Just three months after Google issued a “code red” to address ChatGPT's growing popularity, the tech giant unveiled its own artificial intelligence (AI) chatbot, ‘Bard.'
Bard is Google's response to Microsoft-backed OpenAI's ChatGPT, an AI chatbot that took the Internet by storm late last year due to its capability to generate human-like conversational text in seconds.
Unlike its rival ChatGPT, which only uses data from the large corpus of text it was trained on, Bard will continuously reference information from the Internet to generate answers.
Google CEO Sundar Pichai says that by doing so, Bard can provide up-to-date responses.
In contrast, ChatGPT's responses cannot provide details about recent news and research because its training set ended with data from 2021. This means it cannot offer present-day information, like data or responses based on President Joe Biden's State of the Union address.
In addition to having a completely different data source, Bard also uses Language Model for Dialogue Applications (LaMDA), a language system that Google developed two years before the AI chatbot frenzy started.
LaMDA is built using the Transformer, a neural architecture that is, surprisingly, the basis of many AI chatbots, including ChatGPT's third generation of Generative Pre-trained Transformer (GPT-3).
However, unlike GPT-3, LaMDA is trained using 137B model parameters and 1.56T words. In addition, it follows Google's Quality, Safety, and Groundedness metrics, ensuring the generated responses are high-quality and safe.
In the meantime, Google will use a lighter, smaller version of LaMDA for lower computing power and allow the AI chatbot to scale to new users. Google will also collect feedback.
As of this writing, Bard is only accessible to a group of beta testers. Google claims it will be available to the public in the coming weeks, although they already missed their initially predicted late February launch.
A $100 Billion Error
Despite having a better and more updated data source, Bard had a less promising start.
In the GIF demo shared via a tweet, the AI chatbot was asked, “What new discoveries from the James Webb Space Telescope can I tell my 9-year-old about?”
The AI responded with three answers, including a claim that “JWT took the very first pictures of a planet outside of our own solar system.”
However, European Southern Observatory's Very Large Telescope (VLT) took the first photo in 2004.
Google acknowledged the mistake, saying it “highlights the importance of a rigorous testing process, something that we're kicking off this week with our Trusted Tester program.”
The error cost the tech giant a whopping $100 billion in market value as its stock plunged following the tweet. In comparison, Microsoft's stock increased by 4.2% after announcing the integration of ChatGPT into Bing.
AI-Assisted Search
Finding information hasn't changed much in the less than 25 years since Google launched its first algorithm-based search engine. You enter a few words in the search bar, and the engine pulls websites relevant to that query. But that could change soon, as Alphabet and Microsoft plan to integrate AI with Google and Bing.
Pichai says that “you'll see AI-powered features in search that distill complex information and multiple perspectives into easy-to-digest formats, so you can quickly understand the big picture and learn more from the web: whether that's seeking out additional perspectives, like blogs from people who play both piano and guitar, or going deeper on a related topic, like steps to get started as a beginner.”
Internet users can already test Bing's AI-assisted search by picking one of the available pre-loaded questions like “Help plan my special anniversary trip.”
Once you click it, you will be redirected to the search results. However, instead of just having a list of relevant websites, you will also find a box on the right with a summarized answer to the query. It also has citations.
The box also has a “Let's Chat” button, allowing you to communicate directly with the AI without exiting the search results page. As of this writing, you can't chat with the AI yet, but you can join the waitlist.
Microsoft is also set to integrate AI with its other products, including Outlook, Word, and PowerPoint.
Google and Microsoft are not the only tech companies adopting AI. Opera is also joining the AI-assisted search revolution.
AI Arms Race
The AI chatbot frenzy started last year when ChatGPT became accessible to the public. Unlike other chatbots that usually respond in a monotonous tone, ChatGPT excels in delivering interactive, conversational answers that appear very human-like.
But its most advantageous feature is its user-friendly system, which is like a messaging chatbox. You just type your query, and the AI immediately responds. Even complex topics like programming and arithmetic can be condensed and explained in seconds.
This quick, unobtrusive experience is a welcome change, with SimilarWeb reporting over 600 million users in January alone. However, the high demand has pushed the AI chatbot to its maximum capacity, resulting in inaccessibility during peak hours. This prompted OpenAI, the makers of ChatGPT, to introduce a $20 monthly subscription for priority access.
With the introduction of Google's Bard and emerging independent AI chatbot companies, the AI arms race continues to heat up.
The battle of AI is not limited to chatbots. Even earlier, AI content-writing tools like Copy.ai and Jasper emerged, promising human-like text for blogs, advertising copies, email, and more.
Just like ChatGPT, these AI writing tools use GPT-3, an AI technology with over 175 billion machine learning parameters, to produce convincing human-like text.
Human Intervention Still Needed
Despite the lightning-fast content generation and human-like responses, chatbots and AI writing tools can favor lengthier sentences over conciseness.
They also lack the human's mind complexity and creativity, which often leaves the generated text bland and robotic. And as seen in Google's $100 billion error and ChatGPT's limitation, the relayed information is not always 100% correct.
As such, human intervention in the form of a copy editor is still crucial. AI software may know the meaning of a topic, but it cannot identify nor understand every word it elicits or evaluate the veracity of the millions of pages on the world wide web.
But there is a place for technology in writing. In research, collating facts with citations can be easily done with writing tools. The latest generation of editing tools like Grammarly can make mundane tasks like checking spelling and capitalization faster and easier. AI can give writers and editors more time to focus on what they do best – produce meaningful, fact-checked, and creative copy.
This article was produced and syndicated by Wealth of Geeks.
Bobby Kania is a full-time blogger with multiple blogs spanning topics in education, music, and advice. He currently lives in Seattle with his wife and cat. Learn more at Capitalize My Title.