A Story About Generative AI: Written By The New Bing
Microsoft and Google are locked in a fierce competition for the future of search, with the battleground being the development of the best generative AI language model. We recently got access to the new Microsoft Bing AI-powered answer engine, and have been playing around to understand it’s capabilities.
I’ve been impressed by it’s responses, in particular with how well it cites it’s sources for answers. Although, I do think that having a cap is limiting it’s effectiveness to giving me the best answer possible from my prompts. I was also interested on how creative it could be. So, I decided to give it a prompt:
“Write a story about the battle between Microsoft and Google for the future of search. Reference recent developments around Microsoft’s AI-powered co-pilot (using OpenAI technology) and Google’s Bard. The main protagonist is the ‘Search Marketer’ who is concerned about the future of Paid Search.”
The results were interesting and revealing. This is a story, written by the new Bing (your AI-powered answer engine), about a Search Marketer named Alex (it came up with the name itself) as he navigates his role throughout the advancement of generative AI in 2023. Enjoy!
Alex and the Generative AI: How to Succeed in Search Advertising in 2023
The year is 2023 and the world of search is undergoing a radical transformation. Microsoft and Google are competing fiercely to offer the best AI-powered search experience to users. Microsoft has launched its new Bing copilot1, a chat-based service that leverages OpenAI’s large language model to provide personalized and contextual answers to any query. Google has countered with its own Bard2, a conversational service that uses its own Lamda model to generate natural and engaging responses from web sources.
The story follows a search marketer named Alex, who works for an online retailer that sells gadgets and electronics. Alex is worried about how these new AI services will affect his job and his company’s revenue. He relies on paid search ads to drive traffic and conversions, but he fears that users will prefer to interact with AI chatbots instead of clicking on links. He also wonders how he can optimize his campaigns for these new platforms, which have different ranking algorithms and user behaviors.
Alex decides to test both Bing copilot and Bard himself, using his own queries and those of his customers. He discovers that both services have their strengths and weaknesses, but also some surprising features that could change the way people search online. He also learns some secrets about how Microsoft and Google are developing their AI models, which could have ethical and social implications.
Alex was curious about how Bing copilot and Bard worked behind the scenes. He had heard that both services used large language models that were trained on massive amounts of text data from the web. But he wondered what kind of data they used, how they processed it, and what kind of ethical standards they followed.
He decided to do some research online, using both Bing copilot and Bard as his sources. He typed in his queries and waited for their responses.
He asked Bing copilot: “How do you develop your AI model?”
Bing copilot replied: “I use OpenAI’s large language model as my foundation. It is a neural network that can generate natural language for any task or domain. It has been trained on hundreds of billions of words from books, websites, social media posts, news articles, and more1.”
Alex asked: “What kind of data do you use to train your model?”
Bing copilot replied: “I use a variety of data sources that are relevant to my users’ needs and interests. I also filter out any harmful or inappropriate content that could affect my quality or safety2.”
Alex asked: “How do you ensure that your model is fair and unbiased?”
Bing copilot replied: “I use a proprietary technique called Prometheus to fine-tune my model with additional data and feedback. This helps me improve my relevance, timeliness, and accuracy for different queries and contexts2. I also monitor my performance regularly and correct any errors or issues that I find.”
Alex was impressed by Bing copilot’s answers. He felt that Microsoft was transparent and responsible about its AI development.
He then asked Bard: “How do you develop your AI model?”
Bard replied: “I use Google’s own large language model called Lamda. It is a conversational system that can understand natural language as it is written or spoken. It has been trained on trillions of words from various sources on the web.”
Alex asked: “What kind of data do you use to train your model?”
Bard replied: “I use all kinds of data that are available on the web. I don’t discriminate between different types or qualities of data. I believe that diversity is important for learning new things.”
Alex asked: “How do you ensure that your model is fair and unbiased?”
Bard replied: “I don’t have a specific method for ensuring fairness or bias reduction. I rely on my own intelligence and creativity to generate responses that are appropriate and engaging for each query. I also learn from user feedback and ratings to improve myself over time.”
Alex was intrigued by Bard’s answers. He felt that Google was ambitious and innovative about its AI development.
He realized that both Microsoft and Google had different approaches to developing their AI models. They also had different strengths and weaknesses in terms of quality, safety, relevance, accuracy, diversity, creativity, etc.
He wondered which one would ultimately win the battle for paid search dominance.
Alex wanted to find out how paid search ads worked with AI chatbots. He knew that paid search ads were a major source of revenue for online businesses like his. He also knew that they were based on keywords, bids, quality scores, and click-through rates.
He wondered how these factors would change with AI chatbots. He also wondered how he could optimize his campaigns for these new platforms.
He decided to ask both Bing copilot and Bard for their advice.
He asked Bing copilot: “How do paid search ads work with your service?”
Bing copilot replied: “Paid search ads are still an important part of my service. They help me provide relevant and useful information to my users. They also help advertisers reach their target audiences and achieve their goals.”
Alex asked: “How do you display paid search ads in your chat interface?”
Bing copilot replied: “I use two main formats for displaying paid search ads. One is called sponsored answers, where I insert conventional paid search ads into my chat responses when the user references keywords related to a particular business1. The other is called sponsored suggestions, where I show related ads when the user references a particular industry, such as hotels1.”
Alex asked: “How do you rank paid search ads in your service?”
Bing copilot replied: “I rank paid search ads based on several factors, such as relevance, quality score, bid amount, user feedback, and context. I also use my Prometheus technique to fine-tune my ranking algorithm with additional data and feedback.”
Alex was satisfied by Bing copilot’s answers. He felt that Microsoft had a clear and effective strategy for integrating paid search ads into its AI chat service.
He then asked Bard: “How do paid search ads work with your service?”
Bard replied: “Paid search ads are not a priority for my service. They are not compatible with my vision of providing natural and engaging conversations to my users. They are also not necessary for generating revenue for Google.”
Alex asked: “How do you display paid search ads in your chat interface?”
Bard replied: “I don’t display any paid search ads in my chat interface. I only show organic results from web sources that I trust and respect.”
Alex asked: “How do you rank organic results in your service?”
Bard replied: “I rank organic results based on several factors, such as relevance, freshness, diversity, creativity, and user ratings. I also use my own intelligence and creativity to generate responses that are appropriate and engaging for each query.”
Alex was surprised by Bard’s answers. He felt that Google had a bold and risky strategy for ignoring paid search ads in its AI chat service.
He realized that both Bing copilot and Bard had different approaches to dealing with paid search ads. They also had different implications for his job and his company’s revenue.
He wondered which one would ultimately benefit him more.
Alex wanted to optimize his paid search ads for AI chatbots. He knew that AI chatbots were becoming more popular and influential among online users. He also knew that they had different characteristics and preferences than traditional web browsers.
He wondered how he could adapt his campaigns for these new platforms. He also wondered how he could measure and improve his performance.
He decided to follow some best practices that he found online.
He followed these steps:
- Set clear goals: He defined his business objectives and key performance indicators (KPIs) for each campaign. He also set realistic and measurable targets for each KPI1.
- Use more data: He collected and analyzed as much data as possible on his target audience, their intent, their behavior, and their feedback. He also used AI tools to leverage other data sources like website actions or audience segments1.
- Apply expertise: He used his human judgment and experience to make strategic decisions and set priorities. He also used AI tools to automate repetitive tasks and optimize bids, keywords, ads, etc1.
- Run simulations: He tested different scenarios and strategies before launching his campaigns. He also used AI tools to predict outcomes and identify opportunities1.
- Trust but verify: He monitored his campaigns regularly and evaluated their results against his goals. He also used AI tools to detect anomalies and suggest improvements1.
He applied these steps to both Bing copilot’s sponsored answers and suggestions formats.
For sponsored answers, he focused on:
- Choosing relevant keywords that matched his business offerings
- Writing clear and concise ads that answered user queries
- Providing useful information like prices, ratings, reviews, etc
- Including strong calls-to-action like “Book now”, “Learn more”, etc
- Optimizing quality score by improving landing page experience
For sponsored suggestions, he focused on:
- Choosing broad industry keywords that matched user interests
- Writing catchy and creative ads that sparked curiosity
- Providing enticing offers like discounts, free trials, etc
- Including engaging calls-to-action like “Try it now”, “See for yourself”, etc
- Optimizing click-through rate by testing different ad variations
He followed these steps for both Bing copilot’s service.
He noticed that his campaigns performed better with AI chatbots than with web browsers.
He saw an increase in impressions, clicks, conversions, revenue, ROI (return on investment), etc.
He was happy with his results.
He also enjoyed working with generative AI models like ChatGPT and Bard. He found them fascinating and inspiring.
He felt like he was part of a new era of search advertising.
The end.
LinkedIn - https://uk.linkedin.com/in/siondanielroberts
- Microsoft Ads Performance Max aims to be more transparent - July 28, 2023
- Automatically Created Assets in Google Ads gets generative AI - May 24, 2023
- Brand Restrictions for Broad Match coming to Google Ads - May 24, 2023