Google vs Microsoft: Why Satya Nadella has advantage over Sundar Pichai in the AI war

Dominating the search business for over two decades, Microsoft boss Satya Nadella’s decision to integrate Bing with ChatGPT has possibly got Google CEO Sundar Pichai worried. After all, in contrast to Microsoft's successful launch of ChatGPT, Google's LaMDA (Language Model for Dialogue Applications) failed to generate the buzz or excitement Google was hoping for. And the recent misstep with its BARD AI chatbot has only made it worst as concerns are growing about Google's ability to execute effectively in this area.

To start with, Microsoft has taken a bold approach in deploying its new AI technology, while Google has been more cautious. Microsoft has demonstrated impressive execution in recent years. Its Azure platform has been a standout performer, and its early investment in OpenAI has paid off handsomely. With savvy acquisitions and its capture of the AI market, Microsoft has proven to be an industry leader. Meanwhile, Google's track record in execution has been less impressive, with a series of failures and project cancellations.

Also Read

Google layoffs: Not only humans but even robots are being fired

Google Pixel phones crashing after playing this 'spooky' YouTube video

'Why me': Google's star performer laid off in recent wave of job cuts
 

Adnan Masood, PhD., Chief AI Architect, UST told Business Today: “Although Google has developed a competing chatbot BARD AI, it has not yet been released to the public due to concerns around reputational risk and safety issues with current AI technology. Google is taking steps to address these concerns by enlisting its employees to check the accuracy of BARD's answers. The theory is that if BARD incorporates LaMDA and draws on Google's vast index and social media data, it could potentially provide answers to any query about fresh news.”

Google's leadership in the search industry and history of innovation in machine learning puts it in a strong position to succeed in the new world of language models such as ChatGPT, LaMDA which analyse bodies of text data to provide a basis for their word predictions. However, the emergence of these new language models, which have the potential to replace traditional search methods, provides an opportunity for other companies to compete with Google in a way that was not possible in the past.

“These language models have the ability to reset long-held assumptions about search and, if widely adopted, could render the concept of keywords and traditional advertising models like Google AdSense irrelevant. With the recent misstep of Google's BARD chatbot, there is a perception that Google is now playing catch up to competitors like Bing, and there is potential for other companies to gain a foothold in the search market. Google has a lot at stake!” adds Masood.

 But as the competition between Google and Microsoft intensifies, it's too early to predict the ultimate clear winner as each model presents unique capabilities and challenges that require further improvement to enhance their overall performance.

 "It is important to note that GPT-3 and LaMDA are distinct in their capabilities. While GPT-3 excels in generating human-like sentences, LaMDA is a conversational model that can handle arbitrary user input in a way that is sensible, specific, and interesting. As we continue to research in natural language processing, it's essential to identify the unique challenges that each of these areas poses,” explains Masood.

 Microsoft has developed a proprietary AI model called Prometheus, which combines Bing's index, ranking, and answer search results with OpenAI's GPT models. 

“This allows the ChatGPT models to have a fresher and almost real-time content and data to use for its training models. Prometheus is used in several areas of Bing, such as query interpretation, leveraging Bing's search index, ranking algorithm, and displaying answers and results. Additionally, Bing Chat provides links and citations to where it found the content, which ChatGPT currently does not,” explains Masood. 

Microsoft's blog post also highlighted the importance of merging the Bing Search product with the Bing Chat product to ensure an intuitive user experience.

Google’s BARD build on LaMDA is a model that underwent pre-training on a massive dataset of 1.56 trillion words, a considerable increase from the previous models' training data. “It was then fine-tuned, a two-stage process that fine-tuned its generator and classifier using a dialog dataset to improve the groundedness of its responses. During its interaction with the user, it can also access external information retrieval systems. Although LaMDA's initial outcomes were promising, it is essential to acknowledge that the model still has limitations that must be addressed to improve its overall performance,” he adds.

As a result, Microsoft's Prometheus and Bing Chat currently offer real-time content and search results, while Google's LaMDA exhibits an exceptional ability to handle arbitrary user input. To unlock the full potential of these models, continued research and development in natural language processing will be crucial. 

Be it Microsoft or Google, whoever wins, it’s an exciting technological era where the integration of chat and search in next-generation AI is blurring the lines.