What AI has surpassed GTP-4 and is approaching the top spot in the performance rankings?

Until now there was a very clear dominance in terms of artificial intelligence. The linguistic model GPT-4 From OpenAIwhich generated a giant explosion in 2023, forced big tech companies to direct their efforts – and investments – to catch up. Google launched Bardits conversational artificial intelligence that imitates the style of GPT and, at the beginning, although the results were good, They weren’t even close to ChatGPT..

As time went on, the company behind the Internet search engine became fully involved in the matter and managed to create increasingly complex language models.

This was finally revealed with the launch of Gemini and has now proven to be the real winner. At least later surpass the basic version of GPT-4 and rose to the top of the charts performance rankings as one of the most powerful AIs that exist today.

AI performance: the fight for first place

Today, OpenAI is a leader in the consumer AI space with ChatGPT at the forefront, powered by GPT-4. This technology, in which Microsoft invested $10 billion, was used to power Copilot and integrate with DALL-E 3 as an image generator.

However, Google it’s not willing to be left behind, and its recent moves demonstrate that it intends to be quite successful in this race for AI dominance.

One of Google’s first initiatives was the introduction of bard, an artificial intelligence designed to completely change the way we search. However, it did not fully meet expectations due to various shortcomings and stiff competition.

To correct these shortcomings, the creator recently launched its new artificial intelligence models called Gemini Nano, Gemini Pro and Gemini Ultra, with significant improvements compared to the previous model and which, at first glance, achieved a big advantage. This is reflected on the platform lmsys.orgwhich ranks AI models based on their performance.

In particular, Google Bard with Gemini Pro is currently in second position in the ranking of the most powerful AI, surpassed only by GPT-4 Turbo, the most powerful AI on the market.

This can be seen in the LMSYS ranking of AIs, and although it is not exact, as many will prefer one or the other AI depending on what they use it for, it is clear that it offers a rather interesting view of which companies are emerging in the market. in recent times in the development of the smartest chatbots.

However, this race could take a turn. The success of Gemini Pro only intensifies the curiosity towards it Gemini Ultra, which has not yet been presented and represents the company’s most powerful model. Furthermore, attention is also on OpenAI’s GPT-5, with highly anticipated results. This creates a real race to see who will launch their new AI first.

Experts warn that if OpenAI does not quickly introduce a new model, you risk losing the advantage that you currently have.

It is crucial to note that in addition to Google, Meta is also developing its own model with the potential to surpass GPT-4. Therefore, this competition will ultimately lead to users having better tools for their daily use.

More and more companies are trying to achieve artificial intelligence hegemony, which seems to be increasingly complicated. However, they are starting to take shape a specific type of brand evident behind the most powerful language models: a powerful artificial intelligence, with proprietary code, with a gigantic data sheet and which perhaps violated copyright to learn.

Therefore, although open source models exist, such as Meta’s LLaMA 2, we find that AI training processes tend to be rather opaque and offer us a quite questionable level of guarantees when it comes to knowing how and in what sense they were trained . This inevitably means that AIs face biases and problems in their own structure.

In any case, it can be clearly deduced that Google’s Gemini Pro language model has succeeded in ensuring that its artificial intelligence reaches the necessary potential in this market.

New from Bard: Image creation, data validation, and more

Google’s recent AI update, announced Thursday, allows its users to do just that create high-quality photorealistic images relying on simple written instructions and turning any idea into reality.

Furthermore, new updates have been presented to improve the tool’s capacity: from corroborate the answers given in the conversations and the availability of Gemini Pro in all languages.

The ability to create images, like its rivals Midjurney or Dall-E, comes from the integration of the company’s most advanced text-image processing technology “Image 2”, which allows everyone to rethink the way they create images, what is it get creative help with Bard. For now, it will be available in English in most countries and territories where the tool already works.

To help everyone differentiate original images from those generated with Artificial Intelligence, Bard uses SynthID, a tool developed by Google DeepMind, which adds an imperceptible watermark directly into the pixels of new images.

In turn, it introduced the new “Double Check” feature to corroborate all the answers offered by Bard, which will be available in more than 40 languages ​​and 230 countries and territories.

When you start a conversation, you’ll see a “G” icon below the response, allowing you to Google Bard-generated responses and dig deeper into that information, as well as identify different sources that validate that data. You can also click on the highlighted sentences to get information that supports or contradicts what you answer.