Large model manufacturers have made intensive efforts, and Google has also opened a 'volume': Gemini chat robot has been replaced with a new model, and it can also check the output content with one click.

As large model manufacturers such as Meta and OpenAI are making intensive efforts, Google has also announced a major update -

From now on, the Gemini chat robot will be driven by Gemini 1.5 Flash.

Compared with the previous version, the window length has been increased to 4 times, and the response speed is also faster.

According to Google’s introduction, the 1.5 Flash model behind the new version of the chatbot focuses on lightweighting and speed improvement.

Of course, the quality of model reply has also been improved, and the context window has been increased from the original (based on 1.0 Pro) 8k to 32k.

In addition, the new version of the chatbot also adds a "fact checking" function, which can detect whether the generated content is true with one click, reducing the adverse effects of model illusions.

Some netizens lamented that Google’s performance was really strong today, first, and then Gemini also released an update.

Some people have turned on the wishing mode and look forward to adding AI functions to Google Scholar academic search.

Longer context window and faster speed

The main content of this update is to change the model behind the free version from 1.0Pro to 1.5 Flash.

Gemini 1.5 Flash was first unveiled at the Google I/O developer conference in May.

Through the "distillation" of training data, Gemini 1.5 Flash achieves higher generation quality with a lighter size.

And the small size also makes the model faster and more efficient, and it also supports multi-modal reasoning.

Google introduced that after this model change, the chatbot will become faster, and the context window of the old version of 8k will be expanded to 32k.

However, 1.5 Flash itself supports 1 million contexts. This reduction is indeed not small, but it is a free version after all.

In addition to the model upgrade, another important update is the fact-checking function.

In the latest Gemini chatbot, you can use this function to check the output content with one click.

The system will search and compare the content in the output on Google, and then mark the matches and discrepancies.

Some netizens commented that when they saw OpenAI launching GPT-4o mini, they felt that it was only a matter of time before Google launched the new version.

Indeed, not only OpenAI and Google, but also Meta, Mistral and other manufacturers that are making large models have been making frequent moves recently.

As for the performance of the model, this netizen also said that he has tried 1.0 Pro and 1.5 Flash. The performance of the two is almost the same, and the 1.5 Flash is faster.

So, this wave of operations by Google is also adapted to the recent trend of "model lightweighting" to a certain extent.

So, how is the performance of the Gemini chatbot after replacing 1.5 Flash?

Check model output with one click

Qubit performed a simple test on the new version of the chatbot.

First, let’s take a look at the fact-checking function of this update. The first step is to ask a question just like a normal conversation, and Gemini will answer it normally.

You can see that there is a Google logo under the answer. This is the button for the fact-checking function.

After clicking, the system will automatically search on Google and then compare it with its own output.

After the comparison is completed, the source can be searched and the matching content will be highlighted in green. If there is any discrepancy with the search results, it will be marked with a light red background.

Click on the marked location to see the content link Gemini uses for comparison.

It should be noted that such annotation does not mean that the output content is wrong. For example, in the comparison data cited here, Tom Cruise’s mother is Marry Lee South.

Due to text mismatch, this part of the answer was marked by the system, but in fact both are correct answers.

Since this fact check relies on Internet searches, the quality of the comparison information is also uneven and may not be 100% accurate.

For example, regarding the classic joke "Lin Daiyu pulls up the weeping willow upside down", Gemini clearly gave the correct answer, but was marked red.

Looking at the comparative information cited again, it is indeed a bit difficult to stretch.

So the function of this function is mainly to provide a more convenient verification method, but how to accept it depends on multi-party verification and the user’s own judgment.

In addition, regarding the model itself, we also tested several recently popular problems that often cause large models to hit the wall.

For example, Gemini even converted two numbers into money, but after one operation, the final result was... wrong.

Since this question was discovered, if this is the second funniest answer, no model dares to claim the first place.

I also gave a wrong answer at the beginning, but I corrected it later during the analysis process.

But if you ask questions in English, there is still hope that you can get the answer right.

As for the other question, this answer can actually count letters from Chinese... It also makes people confused, and it is not within the prediction at all.

Finally, regarding the speed improvement mentioned in this update, testing found that Gemini 1.5 Flash takes less time to output the first word than Claude 3 Haiku. The difference in subsequent speeds is not very obvious with the naked eye. .

The above is the performance of Gemini 1.5 Flash in chatbots. Interested readers can try it on their own.

Reference link:

[ 1 ] https://blog.google/products/gemini/google-gemini-new-features-july-2024/

[ 2 ] https://x.com/GeminiApp/status /1816512086232731696

The above is the detailed content of Large model manufacturers have made intensive efforts, and Google has also opened a 'volume': Gemini chat robot has been replaced with a new model, and it can also check the output content with one click.. For more information, please follow other related articles on the PHP Chinese website!