GPT-4 and Claude Opus avoid making the same mistakes in common-sense reasoning. Despite being an AI pioneer, Google seems unable to deliver models that match the caliber of their competitors.
Google was compelled to add AI features because Microsoft incorporated ads into Bing AI answers. Generative AI poses an existential threat to Google's primary revenue source—Search Ads.
Why is Google struggling to compete in AI? While I don't have insider knowledge, here are some potential reasons:
Google prioritizes "safety" over quality. "Safety" encompasses many aspects, including a significant emphasis on being politically correct.
Connected to the previous point, there is an overemphasis on research over product development. Google excels at creating impressive demos but lacks the same culture of continuously operationalizing improvements.
Bureaucracy: Google tends to over-engineer systems. The excessive layers of management hinder their ability to innovate as quickly as their startup competitors.
Real-time results: Google incorporates near real-time results, which do not undergo the same extensive training process as other systems. Nevertheless, a layer of common-sense reasoning would have been feasible if their models were more advanced.