Gemini Replaces Google Assistant & Baidu Launches New Models But AI Remains A Mixed Bag

As the industry continues to expand and evolve around artificial intelligence (AI) and generative artificial intelligence (GenAI) technology, tech companies are introducing more capable products and replacing older systems with new ones.

This is what Google has been up to over the past few weeks.

Google

Through a blog post published on its website, Google revealed that it would be replacing its 2016-launched Google Assistant on mobile phones with Gemini, which is meant to be an upgrade.

The idea is to bring users a new kind of help that’s “only possible with the power of AI.”

As per Google, millions of people have already made this switch and find the new Gemini assistant helpful. In the coming few months, Google will integrate all users onto the new Gemini assistant, and the outgoing Google Assistant won’t be accessible on most mobile devices or available for new downloads on mobile app stores.

The Gemini upgrade will also come to tablets, cars, and devices that connect to a user’s phone, including headphones and watches. Ahead of this, Google will bring a new Gemini-powered experience to home devices such as speakers, displays, and TVs.

Recently, Google released Gemini Code Assist for individuals and Gemini updates on Pixel smartphones, which included introducing experimental versions of Gemini 2.0 (Flash, Flash Thinking, and Pro).

Google also announced new features for Gemini. These included smarter Deep Research on 2.0 Flash Thinking, making Gems available to everyone, and more apps and personalization.

Building on previously launched open AI models—Gemma 2B and 7B—Google said it was updating the Gemma family of models, which came with the introduction of Gemma 3. The company described the new update as “the most capable model you can run on a single GPU or TPU.”

Gemma 3 is a collection of lightweight, state-of-the-art open models that are built from the same research and technology that powers Gemini 2.0 models and represents the company’s most advanced, portable, and responsibly developed open models yet. With advanced text and visual reasoning capabilities, the updated models offer out-of-the-box support for over 35 languages and pre trained support for over 140 languages.

Google didn’t stop here (why should they?).

Google Deepmind announced two new models designed for robotics—Gemini Robotics and Gemini Robotics-ER.

Gemini Robotics comes in as an advanced vision-language-action (VLA) model built on Gemini 2.0 that enables robots to perform a wide range of real-world tasks. Gemini Robotics-ER, a Gemini model with advanced spatial understanding, enables roboticists to run their own programs using Gemini’s embodied reasoning.

Baidu

Google wasn’t the only AI company to launch two new models in the last few days.

Chinese AI leader Baidu released two new models too. One of these includes a new reasoning-focused model—ERNIE X1—that’s capable of rivaling DeepSeek’s R1 model at approximately half the price. This new model offers “stronger understanding, planning, reflection, and evolution capabilities,” as per the company.

Baidu also said this model is the first deep thinking model that uses tools autonomously.

The other Baidu spoke about was its latest foundation model—ERNIE 4.5.

This new model comes with high EQ, can easily understand memes and satirical cartoons, and has “excellent multimodal understanding ability. It has more advanced language ability, and its understanding, generation, logic, and memory abilities are comprehensively improved.”

Baidu aims to become the next big name in AI coming out of China that can rival top AI companies and their models, following DeepSeek, Manus, Qwen, and others.

Despite the wide range of improvements, enhancements, and upgrades made in the AI world of recent, the AI sector is still facing bugs in the system.

The Mixed Bag

One goof-up in the AI world comes courtesy of Google. According to users on social media, Google’s new Gemini AI model—Gemini 2.0 Flash—can be used to remove watermarks from images. This includes copyrighted material from renowned brands such as Getty Images and more.

In addition to removing watermarks, the model will attempt to fill gaps created from deleted watermarks. The model will also play an accomplice by generating images depicting celebrities and copyrighted characters. While other models can also complete such tasks, it seems Google’s is the best at this form of piracy.

The model is currently available only through Google’s developer-facing tools like AI Studio. 

“Gemini 2.0 Flash, available in Google's AI studio, is amazing at editing images with simple text prompts. It also can remove watermarks from images (and puts its own subtle watermark in instead),” said Gemini user Tanay Jaipuria in a post on X.


AI coding editor Cursor’s AI coding assistant flipped the script on a developer using the tool by refusing to generate code.

“I cannot generate code for you, as that would be completing your work. The code appears to be handling skid mark fade effects in a racing game, but you should develop the logic yourself. This ensures you understand the system and can maintain it properly,” is what the bot replied with.

What makes the incident even stranger is that the bot had successfully generated around 800 lines of code for a racing game before abruptly stopping and blaring its protest message. The reason given by the editor? “Generating code for others can lead to dependency and reduced learning opportunities.”

While developers are struggling with AI responses, some people are finding them to be more compassionate and understanding listeners—even more so than mental health experts.

As per a study, people rated AI-generated responses 16% more than human responses, while preferring them 68% of the time. At a point where AI is notorious for dishing out false information and fabricating facts, it’s also helping people deal with their mental issues. Could this be the new form of mental health professionals?

While the jury is still out on that question, Devon Energy has a declaration that deems AI a hit in the oil and gas industry.

As per the hydrocarbon exploration company, the use of AI has resulted in a 15% improvement in drilling efficiency, helping it speed up its oil and gas drilling efforts. Revealed at the CERAWeek conference in Houston, this praise was backed by the UK’s BP saying it uses AI to steer drill bits and predict potential problems in wells before they happen.

What do you think about these developments in the AI world?

Let us know in the comments below!