This month has been packed for Google as it ramps up efforts to outshine OpenAI in the AI space. On Thursday, Google unveiled its latest innovation: Gemini 2.0 Flash Thinking Experimental, an AI model employing runtime “reasoning” techniques, similar to OpenAI’s o1, to deliver more in-depth problem-solving capabilities.
Building on the recently launched Gemini 2.0 Flash, this experimental model operates within Google’s AI Studio platform. However, early testing by TechCrunch’s Kyle Wiggers highlights some issues, such as inaccurately stating that the word “strawberry” contains two R’s. Unlike standard AI models, reasoning models use feedback loops and self-checking mechanisms—first seen in early 2023 projects like “Baby AGI.” These methods enhance accuracy but require significantly more computing time, adding seconds or minutes to response speeds. Google DeepMind’s chief scientist, Jeff Dean, noted that extra computing power improves outcomes, tweeting on X, “We see promising results when we increase inference time computation!”
Since OpenAI introduced “reasoning” AI with o1-preview in September, competitors like DeepSeek and Alibaba have followed suit, launching their own reasoning models. While these models excel in handling complex problems, their high computational costs raise questions about long-term feasibility. For instance, the expense is a key reason OpenAI’s ChatGPT Pro costs $200 per month.
Despite concerns, Google appears committed to this approach. Logan Kilpatrick, part of Google’s AI Studio team, described the Gemini 2.0 Flash Thinking Experimental model as “the first step in our reasoning journey.” It’s clear that Google intends to be a leader in this rapidly evolving AI field.
You can now use your phone line to call ChatGPT when cellular data is unavailable.
New to software development, try the HTML and CSS quiz to test your proficiency
Author
