Hello, Google: AI’s previously recorded offers are the coward road to go out

Each major technical event these days includes artificial intelligence updates, often with a list of live illustrations – and sometimes these experimental offers fail. But some companies evade these restaurants by registering the main offers. I call these cheese moves.

Last year It was made by Google The event, Gemini failed twice during a live demonstration. Although moments like this is undoubtedly embarrassing to companies, they add a layer of originality that you do not get with a pre -registered major event. Unfortunately, Google chose the previously recorded road for Tuesday Android view: I/O Edition. Poetry of coordination in the way that was filmed and shining for my desire, and I was stripped of a feeling of reality that comes with live illustrations and warts.

During Android: I/O Edition, we saw a demonstration of Gemini advice that shares makeup, helping someone find time to lunch in their busy schedule, and give a summary of Jin Austin’s pride. Because these were previously registered reactions, Gemini dealt with requests with APLomb – there are no contracts or horizon problems. But the tests show that artificial intelligence models are routinely mistaken.

According to the artificial intelligence test site LivebenchGoogle Gemini 2.5 Pro Preview In general, about 79 % of the time. This is not bad, but it is not great either. Despite this result, this model of Gemini is still one of the best artificial intelligence models tested for the site, as it lost to only two other models: Openai’s O3 High and O4 Models.

Certainly, there is nothing perfect, devices and programs have errors. But if you give me a calculator and promise it, it works all the time, but in reality, it is wrong 20 % of the time, and this looks like a great contrast.

Since Gemini outperformed most of the other AI models tested LiveBench, there is a good opportunity to use Gemini, even if the live demonstration stops. But since Google has chosen a great offer, I am having difficulty knowing what to believe.

Look, I understand why the company wants its product to work properly in its own event. But showing artificial intelligence tools that make mistakes seem more honest than behavior like a perfect tool. These capabilities are flawed, and this is good, but be honest with people about those defects and show your new features at work. Do not sell me smoke and mirrors.

To learn more about Google, here is what to know Android 16 and Articles 3 Expressive Design.