What was already busy in the past few days New model versionsOpenai reaches a week with a Search inspection from GPT-4.5. The company promotes the new system as the largest and best chat model so far. In the early test, Openai says that people found GPT-4.5 to be a more natural conversation, with the ability to transfer warmth and display a kind of emotional intelligence.
In one of the examples of Openai, a person tells Chatgpt that he is going through a difficult time after the test failure. Where the previous models of the company, including GPT-4O and O3-MiniYou may communicate with the individual before providing a long list of unwanted tips, GPT-4.5 takes different ingenuity. "Do you want to talk about what happened, or do you just need distraction? I am here in both cases," Chatbot says when running by GPT-4.5.
GPT-4.5 gains are the result of the progress made by Openai in the non-supervisory learning. Through unprecedented learning, automatic learning algorithm is given an unheard and left for its own devices to find patterns and visions. GPT-4.5 No "He thinks" Like modern thinking models for the company, but in training the new model, Openai has improved architectural improvements and gave them access to more data and energy calculation. "The result is a model that has a broader knowledge and a deeper understanding of the world, which leads to a decrease in hallucinations," The company says.
Talking about low hallucinations, Openai measured the best GPT-4.5 in this regard. Upon placed SimpleqaA standard designed on Openai that tests large language models on its ability to answer "Clear but difficult but difficult knowledge questions," GPT-4.5 beat O3-MINI, GPT-4O and until O1 with a hallucinatory rate of 37.1 percent. It is clear that the new model does not completely solve the problem of hallucinations, artificial intelligence, but it is a step in the right direction.
Despite its relative strengths on the GPT-4O and O3-MINI, GPT-4.5 is not a direct alternative to these models. Compared to Openai’s thinking systems, GPT-4.5 is "More important, innately smarter model." In addition, it is not a multimedia like GPT-4O, which means that it does not work with features like SoundVideo or screen shot. It’s too "Very large and intense model."
It is better to think of GPT-4.5 as a starting stone for the systems that Openai plans to present in the future. In fact, Sam said earlier this month when it was Share the company’s road mapNoting that GPT-4.5 will be "Our last unprecedented model" – Referring to the fact that the new system does not solve problems by addressing them step by step like thinking models in Openai. Its successor, GPT-5, may integrate many of the latest technologies in Openai, including Border Model O3. Openai repeated that day, saying that he is planning to bring GPT-4.5 "Unique strengths, including broader knowledge, strongest intuition, and “eq”, for all users in future models."
Meanwhile, Chatgpt Pro Subscribers can start using the GPT-4.5 starting today, as the Pro and Team users are scheduled to arrive from next week.
This article was originally appeared on Engadget on https://www.engadget.com/ai/openais-new-gpt-45-model-is-a-better-more-nature-conversionalist-2035185.html?
[og_img]
Source link