighted the need of testing new systems. Bard has been released to a team of specialized testers but has not yet been made available to the general public.
What are Bard and ChatGPT?
Bard is built on a large language AI model, which is a type of neural network that simulates the underlying architecture of the brain in computing form. It is fed massive volumes of text from the internet to educate it on how to respond to text-based stimuli. This, however, may result in the chatbot replicating errors from the knowledge it receives.
Google Bard, an AI-powered chatbot that, like ChatGPT, can react to various inquiries conversationally. Google claims that Bard leverages web information to provide new, high-quality responses. Google’s chatbot is powered by LaMDA, a language model built on Transformer, a neural network architecture.
Surprisingly, ChatGPT is built on the GPT-3 language model, which is built on Transformer as well. Google Research built and released Transformer in 2017.
Where did Bard go wrong?
According to experts, these datasets may contain inaccuracies that the chatbot repeats, as seems to be the case with the Bard demonstration.
The large language model is fed a dataset containing billions of words and constructs a model of the sentences and paragraphs that would ordinarily follow the previous portion of text based on statistical probability.
The networks have no idea what ‘true’ or ‘false’ means. They merely create the most plausible text possible in response to the queries or instructions they are provided. As a result, substantial language models frequently get things incorrect.
Google, once a forerunner in all things AI, has just been overtaken by OpenAI.
Given the growing popularity of ChatGPT, which debuted in November, Google declared the development of its chatbot a “code red” project. The fact that Satya Nadella’s company invested billions more in OpenAI last year only added to the strain on Sundar Pichai’s staff.
Alphabet shares fell as much as 9% during Wednesday’s trading session but recovered slightly to close 7.68% lower. Alphabet’s stock, which fell 40% in value last year, has risen 15% since the beginning of 2023.
Conclusion: Are AI-powered chatbots overhyped?
The advent of artificial intelligence (AI) chatbots has brought about a revolutionary change in the way businesses interact with customers. AI chatbots are computer programs that use natural language processing, machine learning, and artificial intelligence to simulate a conversation with humans. They can be used for customer service, sales, marketing, and many other business activities.
But is the hype around AI chatbots warranted? Many experts believe that AI chatbots are overhyped. They argue that technology is still in its infancy and not ready to replace human interaction. They point out that AI chatbots can’t understand the nuances of human language and therefore, can’t respond to customer questions in a satisfactory manner.
To validate the above-mentioned and the realization that we are most likely in the first generation of AI chatbots, it’s imperative to understand that they are more or less programmed to respond to specific questions. What I mean by the program is that students are taught generic vocabulary, entities, metaphors, synonyms, and so on.
To grasp the context, the chatbot employs a predefined set of flows. Additional training is required for domain-specific use cases, and one needs to be trained on relevant domain terminology as well as the relationship between the words.
That AI chatbots are still in the initial stages of development and have a long path ahead before they can guarantee a complete replacement of humans.
