Every reporter: Wen Qiao Every editor: Lan Suying
Starting tonight, openai online live broadcast, Google 2024 i/o developer conference, Microsoft build annual developer conference, NVIDIA first quarter report and Apple annual developer conference It will all be staged in the coming month.
At 10 a.m. local time on May 13 (1 a.m. on May 14, Beijing time), the openai online live broadcast will be the first to show off the new features of chatgpt and the upgraded version of gpt-4. The company has released a reservation page on the official website. However, relevant details have not yet been announced.
Image source: openai
the information quoted sources as saying that openai is preparing to demonstrate technology that can use sound and text to talk to people and recognize objects and images at the event. This technology will be integrated into the much-anticipated language assistant product.
In addition, a reporter from "Daily Economic News" noticed on the x platform that some netizens picked up the code of chatgpt, indicating that it may have voice call capabilities. Some people joked that the scenes in the movie "Her" are about to become reality.
Foreign media had expected that this live broadcast would release news about search engines or gpt-5, but then Openai CEO Sam Altman refuted these rumors. As competitors such as Google and Meta have successively released new multi-modal large models, foreign media reported that OpenAI may announce GPT-5 before the end of this year in order to maintain its leading position.
openaiās online live broadcast has two highlights worthy of attention
Although openaiās official website did not provide relevant details about this live broadcast demonstration, the reporter found after combing that there are two highlights worth looking forward to in this event.
Highlight 1: Voice Assistant
According to an exclusive report by The Information, OpenAI is preparing to demonstrate technology that can use voice and text to talk to people and recognize objects and images at this upcoming event, and integrate this technology into a voice assistant product middle. This means that the new multi-modal model on which this AI assistant is based can better understand images and audio. The
report quoted people familiar with the matter as saying that the company's developers have demonstrated some of the functions of this AI assistant to some customers, and its logical reasoning capabilities are better than current products.
reported that OpenAI CEO Sam Altman's goal is to eventually develop a virtual assistant similar to Spike Jonze's movie "Her" to make existing voice assistants such as Apple's Siri more intelligent. Altman believes that AI assistants with visual and audio capabilities have the same transformative potential as smartphones. In theory, AI assistants can serve as student mentors, solve car troubles, and many other roles.
However, the report said that the product does not currently run on personal devices, and customers will be able to use the cloud-based version in the short term to deploy or improve features, such as automated customer service agents. It's unclear when OpenAI will make these new features available to its paying customers, but it eventually plans to make them part of the free version of its chatbot ChatGPT.
Highlight 2: chatgpt will have a call function
The reporter noticed that on the x platform, a developer named ananay arora released a screenshot of suspected chatgpt call-related code. He said that openai may be preparing to build call functionality into chatgpt.
Image source: x platform
Additionally, Arora found evidence that OpenAI has configured servers for real-time audio and video communications.
Image source: x platform
Based on this, some netizens commented and speculated, "Does this mean chatgpt can take the initiative to call me without me making a call first?" This scene was seen in the movie "Her" Appear - AI assistant Samantha will take the initiative to call the hero.
Image source: x platform
In addition, hallid.ai co-founder indigo predicted on the x platform that the new version of gpt-4 should be divided into multiple versions according to different parameter scales. He also said: "gpt The historical mission of -3.5 is over, the lightweight version of gpt-4 can be used for free in chatgpt, and the price of the API will drop significantly. "
Image source: x platform
indigo's guess is consistent with the goal of openai.According to the information report, OpenAIās goal is to make the AI āāmodel that supports the aforementioned new features cheaper to run than the most advanced model currently sold, GPT-4 Turbo.
It is reported that OpenAI also plans to release a new pricing model this year, which will provide customers with discounts of up to 50% if they prepay. Currently, OpenAI mainly provides on-demand pricing, charging developers a fee ranging from a few cents to $120 for every 1 million tokens generated by a large model.
It is reported that early payment discounts are common in cloud computing - customers of Microsoft Azure, Google Cloud and Amazon Web Services can reduce costs by reserving server capacity in advance. With more flexible pricing, OpenAI can better compete with competitors.
gpt-5 may be announced before the end of the year, openai still faces challenges in profitability
As Google, meta and other competitors have successively released new multi-modal large models, for example, meta released the open source model llama 3 in April, and Google released gemini in December last year , there is speculation that openai will soon announce gpt-5 in order to maintain its leading position. In addition, a series of recent reports stated that OpenAI is also working on launching an online search engine, aiming to compete with Google Search for market share.
However, in a previous interview with foreign media, altman stated that openai will not release gpt-5 or an ai search engine during this live broadcast.
Still, the company hasn't denied rumors that it's developing an online search product. The technology media Inquirer Tech reported on a domain name and SSL certification called "search.chatgpt.com", saying that this is a strong clue related to the OpenAI online search project. In addition, it is reported that the company may announce GPT-5 before the end of this year.
Although OpenAI invests a lot of energy in AI models, the information also expressed concerns about how the company will create real benefits next. For example, OpenAI has promised developers that by the first quarter of this year, they will be able to make money building custom chatbots through GPTs, but so far the company has not launched an effective way to achieve this goal.
On the other hand, foreign media reported that Apple is close to reaching an agreement with openai on how to integrate chatgpt into the next generation iPhone operating system. However, openai still faces fierce competition: Apple is also in similar negotiations with Google in order to apply gemini to smartphones.
In addition, Altman is also working with former Apple designer Jony Ive to develop a separate AI consumer device. However, the sheer scale of state-of-the-art AI models currently means they need to be run in the cloud and require an Internet connection to work. AI models with visual and audio capabilities may take months or even years to become small enough to be used in run on the device.
Daily Economic News
