On April 1, local time, Openai officially announced that from now on, users can access their free ChatGPT to generate artificial intelligence (AI) chat robots without registration.It is worth noting that the open version of ChatGPT is based on the GPT-3.5 model.

With technology giants and rookies such as Google, META, Anthropic, and Mistral AI, they have continuously launched their unique AI products to generate unprecedented competition in the field of AI.OpenAI announced at this time to relax the use limit. What is the intention?

According to foreign media reports, the data called "AI Times New Lithium Mine" is one of the main considerations of OpenAI.As GPT-5 is about to be born, the company may need to collect data to train more AI models.

In addition, OPENAI also aimed at the long -term search engine market that was monopolized by Google, hoping to open up a new road.This can also take a glance in the recent interviews of the company's CEO Aldman.

For the majority of users, the lower threshold of OpenAI is an exciting news.On the day of OpenAI's official announcement, ChatGPT was downturn due to massive access requests.However, from the perspective of professionals, completely open use may bring the hidden concerns of AI being abused.

Two major considerations: data and search business

Since the first launch of the public at the end of 2022, OpenAI has been asking users to register the OpenAI account for users who want to visit the chat robot.At this time, let's relax the entry threshold, what does Openai mean?Chennai Investment

According to foreign media reports, one of the reasons may be data.As we all know, the super -large model that provides power for generation AI relies on massive data (603138), and data is called the "new lithium mine" in the AI ​​era.With the "New York Times", YouTube, Reddit, Stack Overflow and other content providers no longer allow technology companies to use their data at any time to develop AI at any time. The current high -quality data for technology companies to train large models has become increasingly increasinglimited.

The competition in the field of generating AI is in full swing. In order to feed the big model of its own, technology companies have to use their best solutions to find training data.Openai is open to use ChatGPT, which may be based on this consideration.

Technology media ARS Technica analysis, the reason why OpenAI allows ChatGPT for free is precisely because "your (user) is a product".As GPT-5 is about to be born, opening the use of ChatGPT permissions can help OPENAI collect data and help companies train future AI models.As OpenAI said in a statement, the content provided by the user may help improve its model, but it also said that the user can turn off the function.

In addition to getting more data for training models, OpenAI also targets Google's business that has always occupied the leading position -search engine.

Foreign media commented that this provides users with a barrier -free entrance, which may help ChatGPT as a substitute for Google search.On the other hand, OpenAI can provide users with sales to provide sales by providing users to provide sales, thereby winning potential new customers.

And this is also possible to follow.In an interview with Lex Fridman, a well -known American technology podcast of the United States last month, OPENAI CEO Aldman once mentioned Google Search. He said that the current Google search experience is "boring".The search method that has been known for many years.

However, it is not easy to subvert Google search.According to Statcounter, as of February this year, Google Search still occupies 91.6%of the global search market.In the past 20 years, many "Google killer" came and went again, even Microsoft's bing has not been able to succeed.Pune Stock

As Aldman wants to reconstruct a new type of search, advertising may become the next profit choice.Over the years, Google has been loved by people because of its minimalist advertising experience, but in Aldman's view, the situation is no longer the case."I am convinced that a way of advertising is meaningful to a big model. I am sure that a fair way can be found in this transaction." He said.

The inevitable move under the intensification of competition, experts: There are hidden concerns

Recently, big models such as Gemini 1.5 Pro and Claude 3 have been released one after another. In the face of competitors' stares, Openai has to make adjustments to attract more users.

According to the traffic tracking website Similarweb, although ChatGPT has about 1.6 billion visitors in February, it is still the most visited AI chat robot application in the world, since May 2023 (at that time its traffic reached 1.8 billion network visits peak value peak value at that time) In the past, the growth rate has slowed.

Image source: Similarweb

With the further use of ChatGPT, can the computing power of OpenAI still support it?

In fact, the reasoning cost of GPT models has now fallen sharply.As early as April last year, the media revealed that the daily operating cost of ChatGPT was about $ 700,000.At that time, foreign media Analytics India said that if OpenAI cannot be profitable quickly, the cost will make the company bankrupt in 2024.

However, OPENAI not only did not close because the operating cost was too high, but in January this year, OpenAI also reduced the input price of GPT-3.5 Turbo (GPT-3.5 upgraded version) by 50%. This is the company's first year.Three price reductions.This may mean that the cost of GPT-3.5 Turbo has been further reduced.According to foreign media reports, the cost of GPT-3.5 Turbo is lower than the Claude 2.0 and Claude 2.1 of Anthropic.

For users, it is an exciting news to lower the entry threshold of OpenAI, but from the perspective of professionals, it may bring some hidden dangers.

"Daily Economic News" reporter noticed that AI researcher Simon Willison mentioned on the X platform that if this is a way to bypass the cost of paying OpenAI services, the possibility of AI being abused is greatly improved."I want to know how their grabbing prevention is? I think, the temptation of abuse of free GPT-3.5 API will be quite strong." He wrote.

Image source: X platform

Given that the open-use ChatGPT is based on GPT-3.5, Ethan Molick, a professor of AI at Wharton School of Business, also expressed concerns about this model."So far, people's views on the AI ​​model function are largely affected by GPT-3.5. As we mentioned, GPT-3.5 has much worse ability, and it is moreIt's easy to make a lie. "He said.

He believes that "OpenAi made a mistake, which may make it more difficult for people to see what AI can do because it (GPT-3.5) is far behind GPT-4".

Image source: X platform

Daily Economic News

(Editor in charge: Zhou Wenkai)