Google and OpenAI are paying attention to their artificial intelligence models

Two videos that have quickly gone viral on social media summarize the latest advances in artificial intelligence announced this week by Google and OpenAI, two tech companies vying for leadership in the field. In one you can seeWatch Mira Murati, CTO of OpenAI, speak Italian while her chatbot ChatGPT It is capable of simultaneously translating your words into English.

In another case, an employee at Google’s London office asks: Twins interpret a few lines of code on your computer screen or help find lost glasses while you point the camera around the room. In both cases, AI is able to solve problems with surprising ease.

If until now you had to write for a long time to communicate with AI tips in the text field, latest versions of models allow Interact with your digital assistants more like a human. “We are shifting the paradigm towards more natural and simpler human-machine interactions,” Murati said Monday during the presentation of GPT-4o, an improved version of his model that can interpret voice, video, images and code in a single interface.

The model was trained to provide flexible, real-time communication. Until now, you could talk to ChatGPT using your voice, but the 5.4 to 2.8 second response delay made the process too slow and tedious.

Alphabet answer, matrix Googlewhich yesterday held its annual developer conference – that’s what they agreed to call Project Astra. A new multimodal AI assistant that accepts inputs such as voice, video or text, based on an improved version of the Gemini model. The goal of Astra, a prototype still in development, is to create an intelligent assistant that freely interacts with people in real timewho can be spoken to naturally, who has no delay in responding, and who is capable of reasoning through text, sound, and images.

“While we have made incredible progress in developing artificial intelligence systems that can understand multimodal information, reducing response times to conversational levels is an engineering challenge,” said Demis Hassabis, head of Google’s artificial intelligence research unit DeepMind. The group plans to start adding Astra’s capabilities to its Gemini chatbot and the rest of its products later this year.

Though the biggest announcement of all that happened this week was that after months of testing, Google will finally bring generative artificial intelligence to the heart of its business: its search engine. Starting this week, users in the US and UK (Europe’s expected arrival date is yet to be announced) will see a short search when performing a search. a summary of your request in the AI-generated page header., followed by a series of clickable links. The assistant, which will reach 1 billion users worldwide this year, will also let you create personalized plans and, in the future, offer the ability to search using video.

The company took the step to expand the reach of generative search after a controlled trial of these capabilities to see what impact they would have on its business. ““People visit a wide variety of websites to get answers to more complex questions.”said Liz Reed, Google’s vice president of search.

The technology company already addressed these issues in its latest conference call with analysts following its results presentation. The company noted that during the testing period they saw that the use AI-generated summaries did not have a negative impact on search engine traffic.. It also states that they have been able to reduce the costs associated with generative AI responses, which require more computing power, by 80%.

The search engine changes come as OpenAI and Microsoft, which has incorporated generative artificial intelligence Copilot into its Bing search engine, threaten the foundation on which Google’s business is built. In the last quarter Search advertising brought in $46.156 million., after growing by 14%. This is equivalent to 57% of the tech company’s total revenue.