Google broke too many things in its rush to lead AI

“Move fast and break things,” Zuckerberg’s early Facebook motto perfectly summed up the need to abandon tradition and pursue innovation in a rapidly evolving industry. After a slow start, Google appears to be unofficially taking the same approach as it moves into artificial intelligence. However, recent mishaps suggest that Google’s Gemini alienated users by placing too much emphasis on the “breaking things” part.

It’s been a particularly bad few weeks for Google’s artificial intelligence team. First, the company suspended Gemini’s image generation capabilities after it generated insensitive and historically inaccurate images. Whether it was a byproduct of intentional algorithm tweaks or simply a reflection of the early state of Google’s model, the incident caused public embarrassment. It hammered Alphabet’s stock price again (at least temporarily) and even forced co-founder Sergey Brin to make a rare admission that the company “screwed up.”

Google soon followed suit, revealing that its affordable flagship Pixel 8 smartphone will not receive the Gemini Nano large language model currently used on the top-tier Pixel 8 Pro phone (not to be confused with the Gemini app that replaces Assistant). Two smartphones billed as the future of consumer artificial intelligence, both citing unclear “hardware limitations” — and both seven years old — are now on two distinct trajectories. Even worse for Pixel 8 owners, the Gemini Nano will be coming to phones from other manufacturers as well.

There’s no word yet on if or when Nano will make its way to older Pixel phones, which were also promised regular future updates at launch, although that wasn’t specified. If Google can’t bring most of its existing customers into its AI future, what hope does it have of winning over those wedded to the Apple or Samsung ecosystems?

I can only speculate that the ad hoc nature of Google’s AI roadmap makes it impossible for seemingly incommunicable internal development teams to operate on the same chapter, let alone the same page of the AI ​​playbook. Frustratingly for consumers and the industry at large, Google has by far the largest number of components it can integrate, operating on both the developer and product sides of the AI ​​market. Instead, it risks being beaten by seemingly less well-equipped rivals, such as Samsung’s AI-powered Galaxy S24 series, which uses Google’s own technology.

Unfortunately, these are just the latest in a growing list of tense incidents caused by the explosive arrival of OpenAI’s ChatGPT. Early versions of Google Bard gave incorrect answers to questions, hallucinated facts, and had difficulty with simple math operations. Google’s model has since improved, but in its early days Bard was more of a knee-jerk reaction than a planned product that led the market.

After rushing to get something (anything) to market, Google was left cobbling together last year’s AI plans into something close to a coherent strategy. However, this year’s confusing Bard/Gemini rebranding has left consumers, business users and investors scratching their heads, wondering which Gemini version does what and what subscriptions or services are required. Even the AI ​​business strategy appears hasty.

It’s frustrating, and anyone familiar with Google’s product history should have seen this coming. The brand is notorious for sporadic product launches and short attention spans for those who don’t make the cut. While this is bad enough under normal circumstances, AI development is fraught with additional difficulties that make a cautious approach to innovation even more preferable.

Between the risks of bias and abuse, privacy concerns, regulatory compliance, and well-documented ethical pitfalls, “move fast and break things” is not a recipe for AI success. It may even be destined to fail. In contrast, OpenAI has been more cautious in its rollout and messaging. Despite missteps and setbacks, consistency has contributed significantly to its position as market leader.

The AI ​​arms race still has a long way to go, and recent setbacks can be overcome. Maybe Google’s deep ecosystem has enough potential to make AI integration a winner. In this case, in the distant future, this may all be seen as teething problems. However, Google is no longer a leader in artificial intelligence, and failure to learn from its recent mistakes could do more than damage its reputation.