‘We totally screwed up': Sam Altman explains what went wrong with GPT-5 launch
The ChatGPT maker soon realized its mistake and began providing GPT-4o — the model that previously powered the chatbot — to ChatGPT Plus users, while also offering additional rate limits for GPT-5 Standard and Thinking models.
OpenAI had claimed that its new model was leaps and bounds ahead of previous offerings in areas like coding, reasoning, accuracy, health, and multimodal abilities. However, users complained that the new model gave shorter answers and showed less emotional depth in its responses compared to earlier models.
In the immediate aftermath of the model launch, OpenAI promised to make the new model 'warmer' while also implementing a host of other changes.
Sam Altman on GPT-5's patchy rollout:
Altman, who had been teasing the GPT-5 rollout as the next big leap in AI for months, if not years, appeared to have been surprised by the response the model received from users.
In a recent discussion with reporters on Thursday (as quoted by Bloomberg), the OpenAI chief executive admitted, 'I think we totally screwed up some things on the rollout.'
'We've learned a lesson about what it means to upgrade a product for hundreds of millions of people in one day, and the differences in the kinds of attachment people have with this product versus previous products,' Altman added.
Altman went on to state that ChatGPT API traffic had doubled in the 48 hours following the new model launch and that ChatGPT app usage was at a 'complete high' in the days after the GPT-5 release. He reportedly also agreed that it was the wrong decision to deprecate all the older models.
Last week, OpenAI announced that it had reached 700 million weekly users — a figure the company would have hoped to grow further with the launch of its new models. During its previous major GPT model launch earlier in the year, OpenAI had introduced native image generation capabilities to GPT-4o, which sparked a Studio Ghibli–style craze worldwide and led to a major spike in ChatGPT usage.
That same craze was nowhere to be seen this time around, as social media quickly filled with criticism of the approach taken with the GPT-5 launch. The company even had to restore the model picker after user backlash — notably, removing the model picker entirely had been one of Altman's key promises for GPT-5 in the months leading up to the launch.
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Mint
24 minutes ago
- Mint
ChatGPT boss warns against relying on AI as primary source of information: Here's why
OpenAI released its latest language model, GPT-5, last week, which the company says is significantly better than previous models in areas like accuracy, reasoning, health, and multimodal abilities. However, Nick Turley, Head of ChatGPT, believes the AI chatbot should still be used as a second opinion and not as a primary source of information. In a recent conversation with The Verge, Turley was asked about the continuing problem of hallucination (AI making things up) with GPT-5. While OpenAI claims to have made huge strides in reducing hallucinations, the new model is still prone to errors roughly 10% of the time. Speaking about the persistent challenge, Turley said, 'The thing, though, with reliability is that there's a strong discontinuity between very reliable and 100 percent reliable, in terms of the way that you conceive of the product.' 'Until I think we are provably more reliable than a human expert on all domains, not just some domains, I think we're going to continue to advise you to double check your answer. I think people are going to continue to leverage ChatGPT as a second opinion, versus necessarily their primary source of fact,' he added. Notably, large language models are trained to predict the next likely word based on patterns in their training datasets. This means that when asked about information outside their training data, LLMs often produce plausible-sounding but false answers. Turley acknowledged the problem but stressed that ChatGPT works best when paired with external, verifiable sources like a search engine. 'I still believe that, no question, the right product is LLMs connected to ground truth, and that's why we brought search to ChatGPT and I think that makes a huge difference.' The OpenAI executive further expressed optimism about solving hallucinations: 'I'm confident we'll eventually solve hallucinations, and I'm confident we're not going to do it in the next quarter.' Notably, OpenAI is also reportedly working on developing its browser and the company's CEO Sam Altman recently said in an interview that it is interested in buying Google's Chrome browser if Google is forced to sell its prized possession.
Indian Express
an hour ago
- Indian Express
CEOs want their companies to adopt AI. But do they get it themselves?
In March, Andy Katz-Mayfield, co-founder of the razor brand Harry's, started inviting junior employees to monthly meetings usually reserved for his most senior leaders. The purpose was for lower-level workers to show off how they were using generative artificial intelligence to improve the supply chain, finance and marketing. But Katz-Mayfield had another purpose, too: getting the top executives comfortable with using AI themselves. 'Building familiarity with these tools opens people's eyes,' said Katz-Mayfield, who is also a CEO of Harry's parent company, Mammoth Brands. 'Through demos and stuff, people are like: 'Oh, that's cool. I didn't think about that, but I now realize why this is important for my team.'' Executives refer to the promise of AI with grandiose comparisons: the dawn of the internet, the Industrial Revolution, Carl Friedrich Gauss' discovery of number theory. But while boards and top executives may mandate using AI to make their businesses more efficient and competitive, many of those leaders haven't fully integrated it into their own workdays. As with most technological advances, younger people have taken to AI more quickly than their elders. And the work that people do earlier in their careers — inserting data into spreadsheets, creating decks, coming up with designs — also lends itself to playing around with the technology. Top executives, on the other hand, are often several steps removed from the mechanics. Once they're in the C-suite, days are filled with meetings. Less doing, more approving. So to nudge high-level managers, CEOs who have fully embraced AI are trying new tactics. Some have told senior leaders to use Gemini, Google's AI assistant, before defaulting to Google search. Some are carving out time at corporate retreats to play around with generative AI tools like Creatify. At Mayer Brown, a law firm in Chicago, chair Jon Van Gorp has shared with the partners how he uses a generative AI tool built for legal professionals to help draft contracts and distill the most salient points from his own writing. At a fashion startup called Daydream, Friday lunches are devoted to employees' sharing how they're using generative AI tools; the chief technology officer has shared her Gemini prompts from the week. Mammoth's chief technology officer, Sandeep Chouksey, 41, is well aware of AI and has been playing around with ChatGPT since it came out nearly three years ago. But he found that watching the engineers on his team helped him understand the technology better. He figured his peers needed to get their eyes on it, too, and suggested inviting employees who were working closely with AI to the leadership meetings. The work of senior executives 'doesn't lend itself to actually experimenting with the technology,' Chouksey said. 'I knew that the other leaders needed to see what I was seeing — all the bottom-up work that was happening.' Chuck Whitten is witnessing how company executives are gradually wrapping their heads around the AI phenomenon. He is the global head of digital practices at Bain & Co., a management consulting firm where his job is to advise CEOs about technology. They understand the importance of integrating AI into their companies, he said, but don't yet have a feel for the technology itself. He was in their shoes not too long ago. In 2021, he left Bain after 22 years to become co-chief operating officer at Dell Technologies. He was in that job when ChatGPT rolled out. He describes it as a 'lightning bolt' moment. Part of the reason he returned to Bain was realizing that senior leaders needed assistance entering the 'golden age of artificial intelligence,' he said. 'I think the majority that I see are just experimenting with the basics, sort of trying Copilot or ChatGPT for the occasional email, draft or quick fact check,' Whitten said. 'This is not a tool you can delegate down the hall to the chief information officer. They need to be hands-on in both where the technology is going and how they can apply it today.' According to a survey of 456 CEOs by Gartner, a research and advisory firm, released in May, 77% of the executives thought AI is transformative for business, but fewer than half thought their technology officers were up to the task of navigating the current digital landscape. Every CEO is trying to 'figure out whether they're set up for the future or not and how the world looks on the other side of this technology transformation,' said Tom Pickett, CEO of Headspace, a wellness app. 'They're facing this constant change, which just leads to stress and everyday anxiety.' Pickett, 56, has dealt with his own anxiety by using AI chatbots as much as possible. He joined the company last August and said chatbots had helped him get up to speed in his role. He uses ChatGPT or Gemini to do research and receive advice about business moves, such as potential partnerships with other companies. He said it helped him 'learn 10 times as much or test 10 times as many ideas in a very lightweight way.' In the past, he said, 'I would have had to ask the resident expert or somebody who worked with that company to really give me a debrief,' Pickett said. 'And instead, in five minutes, I'm like, 'Oh, OK, I get this.'' (He said he had also consulted people in his company, but now 'the conversations are more productive.') Sarah Franklin, CEO of Lattice, a human resources software platform, said it can be difficult to get executives to use new tools, and in internal meetings she regularly asks, 'Did you test that message with ChatGPT?' Franklin, who previously was chief marketing officer at Salesforce, has been using generative AI tools since they came on the market. But the technology is moving quickly, and everyone is trying to figure it out on the go. 'Nobody has 10 years of agentic AI experience right now. They at best have six months. So nobody is fully prepared,' Franklin, 49, said. 'What we have right now in the world is a lot of optimism combined with a lot of FOMO.' Fear of missing out can be the mother of innovation, it seems. In January, Greg Schwartz, CEO of StockX, was scrolling the social platform X when he saw several users posting projects that they had made with various AI coding apps. He downloaded the apps. He hadn't written a line of code in years. But using the apps got his mind racing. During a corporate retreat in March, he decided to push 10 senior leaders to play around with these tools, too. He gave everyone in the room, including the heads of supply chain, marketing and customer service, 30 minutes to build a website with the tool Replit and make a marketing video with the app Creatify. 'I'm just a tinkerer by trait,' Schwartz, 44, said. 'I thought that was going to be more engaging and more impactful than me standing in front of the room.' There was a 'little bit of shock' when he presented the exercise, he said. But he tried to remind people it was a fun activity. They weren't being graded. Their discomfort is normal, said Ethan Mollick, a professor at the University of Pennsylvania's Wharton School and author of the newsletter One Useful Thing and the book 'Co-Intelligence: Living and Working With AI.' 'AI is weird and off-putting,' Mollick said. 'There's a lot of psychological resistance to using the systems even for people who know they should be doing it.' Many organizations, he added, have a 'real failure of imagination and vision' when it comes to the power of these systems. 'The main issue is that leaders have to take a leading role,' Mollick said. 'They all say AI is the future, use AI to do stuff. And then they don't make any decisions or choices.' About half of companies do not have road maps for integrating AI, according to a Bain survey. Whitten at Bain said that about only 20% of companies were scaling their AI bets and that most didn't have benchmarks for how workers should use AI. At Mammoth Brands, Katz-Mayfield said that he and his team had discussed providing incentives to employees who use AI but that they hadn't needed to. The energy around experimenting is working for the company. In the last meeting it had five demos on the docket but didn't get to all of them because senior leaders were 'asking so many questions and wanting to see different things.' 'If the leadership team is excited and engaged in that stuff,' Katz-Mayfield said, 'that's probably more than half the battle.'
United News of India
an hour ago
- United News of India
Cohere achieves USD 6.8 bln valuation after successful investment rounds
Business Economy New Delhi, Aug 16 (UNI) Cohere, a security-first AI enterprise company, achieved a USD 6.8 billion valuation after a successful investment round. Reportedly, Radical Ventures and Inovia Capital also participated in this funding round, besides AMD, Nvidia, and Salesforce Ventures. Canadian company Cohere distinguishes itself in this heated marketplace due to its unique operational approach. It's an AI company that builds and provides access to large language models (LLMs) and related AI tools. These tools mainly focus on offering that kind of AI solutions to companies which will improve operations, automate tasks, enhance customer service, and ultimately drive sales for the business. Cohere leverages Transformer-based Large Language Models (LLMs), which allow a range of generative AI and Natural Language Processing (NLP) applications, including summarization, content creation, search, and chatbots. Moreover, the AI market is currently going through intense competition due to the launch of GPT-5 by OpenAI. But currently, multiple reports indicate that users are not impressed by the operational capabilities of GPT-5 which can be a major opportunity for Cohere to dominate. UNI SAS GNK
