Google Bard’s Journey: From Initial Hurdles to Rising Contender
--
The annual I/O 2023 conference by Google (held in Mountain View, CA on May 10), a hub for developer innovations, was brimming with substantial breakthroughs, particularly in the realm of artificial intelligence. The tech behemoth showcased a range of AI improvements across its applications and services, with a special focus on its advanced language model, PaLM 2, and the updated version of Bard, Google’s experimental AI chatbot. To truly appreciate Bard’s evolved capacities, it’s imperative to trace its origins and comprehend its foundational framework.
The Rise of Google’s Bard
Google’s Bard, introduced in February, represented the company’s pioneering stride into the domain of AI-fueled conversational bots, a field already marked by OpenAI’s popular ChatGPT. Initially, Bard was empowered with a downscaled variant of Google’s Language Model for Dialogue Applications, LaMDA. It was crafted to interact with users in a conversational manner, exhibiting abilities such as idea generation, essay writing, code writing, and even solving mathematical problems.
Despite its innovative approach, Bard’s first iteration faced critique for its restricted capabilities and factual inconsistencies. Sundar Pichai, Google’s CEO, acknowledged these initial limitations, stating that they were purposeful steps towards a gradual enhancement of Bard’s functionality through more powerful LLMs.
At the I/O 2023 conference, Google made good on its promise by imbuing Bard with the newest iteration of the Parallel Language Model, PaLM 2, marking a significant advancement from LaMDA and boosting Bard’s abilities.
Google I/0 2023 Conference Source: Google
Initially, Bard’s access was limited to a chosen set of testers from the US and the UK. While it was expanded to a waitlist in March 2023, Bard was still out of reach for the general public. Google has now increased Bard’s availability to over 180 countries.
Bard’s Advanced Features
The incorporation of PaLM 2 has significantly elevated Bard’s capabilities. It now excels in mathematics, logic, and reasoning, and can generate, elucidate, and debug code in more than 20 programming languages, proving a useful tool for developers.
Bard’s latest update introduces a more visually appealing and interactive user interface. Users can now input images for Bard to respond to, using tools such as Google Lens. It can also produce amusing captions, increasing user engagement.
Bard’s responses can be exported directly to Gmail and Google Docs. It can also search the web for images, use knowledge graphs for relevant information, and employ Google Maps for location-based queries. Its integration with Google Sheets further boosts its practicality.
To enhance its collaborative abilities, Google aims to incorporate Bard with external services like Adobe Firefly. This will enable users to generate new images from text prompts for editing. Google is also fostering partnerships between Bard and other companies like Kayak, OpenTable, ZipRecruiter, Instacart, Wolfram, and Khan Academy.
Google Bard vs. ChatGPT
Google’s venture into the AI chatbot space with the introduction of Bard initially faced significant hurdles, as it stumbled in matching up against competitors such as Bing Chat and OpenAI’s ChatGPT. The launch was marred by an unfortunate incident where a Google advertisement featuring Bard provided an incorrect response, leading to a brief but significant 8% drop in Google’s stock.
However, the tide seems to be turning for Bard.
One of Bard’s most significant advantages over ChatGPT is its internet access. Unlike ChatGPT, which has a training data cut-off in 2021, Bard’s training dataset includes real-time information from the internet, enabling it to deliver comprehensive and up-to-date responses.
While Bard’s internet access is a significant advantage over ChatGPT, it also presents potential risks which must be carefully managed.
Data privacy is a paramount concern. Bard’s real-time internet access necessitates the anonymization of user interactions to avoid inadvertent access or storage of sensitive personal information. As such, Google must enforce stringent data privacy standards and adhere to relevant regulations to safeguard user data.
The accuracy of real-time information is another substantial challenge. Bard’s training on internet data requires it to distinguish between reliable and untrustworthy sources in an online environment laden with misinformation and unverified content. In the absence of robust fact-checking mechanisms, Bard could risk disseminating misleading or incorrect information.
Chatbots are also known to “hallucinate,” meaning it’s generating responses without accurate or reliable data. The chatbot provides answers or claims that are not based on reality or supported by factual knowledge. This metaphorical term highlights instances where the chatbot generates imaginative, fictional, or unrelated responses. Hallucination can occur due to limitations in training data, biases, or algorithm errors. It’s unintentional and not indicative of conscious thought or imagination. Developers strive to improve chatbot models to minimize these instances and ensure responses are grounded in reality and based on accurate information.
Furthermore, training Bard on internet data could result in the inadvertent absorption and reproduction of biases present in the data. This could lead to responses that propagate harmful stereotypes or misinformation. Google must therefore continuously implement and refine strategies for detecting and mitigating bias in Bard’s training and operational stages.
Google plans to integrate Bard into various apps and services to make it readily accessible to smartphone users. OpenAI has also recently released a mobile version of ChatGPT, bringing the chatbot to iOS users. At the same time, Microsoft has integrated Bard into Edge, GitHub, and Office through a limited initial program by invitation only.
Google is boasting Bard’s ability to provide image-based responses and accept image prompts. This feature, combined with the chatbot’s voice command functionality, offers a multitasking-friendly user experience. However, it’s worth noting that there are still some bumps in the road. For example, in my experience, requesting Bard to produce an image returns the following error message:
”I am a large language model, also known as a conversational AI or chatbot trained to be informative and comprehensive.”
Another feature of Bard is its array of plugins. At Google’s event, the company alluded to several plugins coming down the pipeline, including ones for Walmart, Spotify, Uber Eats, Adobe Firefly, and various Google Apps. OpenAI has already introduced a wide range of plugins for ChatGPT, which are exclusively available to ChatGPT Plus subscribers.
Lastly, in the realm of coding support, Bard has the capability to assist in code generation, explanation, and debugging across more than 20 programming languages, such as C++, Python, Java, TypeScript, and JavaScript. This goes beyond ChatGPT’s current capabilities, although OpenAI’s Codex may provide a more equivalent comparison.
These improvements position Google Bard as a rising contender in the AI chatbot space, promising an exciting showdown between it and existing leaders like ChatGPT.
The Current State of Chatbots
Google’s I/O 2023 conference unveiled significant advancements in artificial intelligence, particularly with the relaunch of the much-improved Bard, now powered by the advanced PaLM 2 model. The journey of Bard, from its humble beginnings to its current upgraded form, demonstrates Google’s commitment to enhancing AI-powered conversational capabilities.
Despite initial hurdles, Bard’s enhanced features and broadened availability promise to position it as a formidable competitor in the AI chatbot space. Its significant upgrades, such as real-time internet data access, integration into various apps and services, image and voice response capabilities, diverse plugin support, and extensive coding assistance, help set it apart from other AI chatbots like ChatGPT.
However, it’s also important to acknowledge the evolving nature of this technology. OpenAI’s ChatGPT, for instance, has demonstrated consistent progress and will likely continue to evolve in response to emerging competitors like Bard. The AI chatbot space is poised to witness dynamic growth and innovation, and the competition between AI leaders like Google and OpenAI will ultimately benefit users by accelerating the development of more powerful and versatile AI tools. As Google’s Bard and OpenAI’s ChatGPT continue their race, we can eagerly anticipate more advancements in this fascinating field.
The choices are not limited to Microsoft’s and Google’s offering though. New entrants are working hard to compete. Some of the other top chatbots on the market include: Bing, Jasper, YOU, Writesonic, and Socratic.
