Not the AI search that everyone was expecting, the focus of competition suddenly shifted to AI intelligent assistants.
Recently, OpenAI has launched the all-around model GPT-4o, which can accept multiple inputs and generate corresponding outputs, demonstrating new capabilities in millisecond level response and multimodal interaction. Meanwhile, Google showcased its AI assistant Astra and flagship model Gemini at its I/O developer conference.
Some industry insiders believe that OpenAI did not achieve the expected breakthrough this time, but instead integrated existing technologies. In addition, Google's layout and innovation in the AI search field, as well as its efforts in optimizing smartphone operating systems, demonstrate its deep accumulation and strategic layout in the AI field.
As the competition reaches its peak stage, it seems that it has bid farewell to pure technological competition and is more about competition in applications and user experience. When the influencing factors become complex, how likely is OpenAI, which focuses on investing in the forefront of large models, to become a winner?
Assault on Google, OpenAI announces AI personal assistant first
The anticipated war surrounding AI search did not start, and the focus shifted to AI intelligent assistants.
On May 13th local time, OpenAI held a press conference the day before the Google I/O Conference and released its latest product, GPT-4o, which stands for Omni, meaning "versatile". According to the OpenAI official website, GPT-4o is a step towards more natural human-computer interaction, as it accepts any combination of text, audio, and image as input content and generates any combination of text, audio, and image as output content.
OpenAI Chief Technology Officer Mira Murati stated at a press conference that the speed of the GPT-4o is twice that of the existing GPT-4Turbo, but the cost is only half of it. GPT-4o can perform real-time inference on text, audio, and images, with response times almost reaching human level.
During the 26 minute live broadcast, GPT-4o demonstrated a series of new abilities, including millisecond level response, recognition of human emotions for audio and video interaction, and multimodal input/output capabilities. At the same time, GPT-4o covers desktops and apps and is completely free to provide to users.
Google, on the other hand, demonstrated its versatile AI capabilities at its I/O developer conference. At the meeting, more than ten products were continuously released and updated, including the AI assistant Astra, the cultural image model Imagen3, the cultural video model Veo benchmarking Sora, and the flagship large model Gemini.
In Google's demonstration video, when using the AI assistant Astra, simply turn on the phone's camera, aim at any item, and the AI can accurately say the name of the item. As long as the camera of the phone is aimed at an object, Gemini can recognize it, such as a red apple, and can also answer questions such as "What can sound in the lens?".
In addition, Google has stated that it will expand Gemini's multimodal features during the summer, including the ability to engage in deep two-way conversations using voice, a feature called Live. Through GeminiLive, users can converse with Gemini and choose the sound it responds to from various natural sounds. Users can even speak at their own pace, or interrupt and clarify questions during the answering process, just like in any human conversation.
According to media reports cited by the Daily Economic News, Apple has recently been revealed to be finalizing an agreement with OpenAI to introduce some of its technologies into the iPhone this year. At this press conference, Google's Vice President of Product Management, Sameer Samat, clearly stated that Google will further optimize the Android operating system through Gemini. This optimization will first be reflected on Google's own phone Pixel.
GPT-5 absent, OpenAI slows down?
For the sudden update of OpenAI, the industry is no longer unified in admiration. "Although the press conference was stunning, Google should not panic after reading it," said Fu Sheng, Chairman and CEO of Cheetah Mobile and Chairman of Orion Starry Sky, on his personal Weibo account.
Weibo screenshot
In a short video released on May 14th, Fu Sheng mentioned that "all domestic artificial intelligence practitioners are staying up late waiting for nuclear bombs to be fired on the other side of the ocean, but they didn't expect the bombs to be fired and instead pulled out a pile of dropped cannons.". He said that although such comments are a joke, what is more disappointing is that OpenAI did not release GPT-5.0, not even GPT-4.5, but instead released GPT-4o, "which combines a series of engines, such as images, text, and sound, so you don't need to switch back and forth.".
However, Fu Sheng also stated that "OpenAI has made great efforts to enable more users to use it this time, with a series of applications, API price reductions, and GPT free. Of course, we hope that OpenAI can make this industry develop better, and we can also learn seriously. This press conference truly tells us that applications have great potential, and everyone should work hard.".
"The release of GPT-4o has made significant progress compared to before. Every time it (OpenAI) is upgraded, some companies will 'die'. This time, some teams that do GPT real-time voice interaction can be announced to be disbanded." On the second day after the release of GPT-4o, a large model industry entrepreneur exclaimed to a reporter from Economic Daily News.
Shenyang, the director of the Metaverse Culture Laboratory at the School of Journalism at Tsinghua University, also tried GPT-4o the next day. In his video account, he mentioned that GPT-4o excels in details such as hair delicacy and light and shadow effects at the level of cultural images.
With the confrontation between the two sides on smart assistants, Shenyang believes that the current competitive landscape has become clearer. Google is further promoting its Gemini based AI assistant, and Apple has also reached an initial cooperation with OpenAI to install ChatGPT on Apple phones.
Shenyang stated that with this press conference as a turning point, ChatGPT has been transformed into a soulmate, which is actually Siri. Therefore, the industry landscape has become clear, and Apple is using the built-in ChatGPT to compete with Google Gemini's mobile assistant. Meta will also launch a mobile assistant based on Llama. For the industry, AI assistants are expected to move from a user base of 100 million to a user base of 1 billion.
"When GPT-4o was released yesterday, it felt very powerful, but today I think Google's latest release has completely caught up with its achievements. I think OpenAI should be even more nervous in the future because application companies and super large platform companies have all caught up, and its advantages are becoming less and less." Li Mingshun, founder of Shunfu Capital and chairman of Xingxing AI, told a reporter from Daily Economic News that currently, the user growth of OpenAI is not very obvious, and technology leadership and cost advantages may not be the best. At the same time, the era of strong applications is coming faster. In this context, Google has combined all of its applications with large models to form stronger user stickiness and dependence., or even better.
In Li Mingshun's view, in the next stage, platform application companies in the United States, including Microsoft, Apple, Dell, as well as Tencent, ByteDance, and Alibaba in China, will gradually combine their own application and big model capabilities to launch super applications, gradually entering the era of comprehensive competition. It will be even more difficult to rely on a big model to make a name for oneself.
Google's Anti Encirclement Campaign Against OpenAI
Eating the cake of giants is not so easy.
Prior to the press conference, the market was filled with smoke bombs surrounding OpenAI's search layout, with reports suggesting that OpenAI is likely to launch a new search engine based on ChatGPT technology. Meanwhile, a webpage called "GPTSearch" has been launched, but currently only members can access it. Renowned journalist Pete Huang also tweeted a teaser stating that GPTSearch will be officially launched on May 9th.
In the end, Google held its ground in this round of negotiations. Google CEO Sundar Pichai mentioned in his speech that one of the most exciting changes brought by Gemini is in Google search, where "one of our biggest investment and innovation areas is our founding product - search.".
From the press conference, it appears that Google has taken the lead in combining AI capabilities with its search engine. Google announced that the "AI Overviews" feature, which can summarize Google search engine results, will be launched in the United States this week. In this feature, Google will display AI generated answers to users.
According to Google, AI Overview is designed to respond to complex searches and help users seek solutions. For example, when people search for vegetarian preparation or travel plans, the answers provided by AI will appear at the top of the search page.
Google has also enhanced the visual functionality of search, supporting asking questions through videos. At the I/O conference, Google demonstrated that when faced with a record player malfunction, users can shoot videos and ask questions at the same time, and obtain an AI overview including repair steps and resources through new searches.
Despite being the first to target Google's new products at the level of intelligent assistants, this revolutionary feature, which has been highly anticipated since the release of GPT-3.5, has yet to take further action after a series of smoke bombs. In the search market, Google's fundamentals remain stable, while its comprehensive AI capabilities are forming a looming encirclement.
On the other hand, for OpenAI, there are still internal concerns about competing with giants for territory.
After the aftermath of last year's infighting, just one day after the release of GPT-4o, Ilya Sutskever, co-founder and chief scientist of OpenAI, who had disappeared from the public eye for a long time, officially announced her departure from OpenAI. Last November, there was turbulence in the management of OpenAI, and Sutskever is considered the driving force behind this controversy. Not long ago, one of the founding members of OpenAI, AndrejKarpathy, also resigned before the release of Sora.
In this new technological revolution, as the focus of competition shifts from large model technology to more responsible application side, OpenAI, which once led the way with a dark horse stance, has begun to slow down, and a new turning point may have emerged.