Last week, on the GPT-5 launch day, AI Hype was at an all-time high.
In already a press briefing, Openai CEO Sam Altman said GPT-5 is “something that I want to return yet,” a milestone for the first iPhone with a retina display. Night before announcement, Altman Posted An image of the death star, and even more publicity. On X, a user wrote This expectation “seems like the eve of Christmas.” Everyone’s eyes were on the chatter-makers because people in industries were waiting to see if the publicity would be distributed or disappointed. And from most accounts, the larger revelations will be reduced.
The promotional for the long-term new models of Openai was built over the years-since the 2023 release of GPT-4. In the Reddit Ama with Altman and Employees last October, users continuously asked about the release date of GPT-5, looking for details about its characteristics and will separate it. A reditor asked, “Why is GPT -5 so long?” Altman replied that the calculation was a limit, and “all these models have become quite complicated and we cannot ship as much things as we want.”
But when the GPT-5 appeared in the Chatgpt, the users were largely unaffected. The major progress they were expecting seemed to mostly look old, and the major benefits of the model were in areas such as costs and speeds. In the long run, however, it can be a solid financial condition for openi – although a less attractive.
People expected the world of GPT-5. (An X user Posted After Altman’s Death Star Post, “All transferred expectations.”) And Openai did not reduce those estimates, Calling GPT-5 Its “best AI system yet” and “significant jump in intelligence” with sophisticated performance during coding, mathematics, writing, health, visual perception, and more. Altman said in a press briefing that chatting with the model “wants to talk to a PhD-level specialist.”
This promotion is designed for a contrast with reality. For example, PhD-level intelligence will be a model with information Repeat The word Blueberry had three “B”, as some social media users found? And it will be Should not be able to identify How many state names consisted of the letter “R”? would it be Wrong label “New Jeffst,” “Mican,” “New Nakamia,” “Crizona,” and “Miroinia,” and an American map with states made as Label Nevada as the expansion of California? People using bots for emotional support found the new system as far and distant, so loudly opposed that Openai brought support to an old model. Mems Abed – Ek to depict GPT-4 and GPT-4O as a covalent dragon with GPT-5 as a simulton.
The court of expert public opinion was either not forgiving. Gary Marcus, a major AI industry voice and Emeritus Professor of Psychology at the University of New York, The model is called “Excessive, overHipped and weak.” Peter Wildford, Co-Founder of Institute for AI Policy and Strategy, wrote In his review, “Is it a huge smash that we were looking for? Unfortunately, no.” ZVI Mowshowitz, a popular AI industry blogger, Called it “A good, but not great, model.” A reditor on official GPT-5 Reddit Ama wrote“Someone told that Sam is 5 hot garbage.”
In the days after the release of GPT-5, the attack of uncontrolled reviews is slightly a bit. The consensus is that although GPT-5 was not as important as an advancement as people were expecting, it offered upgradation in cost and speed, as well as low hallucinations, and switch systems that it was introduced-to direct their queries on the backynd for the latest that models that are the most understandable to answer it, so you do not need to decide. Altman bowed into that story, Write“GPT -5 is the clever model we have ever done, but the main thing we have pushed is the real world utility and large -scale access/strength.”
Openai researcher Christina Kim Posted On the x that with the GPT-5, “the real story is utility. What about the people cares about the appearance, creative writing, and navigating health information-with more stability and low friction. We also cut the hallucinations. It is better calibrated, ‘I don’t know,’ separate the facts, ‘and when you want you can give up the answer when you want.”
There is a comprehensive understanding that, to keep it clearly, GPT-5 has reduced the chatgpt. The viral social media posts complained that the new model lacked the nuances and depth in its writing, which was coming in the form of robots and colds. Even in its marketing material of GPT-5, the GPT-4o and GPT-5-Generated wedding toasts’ side-by-side comparison looks like an uncontrolled win for new models-I individually prioritize one from 4o. When Altman Asked by redditors If they feel that GPT-5 was better in writing, it was raised with an attack of comments defending the retired GPT-4O model instead; Within a day, he received for pressure and at least temporarily returned it to the chatter.
But there is a front where the model appears shining: coding. A recurrence of GPT-5 Currently top The most popular AI model leaderboard in the coding category is coming in another with the cloud of anthropic. Openai’s launch promotion showed the AI-Janit game (a rolling ball mini-game and a typing speed race), a pixel art tool, a drum simulator and a lofty visualizer. When I tried to vibi a puzzle game with the tool, it had a bunch of glits, but I was successful with simple projects such as an interactive embroidery text.
This is a major victory for Openai, as it has been head-to-head with contestants such as anthropic, Google and others in AI coding wars. Businesses are ready to spend a lot on AI coding, and it is one of the most realistic revenue generators for cash-burning AI startups.
Openai also highlighted the skills of GPT-5 in healthcare, but it is mostly unused in practice-we will probably not know how successful it is for a while.
AI benchmarks have come after low and low in recent years, as they often change and some companies are as a result of cherry-picks, resulting in they appear. But overall, they can give us an appropriate picture of GPT-5. The model performed better than its predecessors on many industry tests, but according to many industry people, it was nothing to write about the improvement house. As wildford Keep it“When it comes to this formal evaluation, it seems that GPT -5 will be highly expected -a vague Death Star Meme worthy of anything worthy of anything, small, incremental growth.”
But if there is anything to say anything about it in recent history, those small, incremental growth may be more likely to translate into solid benefits than individual consumers. AI companies know that their biggest money routes are enterprise clients, government contracts and investment, and inkrymels proceed to the solid benchmark, as well as to reduce coding and invest in hallucinations fighting and investing in hallucinations, the best way to get out of the three.
