Everything you need to know about the latest ChatGPT update
OpenAI unveiled a new flagship artificial intelligence (AI) model, GPT-5, yesterday.
The upgrade is available to all 700million users for free, though paid subscribers have fewer usage limits.
'Our smartest, fastest, most useful model yet, with built-in thinking that puts expert-level intelligence in everyone's hands,' the start-up said.
The virtual assistant is now powered by a so-called reasoning model, which spends more time 'thinking' through a problem.
OpenAI says the bot is far faster than before, is better at coding and is now more customizable.
Users can change the chat colours and choose from preset 'personalities', like 'cynic', 'listener', and 'robot'.
Voice interactions have also been improved, and pro users will will soon be able to connect Gmail, Calendar and Contacts directly to ChatGPT.
To show off what this means, OpenAi compared GPT-5 to its previous model, GBT-4o, when it comes to writing poetry or a wedding toast.
It can also build simple software apps, such as video games or pixel art-makers, from short text prompts.
Health experts and officials have long cautioned people against using ChatGPT to ask medical questions, with the bot often citing fake scientific sources and giving shoddy diagnoses.
OpenAI claims the new bot is its 'best model yet for health-related questions', though it stressed it does not replace health professionals.
CEO Sam Altman said: 'GPT-5 is the first time that it really feels like talking to an expert in any topic, like a PhD-level expert.'
Chatbots are increasingly used by people for everyday tasks, like writing out emails they meant to send two weeks ago or asking for advice.
But this technology, called generative AI, has never been the best at ensuring that the information it provides is true.
It happens so frequently that researchers have a word for it – hallucinating.
The chatbots are called large language models (LLM) that learn by analysing texts on the internet to see how humans put words together.
But what it can't learn is how to tell if something is true or not – as they try to guess the next word in a sequence of words, they sometimes generate phoney information as their patterns get muddled up with fake news.
So OpenAI engineers carried out two tests to see how often ChatGPT hallucinated.
The first, called Long Fact, involved asking the bot to answer questions about concepts or objects. They then asked it to make biographies of public figures, an examination called FActScore.
OpenAI said that for both, GPT-5 makes about 45% fewer factual errors than its predecessor, GPT-40, and makes things up six times less than o3.
A system card, which describes the algorithm's capabilities, says: 'We've taken steps to reduce GPT-5-thinking's propensity to deceive, cheat, or hack problems, though our mitigations are not perfect and more research is needed.
'In particular, we've trained the model to fail gracefully when posed with tasks that it cannot solve.'
While OpenAI said this is a 'major leap' for the system, Michael Rovatsos, an AI professor at the University of Edinburgh, wasn't too impressed.
He said: 'While it's too soon to tell, it sounds like no significant progress has been made on the core AI model, but instead, OpenAI is focusing on making it more useful by better controlling how it behaves through additional 'wrapper' technology. More Trending
'As an analogy, this would mean that what matters is not whether you have the best nuclear reactor, but whether you can actually build a safe and efficient power plant around it.'
Edoardo Ponti, an assistant professor in Natural Language Processing at the Scottish university, said the upgrade is 'far from dramatic'.
'The presentation was partly weakened by flaws in the result reports and a hallucinated demo,' Ponti said.
'Moreover, it left a bit unclear where GPT-5 stands with respect to models from OpenAI's competitors.'
Get in touch with our news team by emailing us at webnews@metro.co.uk.
For more stories like this, check our news page.
MORE: Creepy AI pics of Paul McCartney visiting Phil Collins in hospital go viral after star issued health update
MORE: We chatted with the UK's first AI MP and it said something very unexpected
MORE: Sir Rod Stewart's 'tacky' AI tribute comes far too soon after Ozzy Osbourne's death
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
The Guardian
an hour ago
- The Guardian
OpenAI will not disclose GPT-5's energy use. It could be higher than past models
In mid-2023, if a user asked OpenAI's ChatGPT for a recipe for artichoke pasta or instructions on how to make a ritual offering to the ancient Canaanite deity Moloch, its response might have taken – very roughly – 2 watt-hours, or about as much electricity as an incandescent bulb consumes in 2 minutes. OpenAI released a model on Thursday that will underpin the popular chatbot – GPT-5. Ask that version of the AI for an artichoke recipe, and the same amount of pasta-related text could take several times – even 20 times – that amount of energy, experts say. As it rolled out GPT-5, the company highlighted the model's breakthrough capabilities: its ability to create websites, answer PhD-level science questions, and reason through difficult problems. But experts who have spent the past years working to benchmark the energy and resource usage of AI models say those new powers come at a cost: a response from GPT-5 may take a significantly larger amount of energy than a response from previous versions of ChatGPT. OpenAI, like most of its competitors, has released no official information on the power usage of its models since GPT-3, which came out in 2020. Sam Altman, its CEO, tossed out some numbers on ChatGPT's resource consumption on his blog this June. However, these figures, 0.34 watt-hours and 0.000085 gallons of water per query, do not refer to a specific model and have no supporting documentation. 'A more complex model like GPT-5 consumes more power both during training and during inference. It's also targeted at long thinking … I can safely say that it's going to consume a lot more power than GPT-4,' said Rakesh Kumar, a professor at the University of Illinois, currently working on the energy consumption of computation and AI models. The day GPT-5 was released, researchers at the University of Rhode Island's AI lab found that the model can use up to 40 watt-hours of electricity to generate a medium-length response of about 1,000 tokens, which are the building blocks of text for an AI model and are approximately equivalent to words. A dashboard they put up on Friday indicates GPT-5's average energy consumption for a medium-length response is just over 18 watt-hours, a figure that is higher than all other models they benchmark except for OpenAI's o3 reasoning model, released in April, and R1, made by the Chinese AI firm Deepseek. This is 'significantly more energy than GPT-4o', the previous model from OpenAI, said Nidhal Jegham, a researcher in the group. Eighteen watt-hours would correspond to burning that incandescent bulb for 18 minutes. Given recent reports that ChatGPT handles 2.5bn requests a day, the total consumption of GPT-5 could reach the daily electricity demand of 1.5m US homes. As large as these numbers are, researchers in the field say they align with their broad expectations for GPT-5's energy consumption, given that GPT-5 is believed to be several times larger than OpenAI's previous models. OpenAI has not released the parameter counts – which determine a model's size – for any of its models since GPT-3, which had 175bn parameters. A disclosure this summer from the French AI company Mistral finds a 'strong correlation' between a model's size and its energy consumption, based on Mistral's study of its in-house systems. 'Based on the model size, the amount of resources [used by GPT-5] should be orders of magnitude higher than that for GPT-3,' said Shaolei Ren, a professor at the University of California, Riverside who studies the resource footprint of AI. GPT-4 was widely believed to be 10 times the size of GPT-3. Jegham, Kumar, Ren and others say that GPT-5 is likely to be significantly larger than GPT-4. Leading AI companies like OpenAI believe that extremely large models may be necessary to achieve AGI, that is, an AI system capable of doing humans' jobs. Altman has argued strongly for this view, writing in February: 'It appears that you can spend arbitrary amounts of money and get continuous and predictable gains,' though he said GPT-5 did not surpass human intelligence. Sign up to TechScape A weekly dive in to how technology is shaping our lives after newsletter promotion In its benchmarking study in July, which looked at the power consumption, water usage and carbon emissions for Mistral's Le Chat bot, the startup found a one-to-one relationship between a model's size and its resource consumption, writing: 'A model 10 times bigger will generate impacts one order of magnitude larger than a smaller model for the same amount of generated tokens.' Jegham, Kumar and Ren said that while GPT-5's scale is significant, there are probably other factors that will come into play in determining its resource consumption. GPT-5 is deployed on more efficient hardware than some previous models. GPT-5 appears to use a 'mixture-of-experts' architecture, which means that it is streamlined so that not all of its parameters are activated when responding to a query, a construction which will likely cut its energy consumption. On the other hand, GPT-5 is also a reasoning model, and works in video and images as well as text, which likely makes its energy footprint far greater than text-only operations, both Ren and Kumar say – especially as the reasoning mode means that the model will compute for a longer time before responding to a query. 'If you use the reasoning mode, the amount of resources you spend for getting the same answer will likely be several times higher, five to 10,' said Ren. In order to calculate an AI model's resource consumption, the group at the University of Rhode Island multiplied the average time that model takes to respond to a query – be it for a pasta recipe or an offering to Moloch – by the model's average power draw during its operation. Estimating a model's power draw was 'a lot of work', said Abdeltawab Hendawi, a professor of data science at the University of Rhode Island. The group struggled to find information on how different models are deployed within data centers. Their final paper contains estimates for which chips are used for a given model, and how different queries are parceled out between different chips in a datacenter. Altman's June blog post confirmed their findings. The figure he gave for ChatGPT's energy consumption per query, 0.34 watt-hours per query, closely matches what the group found for GPT-4o. Hendawi, Jegham and others in their group said that their findings underscored the need for more transparency from AI companies as they release ever-larger models. 'It's more critical than ever to address AI's true environmental cost,' said Marwan Abdelatti, a professor at URI. 'We call on OpenAI and other developers to use this moment to commit to full transparency by publicly disclosing GPT-5's environmental impact.'
The Guardian
4 hours ago
- The Guardian
‘It's missing something': AGI, superintelligence and a race for the future
A significant step forward but not a leap over the finish line. That was how Sam Altman, chief executive of OpenAI, described the latest upgrade to ChatGPT this week. The race Altman was referring to was artificial general intelligence (AGI), a theoretical state of AI where, by OpenAI's definition, a highly autonomous system is able to do a human's job. Describing the new GPT-5 model, which will power ChatGPT, as a 'significant step on the path to AGI', he nonetheless added a hefty caveat. '[It is] missing something quite important, many things quite important,' said Altman, such as the model's inability to 'continuously learn' even after its launch. In other words, these systems are impressive but they have yet to crack the autonomy that would allow them to do a full-time job. OpenAI's competitors, also flush with billions of dollars to lavish on the same goal, are straining for the tape too. Last month, Mark Zuckerberg, chief executive of Facebook parent Meta, said development of superintelligence – another theoretical state of AI where a system far exceeds human cognitive abilities – is 'now in sight'. Google's AI unit on Tuesday outlined its next step to AGI by announcing an unreleased model that trains AIs to interact with a convincing simulation of the real world, while Anthropic, another company making significant advances, announced an upgrade to its Claude Opus 4 model. So where does this leave the race to AGI and superintelligence? Benedict Evans, a tech analyst, says the race towards a theoretical state of AI is taking place against a backdrop of scientific uncertainty – despite the intellectual and financial investment in the quest. Describing AGI as a 'thought experiment as much as it is a technology', he says: 'We don't really have a theoretical model of why generative AI models work so well and what would have to happen for them to get to this state of AGI.' He adds: 'It's like saying 'we're building the Apollo programme but we don't actually know how gravity works or how far away the moon is, or how a rocket works, but if we keep on making the rocket bigger maybe we'll get there'. 'To use the term of the moment, it's very vibes based. All of these AI scientists are really just telling us what their personal vibes are on whether we'll reach this theoretical state – but they don't know. And that's what sensible experts say too.' However, Aaron Rosenberg, a partner at venture capital firm Radical Ventures – whose investments include leading AI firm Cohere – and former head of strategy and operations at Google's AI unit DeepMind, says a more limited definition of AGI could be achieved around the end of the decade. 'If you define AGI more narrowly as at least 80th percentile human-level performance in 80% of economically relevant digital tasks, then I think that's within reach in the next five years,' he says. Matt Murphy, a partner at VC firm Menlo Ventures, says the definition of AGI is a 'moving target'. He adds: 'I'd say the race will continue to play out for years to come and that definition will keep evolving and the bar being raised.' Even without AGI, the generative AI systems in circulation are making money. The New York Times reported this month that OpenAI's annual recurring revenue has reached $13bn (£10bn), up from $10bn earlier in the summer, and could pass $20bn by the year end. Meanwhile, OpenAI is reportedly in talks about a sale of shares held by current and former employees that would value it at about $500bn, exceeding the price tag for Elon Musk's SpaceX. Some experts view statements about superintelligent systems as creating unrealistic expectations, while distracting from more immediate concerns such as making sure that systems being deployed now are reliable, transparent and free of bias. 'The rush to claim 'superintelligence' among the major tech companies reflects more about competitive positioning than actual technical breakthroughs,' says David Bader, director of the institute for data science at the New Jersey Institute of Technology. Sign up to TechScape A weekly dive in to how technology is shaping our lives after newsletter promotion 'We need to distinguish between genuine advances and marketing narratives designed to attract talent and investment. From a technical standpoint, we're seeing impressive improvements in specific capabilities – better reasoning, more sophisticated planning, enhanced multimodal understanding. 'But superintelligence, properly defined, would represent systems that exceed human performance across virtually all cognitive domains. We're nowhere near that threshold.' Nonetheless, the major US tech firms will keep trying to build systems that match or exceed human intelligence at most tasks. Google's parent Alphabet, Meta, Microsoft and Amazon alone will spend nearly $400bn this year on AI, according to the Wall Street Journal, comfortably more than EU members' defence spend. Rosenberg acknowledges he is a former Google DeepMind employee but says the company has big advantages in data, hardware, infrastructure and an array of products to hone the technology, from search to maps and YouTube. But advantages can be slim. 'On the frontier, as soon as an innovation emerges, everyone else is quick to adopt it. It's hard to gain a huge gap right now,' he says. It is also a global race, or rather a contest, that includes China. DeepSeek came from nowhere this year to announce the DeepSeek R1 model, boasting of 'powerful and intriguing reasoning behaviours' comparable with OpenAI's best work. Major companies looking to integrate AI into their operations have taken note. Saudi Aramco, the world's largest oil company, uses DeepSeek's AI technology in its main datacentre and said it was 'really making a big difference' to its IT systems and was making the company more efficient. According to Artificial Analysis, a company that ranks AI models, six of the top 20 on its leaderboard – which ranks models according to a range of metrics including intelligence, price and speed – are Chinese. The six models are developed by DeepSeek, Zhipu AI, Alibaba and MiniMax. On the leaderboard for video generation models, six of the top 10 – including the current leader, ByteDance's Seedance – are also Chinese. Microsoft's president, Brad Smith, whose company has barred use of DeepSeek, told a US senate hearing in May that getting your AI model adopted globally was a key factor in determining which country wins the AI race. 'The number one factor that will define whether the US or China wins this race is whose technology is most broadly adopted in the rest of the world,' he said, adding that the lesson from Huawei and 5G was that whoever establishes leadership in a market is 'difficult to supplant'. It means that, arguments over the feasibility of superintelligent systems aside, vast amounts of money and talent are being poured into this race in the world's two largest economies – and tech firms will keep running. 'If you look back five years ago to 2020 it was almost blasphemous to say AGI was on the horizon. It was crazy to say that. Now it seems increasingly consensus to say we are on that path,' says Rosenberg.
Geeky Gadgets
7 hours ago
- Geeky Gadgets
ChatGPT 5 Review : The AI Revolution We Hoped For or a Step Backward?
What if the most anticipated AI model of the year wasn't quite what you expected? OpenAI's ChatGPT 5 has arrived in a variety of different AI models, promising new advancements in technical problem-solving and software creation, yet sparking fierce debates about its broader impact and limitations. On one hand, it features an innovative routing system designed to optimize task performance, and on the other, it faces criticism for inconsistent results and its failure to significantly advance toward Artificial General Intelligence (AGI). This tension between potential and reality has left many wondering: is GPT-5 a step forward or a signal that scaling AI models is reaching its limits? In the video below, Wes Roth unpack ChatGPT 5's AI models and the key strengths and weaknesses of each, exploring its impressive capabilities in coding and tool development alongside its struggles with reasoning and basic math. You'll also discover how its unique routing system works—and why it has sparked frustration among users. Beyond the technical details, we'll examine the broader implications of GPT-5's performance for the future of AI development and innovation. Whether you're a developer, a tech enthusiast, or simply curious about the state of AI, this exploration will leave you with a clearer understanding of where GPT-5 shines, where it stumbles, and what it means for the next chapter of artificial intelligence. ChatGPT 5: Strengths and Challenges Polarized Reception The reception of GPT-5 has been notably divided, reflecting both its potential and its shortcomings. Strengths: GPT-5 demonstrates enhanced capabilities in generating software applications and solving complex, technical tasks, making it a valuable tool for developers. GPT-5 demonstrates enhanced capabilities in generating software applications and solving complex, technical tasks, making it a valuable tool for developers. Criticisms: Many experts argue that ChatGPT 5 represents only an incremental improvement over its predecessor, falling short of the fantastic progress expected in the pursuit of AGI. This polarized response has sparked debates about whether the current approach of scaling AI models is nearing its limits, raising questions about the future direction of AI research and development. Innovative Routing System: Promise and Pitfalls One of GPT-5's most distinctive features is its innovative routing system, designed to optimize task allocation by assigning specific tasks to specialized sub-models based on their complexity and computational requirements. This system aims to enhance both efficiency and performance, offering a more dynamic approach to handling diverse tasks. Despite its promise, the routing system has encountered significant challenges: Inconsistent task allocation: The auto-switcher, responsible for selecting the appropriate sub-model, has been criticized for producing suboptimal outputs in certain scenarios. The auto-switcher, responsible for selecting the appropriate sub-model, has been criticized for producing suboptimal outputs in certain scenarios. Unpredictable results: Users have reported frustration with the system's inability to consistently deliver reliable outcomes, particularly for critical or high-stakes tasks. OpenAI has acknowledged these issues and is actively working to improve the system's transparency and reliability. Enhancements in this area could significantly bolster user confidence and the model's overall utility. OpenAI GPT-5 Overview & Performance Tested Watch this video on YouTube. Here is a selection of other guides from our extensive library of content you may find of interest on GPT-5. Strengths: Coding and Tool Creation GPT-5 excels in domains requiring technical expertise, particularly in coding and tool development. Its strengths include: Software generation: The model can create functional software applications, streamlining development processes for programmers. The model can create functional software applications, streamlining development processes for programmers. Problem-solving: It effectively addresses complex challenges through code-based solutions, making it a powerful resource for technical industries. It effectively addresses complex challenges through code-based solutions, making it a powerful resource for technical industries. Integration capabilities: ChatGPT 5 demonstrates proficiency in following instructions and integrating seamlessly with external tools, enhancing its versatility. These capabilities position GPT-5 as a valuable asset for developers, businesses, and industries focused on automation and programming. Its ability to handle intricate tasks with precision underscores its potential to drive innovation in technical fields. Weaknesses: Reasoning and Basic Math Despite its technical strengths, GPT-5 exhibits notable weaknesses in fundamental areas such as reasoning and basic mathematics. These limitations include: Logical reasoning: The model struggles with tasks requiring deductive reasoning or nuanced problem-solving, often producing inconsistent or incorrect results. The model struggles with tasks requiring deductive reasoning or nuanced problem-solving, often producing inconsistent or incorrect results. Numerical accuracy: Simple mathematical calculations can pose challenges, particularly when the default routing system is relied upon. While premium features like 'max reasoning effort' can enhance performance in these areas, their accessibility is limited, potentially restricting the model's broader usability. Addressing these weaknesses will be critical for expanding GPT-5's applicability across a wider range of tasks. Transparency and User Feedback A recurring concern among GPT-5 users is the lack of transparency in its routing system. The inability to predict or control how tasks are assigned to sub-models has led to frustration and diminished trust in the model's reliability. In response, OpenAI has committed to several improvements: Enhanced transparency: Efforts are underway to provide users with greater insight into the task allocation process, allowing more informed interactions with the model. Efforts are underway to provide users with greater insight into the task allocation process, allowing more informed interactions with the model. Refined decision-making: OpenAI is working to improve the consistency and accuracy of the routing system's decision-making mechanisms. These initiatives aim to address user concerns and establish a stronger foundation of trust, making sure that GPT-5 can deliver more predictable and reliable performance across diverse applications. Broader Implications for AI Development GPT-5's mixed performance has sparked broader discussions about the trajectory of AI development. While its advancements in specific areas are undeniable, its limitations highlight the challenges of achieving significant breakthroughs through scaling alone. Key insights include: Need for innovation: ChatGPT 5 underscores the importance of exploring novel approaches to AI design, moving beyond incremental improvements to achieve fantastic progress. ChatGPT 5 underscores the importance of exploring novel approaches to AI design, moving beyond incremental improvements to achieve fantastic progress. Balancing strategies: A combination of scaling and innovative methodologies will be essential for overcoming current limitations and advancing AI capabilities. These considerations reflect the complexity of pushing AI systems beyond their current boundaries, emphasizing the need for a more holistic approach to development. Future Outlook OpenAI has expressed a commitment to addressing ChatGPT 5's shortcomings while building on its strengths. The model's proficiency in coding and tool creation suggests promising applications across various industries, from software development to automation. However, its struggles with reasoning and basic math remain critical areas for improvement. As the AI community continues to explore new methods and technologies, GPT-5 serves as both a milestone and a reminder of the challenges inherent in advancing artificial intelligence. Achieving AGI and other ambitious goals will require not only scaling existing models but also rethinking fundamental design principles to unlock new possibilities. Media Credit: Wes Roth Filed Under: AI, Top News Latest Geeky Gadgets Deals Disclosure: Some of our articles include affiliate links. If you buy something through one of these links, Geeky Gadgets may earn an affiliate commission. Learn about our Disclosure Policy.
