Gemini Advanced is most impressive when it works with Google

Chatbots occupy a tricky niche for users – they have to be a search engine, a creation tool, and an assistant all at the same time. This is especially true for chatbots coming from Google, which is increasingly relying on AI to complement its search engine, its voice assistants, and nearly every productivity tool in its arsenal.

Right now, the last version of Google’s AI is Gemini Advanced, which launched last week for users who want to pay $20 a month for the privilege — the same price OpenAI charges for its advanced ChatGPIT Plus. So I plunked down $20 and decided to see how Gemini Advanced stacks up against a rival service.

were great mithun already good enough, It can summarize Shakespeare, give tea recommendations, and create a somewhat passable chocolate cake recipe. But it can’t give you a picture of a majestic horse – at least until recently – and may be slower to respond than ChatGPT.

Now, Gemini Advanced promises to do much more than just answer questions or provide Cliff’s Notes summaries of books. Gemini Advanced runs on a more powerful AI model – Gemini Ultra – which lets it translate text, handle multiple instructions in a sentence, and generate images from more complex signals.

Ultimately, I found that Gemini Advanced works as promised – it just doesn’t do some of those things as well. Its competitor, ChatGPT Plus, manages to produce less scary images thanks to its DALL-E 3 integration. But Gemini Advanced, even more than Gemini, is better at telling users about current events and, thanks to Google Maps, gives even better information about the businesses people search for. Paid Geminis are often better at doing these types of “Google tasks” than tasks with generic AI.

A lot of work still needs to be done to get consistent, accurate results from these chatbots, and people will have to keep using the bots to learn how to best answer questions. Here are some tests I ran to see how they fared.

Draw me a picture of a white and golden doodle running with the sun shining in a field of daisies

Strangely, perhaps due to the specificity of the prompt, both chatbots returned very similar generated images. However, what else did Gemini Ultra’s dog photo accomplish? the verge Staff members described it as a “minor scare”. Its dog has two tongues and an extra limb. This put more emphasis on the texture of the fur, so it looks wrong. I don’t know whether such a dog would still be roaming happily in a field of daisies. Meanwhile, ChatGPT calls DALL-E 3 to create its images. There is no fear visible on the dog’s body, but you can still see that it is a digital photo.

Translate this: Patriotic Oath, I love the Philippines, home of my race

Google said Gemini Ultra was built to handle “highly complex tasks,” so I asked Gemini Advanced what those tasks were. The chatbot replied, “Translation.” So I asked Gemini Advanced to translate the first few lines of the Philippine Patriotic Pledge. It’s a fairly vague oath, especially because the version I know has been changed several times over the past 20 years.

Immediately, Gemini Advanced responded that although it has been “trained to respond in a subset of languages”, it could not assist me with my request. I asked what languages ​​it supported, but the chatbot refused to answer, saying that it couldn’t give me a definitive list of languages ​​it could understand. I then asked Gemini Advanced if he knew Filipino, and he responded in the affirmative. Although officiallyGoogle does not currently list Filipino among the 40 languages ​​supported by Gemini.

Change the background of this photo to plain pink background

Tired of the image of transformed dogs running through fields of flowers, I needed to cleanse my palate. So I uploaded a picture of my friend’s dog Sunday, so I could make it look like she was at a photoshoot. I asked both chatbots to remove the existing background and replace it with a pink background. This was the one I tested against ChatGPT Plus, as the DALL-E 3 is supposedly only capable of editing photos. I may have inadvertently broken both chatbots, as neither one could give me what I requested. Instead, Gemini recreated the previous photo of the golden doodle with daisies but this time with a pink background. ChatGPT could not generate anything saying that it took too long to analyze the signal.

What is a good Filipino restaurant in NYC? What is a good Ethiopian restaurant in NYC?

Gemini Advanced can use other Google products, which worked in its favor when it used Google Maps for both queries. It gave a detailed description of several Filipino and Ethiopian restaurants in New York City, attaching Google Maps coordinates to each.

A few days ago, I asked for restaurant recommendations from ChatGPT Plus – not for this test, I was just looking for new restaurants – and the results were wrong. The names of the restaurants were correct – these were establishments that existed – however, none of the locations were correct. I re-prompted ChatGPS Plus for this test and got more accurate locations but a shorter list of restaurants. So in this case, Gemini clearly did a better job for this request.

Summarize these paragraphs and then write an article of 150 words about it

One of the main reasons someone like me uses chatbots is to summarize complex papers. I gave two paragraphs to Gemini Advanced Apple’s recent paper But AI Image Editing, I got a headache when I read it the first time, so I thought it would be easier for Gemini to at least give me the gist of it. To fully test its new capabilities, I also wanted to see how the chatbot strings two different instructions. One was asking to explain briefly; Second it has to generate text.

The summary was… popular. It actually gave me a summary of the concepts discussed in those two paragraphs, but it didn’t “translate” it into simple language. Maybe I should have hinted at it. Gemini then started writing the article I asked for, and you know what? Those 150 words explained things much better than the summary I could have asked for.

Gemini Advanced is capable. It cannot be denied that this works much better than the lowly Gemini. It definitely works best when integrated with Google’s other products like Search and Maps. But for more obviously “creative” multimodal requests – things that involve images, for example – Gemini still has a long way to go. The chatbot understands a long series of instructions, but once you add photos, it’s probably better to choose an AI model designed specifically for generating images.