ChatGpt is no longer the best AI

That’s not always the perfect generative artificial intelligence more powerful. In fact, from the users’ point of view, the best chatbot is the one that is best suited to the task it has to carry out.

For this reason one might say that ChatGpt (OpenAI) is optimal for producing creative text, for example, and Gemini (Google) to obtain an explanation of a complex concept or – even better – a precise translation. Copilot (Microsoft), however, is also capable of creating images, moreover for free. And then there are AI like Perplexity And You.com which represent the ideal destination for those looking for information on recent events.

But the thing is There are not only consumers.

Big techs compete in the field of AI to conquer even the largest fish: billion-dollar companies, institutions and small and medium-sized businesses. In the future, they will all need an artificial intelligence model to improve their services and manage their businesses. And each of them, inevitably, will be attracted by the technology they can boast the best performances.

This is why it makes a lot of noise that Gpt-4, the most capable version of ChatGptis no longer the most powerful AI model around.

Recently, in fact, OpenAI’s most advanced artificial intelligence (Gpt-4 Turbo) was surpassed for the first time by Claude 3 Opus in the Chatbot Arena, a public ranking system, used by researchers, that lists the performance of various algorithms or models based on specific metrics or benchmarks.

The Chatbot Arena was first established on May 3, 2023. And the Gpt-4 model was included seven days later. Since then OpenAI artificial intelligence has always occupied the top of the rankings. Until a few days ago, when it was ousted by an advanced version of Claude, the AI ​​developed by an American startup called Anthropic.

Anthropic was founded just three years ago by two Italian-American brothers: Dario and Daniela Amodei, who are now CEO and president of the company, respectively. The Amodei, in the past, worked for OpenAI. They left because they wanted to develop a more “ethical” and more “responsible” artificial intelligence. They created it and named it Claude in homage to Claude ShannonAmerican mathematician and engineer considered “the father of information theory”.

Competition from Claude – which is not yet available in Italy – is becoming increasingly fierce and is starting to worry OpenAI.

From a consumer perspective, the difference between Claude and ChatGpt is still minimal and mostly boils down to Claude’s ability to analyze longer texts: those who use the “Pro” version, for example, can “upload” documents that contain up to 200 thousand tokens, approximately 350 pages of text.

But Anthropic’s technology is set to get even better thanks to the river of dollars that is coming from Amazon. The Seattle giant, in fact, has just announced that it has invested another 2.75 billion dollars in the Amodei brothers’ company. Amazon’s total investment in Anthropic thus rises to $4 billion. And the company’s estimated valuation jumps to $18.4 billion.

OpenAI is still ahead on the investment front – Microsoft has poured 13 billion dollars into its coffers – and as regards the market valuation, estimated at 80 billion dollars. But the increasingly solid alliance between Anthropic and Amazon is starting to put some pressure on the San Francisco startup led by Sam Altman.

Also because ChatGpt’s problems are different. And on multiple fronts.

From the beginning of 2024, users registered for the version “Plus” by ChatGpt, who therefore pay 20 euros per month (plus taxes) to access the best performing model, which is also capable of generating images in addition to text, they complain a certain “laziness” on the part of the AI.

ChatGpt’s responses, users claim, sometimes appear incomplete and generally lackluster. It’s not about sensations. Sam Altman himself, last February, referred to the problem of his creation, writing on X that an update would make ChatGpt “much less lazy”. And the responses, for a while, actually seemed better. Until a few days ago, when “laziness” became a hot topic again on one of the main communities linked to OpenAI.

“The reality is that ChatGpt has become unusable,” wrote a developer on the OpenAI forum. While others have pointed out that Claude 3, the latest version of Anthropic’s AI, does a better job with writing codes.

Added to the performance problems are those concerning the “traffic” of the platform, which is an indication of the health status of a chatbot. On this front ChatGpt – which boasts more than 100 million active users per month – is still ahead of Anhtropic and even Google Gemini.

But there wasn’t, for example, a surge afterward the launch of the Gpt Storethe showcase dedicated to personalized AI – called Gpt – that users can create in minutes.

A recent analysis of Financial Times highlighted that the research and use of Gpt are only responsible for the1.5% of visits received by ChatGpt in February 2024 (referring only to the desktop version).

OpenAI could give new impetus to its popular AI by soon releasing a new, even more advanced model. It is rumored that the long-awaited update to Gpt-5 it could arrive by next summer.

By the end of 2024 it should also debut Sorathe new artificial intelligence from OpenAI capable of creating realistic videos starting from simple text.

v