VENDORiQ: Google’s AI Hype Hides a Bigger Problem

Why It’s Important 

Gemini is touted as ‘the most powerful AI ever built’ due to its ability to handle large amounts of data and complex prompts. 

However, there have been concerns about its readiness for public use because of the discrepancies between Google’s demo and its actual capabilities. This mismatch in Gemini’s demo is due to Google’s carefully curated and edited representations of the model’s capabilities in the demo video, which did not accurately reflect the real-time, spontaneous interactions with the model. This ‘curated’ demo approach has been confirmed by Google, indicating that the interactions shown in the video were not spontaneous real-time interactions as implied. 

This has (fairly) raised concerns about the accuracy and transparency of the promotional material for this new AI.

Google’s stumble in demonstrating Gemini is a symptom of a much bigger problem.  All vendors are hyping AI with bold vision statements and predictions. Google is literally forced to enter the AI hype race to compete against the other AI vendors. Worse, industry pundits, who often have little genuine understanding of the algorithms and methods underpinning different generative AI capabilities, increasingly attempt to outdo each other with technological predictions. For example, IBRS recently encountered an AI pundit who insisted AI would replace all ERPs because “all enterprise data can just be kept in spreadsheets, and the AI will take care of everything else”. No. Really.

Rather than focusing on ridiculous predictions, focus on what matters: how using multiple AI algorithms will impact how work is done. 

Who’s Impacted 

• CEO 
• Business strategists
• CTO 
• AI developers
• Workforce/change management teams

What’s Next? 

When reviewing AI claims, always ask for transparency. Demand answers about how demonstrations were created and how the predicted impact of AI was calculated.

Most importantly, IBRS recommends recalibrating expectations for AI.

First, understand that the future will not be a single ‘all-encompassing’ AI solution. Rather, the future will be a collection of many algorithms, each with specific purposes. For example, large language models (LLMs) like GPT-4 will not be ‘intelligent agents’ that ‘provide answers’. The underlying algorithm was never intended for that use. They need to be chained with other algorithms, such as machine learning, keyword extraction, and even traditional SQL and NoSQL database lookups that provide the ‘answers’ on which they can then generate responses.

Second, realise that AI is not a panacea for efficiency or broken processes. Even with complex chains of algorithms, workflow redesign needs to be intentional. Principles of process mapping, modelling and change management are still required to get the most from AI investments. This is the reason why IBRS has historically placed AI firmly in its workplace and automation research agenda. At the end of the day, AI is automation.

Third, consider that AI will significantly change where ‘value’ is created within businesses. In the past, the notion was that ‘content was king’.  With generative AI, fed highly custom data, that may no longer hold. Likewise, AI-power self-service agents may displace the value of previous high-touch customer activities. Many jobs will change as a result. Organisations must look at how multiple algorithms can be combined to intentionally disrupt their current value proposition, and plan a transition. This transition will involve incrementally adding AI capabilities into existing systems and workflows, with the mid to long-term goal of creating new opportunities and new forms of value.

Related IBRS Advisory  

1. VENDORiQ: Proposals For AI Regulation
2. VENDORiQ: OpenAI Launches New Features