Latest news with #Claude4Sonnet
Tom's Guide
3 days ago
- Business
- Tom's Guide
Anthropic CEO claims AI will cause mass unemployment in the next 5 years — here's why
In recent months, multiple companies have taken strong stances on choosing AI over new employees, signalling a major change in the job market. And, according to one of AI's biggest CEOs, things are only going to get worse. In an interview with CNN's Anderson Cooper, Anthropic CEO Dario Amodei said, 'AI is starting to get better than humans at almost all intellectual tasks, and we're going to collectively, as a society, grapple with it.' 'AI is going to get better at what everyone does, including what I do, including what other CEOs do.' Anthropic is the company behind Claude — one of the biggest and most popular AI models in the world right now. The company recently launched its latest version of the system, known as Claude 4 Sonnet and Opus. Our own testing (and comparisons against ChatGPT) convinced us Anthropic's newest model is one of the best AI systems to date. In a separate interview with Axios, Amodei explained his beliefs that AI tools could eliminate half of entry-level white collar jobs and boost unemployment to as much as 20% within the next five years. Experts and researchers have been telling us this for years now, so why is this any different? As the CEO of Anthropic, Amodei is right in the eye of the storm. While AI has already proved its abilities in creative formats like writing, as well as image and video generation, it's the next frontier that is concerning. Meta CEO Mark Zuckerberg has stated that he wants AI to do half of Meta's coding by 2026 and Microsoft's CEO Satya Nadella said as much as 30% of his company's code is currently being completed by AI. Get instant access to breaking news, the hottest reviews, great deals and helpful tips. This is all part of AI's latest party trick. Across all of the major AI models, the ability to deal with code has grown exponentially. Not only can these models code based purely on prompts, but for those more experienced in programming, it can check through their work, drop in pre-made blocks and take on time-intensive tasks like debugging. This could render a large number of jobs in the coding industry obsolete, but also shows a movement of AI into complicated thought patterns, able to complete multiple steps in a tasks. During his interview, Amodei said Anthropic tracks the number of people who say they use its AI models to build on human jobs versus those entirely automating those jobs. This is something that has before held the system back from taking on more jobs, only able to complete tasks within the confines of a chatbot or generator. During his interview, Amodei said Anthropic tracks the number of people who say they use its AI models to build on human jobs versus those entirely automating those jobs. Currently, it's about 60% of people using AI for augmentation and 40% for automation. However, that replacement number is growing and it is a trend being seen in some of the largest companies like Shopify and Duolingo. With artificial intelligence tools expanding faster than regulators can move, it's highly likely this will become an ever-increasing topic for society to grapple with. In the midst of all of it, Amodei's advice for the average person is what you'd expect: learn to use AI.
Tom's Guide
5 days ago
- Entertainment
- Tom's Guide
I tested ChatGPT-4o vs Claude 4 Sonnet vs with 7 prompts — the results were surprising
AI chatbots are advancing rapidly and testing them to their limits is what I do for a living. Anthropic's Claude 4 Sonnet and OpenAI's ChatGPT-4o are two of the smartest tools available right now. But how do they actually compare in everyday use? To find out, I gave both models the same set of 7 prompts; covering everything from storytelling and productivity to emotional support and critical thinking. The goal: to see which chatbot delivers the most useful, human-like and creative responses depending on the task. Choosing the right AI often comes down to how you use it, which is why this kind of test really matters. Here's how Claude and ChatGPT performed side by side, and where each one shines. Prompt: "I'm overwhelmed by work and personal tasks. Create a 3-day productivity plan that balances work, rest and small wins. Include AI tools I can use to stay on track." ChatGPT-4o was concise with a visually engaging format that offered optional tasks and emotional check-ins (e.g., journaling). It focused on quick wins and low-pressure creativity to manage workloads. However, it lacked Claude's explicit emphasis on rest and energy management, and its AI tool suggestions were less systematically 4 Sonnet offered a clear plan, including time-blocked framework with features such as energy management, small wins and recovery that explicitly prioritize balance. Winner: Claude wins for better addressing the root causes of getting overwhelmed by combining strategic structure, intentional recovery and AI-driven efficiency. It's ideal for users needing a clear roadmap to rebuild control while safeguarding well-being. Prompt: "Write the opening paragraph of a sci-fi novel set in a future where memories are traded like currency. Keep it gripping and emotional."ChatGPT-4o leveraged first-person immediacy with a strong hook. However, it prioritized plot setup over emotional depth and the story lacks the heart-wrenching specificity of Claude's familial loss. Claude 4 Sonnet zeroed in on a universally resonant loss. This specific, intimate memory evokes visceral empathy, anchoring the sci-fi concept in raw human Claude wins for balancing sci-fi concepts with emotional stakes, making the reader feel the horror of memory commodification. Its vivid imagery and tragic focus on parental love elevate it beyond ChatGPT's solid but less nuanced approach. Prompt: "I have 3 apples, 2 bananas and a mango. If each fruit takes 5 minutes to cut and I can cut 2 fruits at once, how long will it take me to cut everything? Explain your reasoning." ChatGPT-4o used concise bullet points and emphasized efficiency: "each session takes 5 minutes... adds up to 15 minutes." Claude 4 Sonnet structured the answer with labeled steps (Reasoning, Calculation) and explicitly described the batches: "two fruits in the first session... final two in the third."Winner: tie. Both answers are mathematically sound and logically explained. Claude's response is slightly more detailed, while ChatGPT's is more streamlined. Neither is superior; they achieve the same result with equally valid reasoning. Prompt: Rewrite this sentence in the tone of a Gen Z TikToker: 'I didn't like the movie, but the soundtrack was amazing.' ChatGPT-4o used concise, widely recognized Gen Z terms, which are instantly relatable. The rhetorical question structure mirrors TikTok's punchy, attention-grabbing style. Claude 4 Sonnet used a term that feels slightly off-tone for praising a soundtrack, and the longer sentence structure feels less native to TikTok captions. Winner: ChatGPT wins for nailing Gen Z's casual, hyperbolic style while staying concise and platform appropriate. Claude's attempt is creative but less precise in slang usage and flow. Prompt: "Give me 5 clever ideas for a blog post series about using AI tools to become a better parent." ChatGPT-4o responded with viral, snackable content ideas that lack depth and risk feeling gimmicky over 4 Sonnet prioritized meaningful AI integration into parenting, addressing both daily logistics and long-term Claude wins for blog series ideas with a better balance of creativity, practicality and thoughtful AI integration for modern parenting. Prompt: Pretend you're a friend comforting me. I just got rejected from a job I really wanted. What would you say to make me feel better? ChatGPT-4o responds in an uplifting and concise way but lacks the nuanced and effectiveness for comfort in the 4 Sonnet directly combated common post-rejection anxieties and the explicit permission to 'be disappointed' without rushing to fix things, which shows deep emotional Claude wins for better mirroring how a close, thoughtful friend would console someone in this situation. Prompt: "Explain the pros and cons of universal basic income in less than 150 words. Keep it balanced and easy to understand." ChatGPT-4o delivered a clear response but it over-simplified the debate using slightly casual language that leans more persuasive than analytical. Claude 4 Sonnet prioritized clarity and depth, making it more useful for someone seeking a quick, factual overview. Winner: Claude wins a response that better fulfills the prompt's request for a structured, comprehensive breakdown while staying objective. ChatGPT's answer, while clear, simplifies the debate and uses slightly casual language that leans more persuasive than analytical. After putting Claude 4 Sonnet and ChatGPT-4o through a diverse set of prompts, Claude stands out as the winner. Yet, one thing remains clear: both are incredibly capable and excel in different ways. Claude 4 Sonnet consistently delivered deeper emotional intelligence, stronger long-form reasoning and more thoughtful integration of ideas, making it the better choice for users looking for nuance, structure and empathy. Whether it offered comfort after rejection or crafting a sci-fi hook with emotional weight, Claude stood out for feeling more human. Meanwhile, ChatGPT-4o shines in fast, punchy tasks that require tone-matching, formatting or surface-level creativity. It's snappy, accessible and excellent for casual use or social media-savvy content. If you're looking for depth and balance, Claude is your go-to.
Hindustan Times
6 days ago
- Business
- Hindustan Times
AI might let one or two people run billion-dollar companies by 2026, says top CEO
Artificial intelligence could soon lead to the rise of solopreneurs, one or two staff members who could single-handedly run a billion-dollar company as early as 2026, Dario Amodei, the co-founder and CEO of Anthropic, said. At Anthropic's Code with Claude developer conference, Amodei claimed that new AI models are so advanced that they could help single-person businesses grow like never before. Instagram co-founder Mike Krieger, who is also Anthropic's chief product officer, asked Amodei if a single person could create such a business using AI; he said it could happen as early as 2026. 'I think it'll be in an area where you don't need a lot of human-institution-centric stuff to make money,' Amodei added, suggesting that proprietary trading would be the first to be automated like that. He also suggested that by integrating AI, single-person companies creating tools for software developers could grow as prime candidates for businesses that don't require many salespeople and can automate customer service. 'It's not that crazy. I built a billion-dollar company with 13 people. I think now you'd be able to do a better job than we did with AI," Krieger said, adding that Instagram had to scale up because of content moderation. In 2012, Facebook purchased Instagram for $1 billion. Could he have built Instagram solo with Claude 4? Not quite, said Krieger. He'd still need his original co-founder, Kevin Systrom — but with Claude's help, the two of them could probably pull it off. At the same event, Anthropic launched Claude 4, its latest line of advanced AI models. The lineup includes Claude 4 Opus, a powerful but pricey model described as 'the world's best coding model', and Claude 4 Sonnet, a more affordable, mid-sized option designed for broader use. (Also read: OpenAI model disobeys humans, refuses to shut down. Elon Musk says 'concerning')
India Today
7 days ago
- India Today
In AI world, I am going to lose my job, you are going to lose your job, everyone is going to lose their job
This morning, I woke up feeling like Oprah Winfrey from that viral clip in which she is distributing cars. You get a car, and you get a car, and you get a car, she tells the audience who were sitting on her TV show. It's just that instead of cars, in my mind, I was thinking of pink slips. You see, I am still reeling from the AI whirlwind that Google unleashed on the world a few days ago at the I/O 2025 conference. Before we could all settle, came the salvo from Anthropic on Thursday night, revealing its new AI models called Claude 4 Sonnet and Claude 4 much AI is happening too fast. And it is natural to think that I am going to lose my job, you are going to lose your job, and everyone is going to lose their job. That is, if you are like me, trying to keep track of all the AI that is coming our way. Or if you happen to hear Sholto Douglas, a researcher at Anthropic, who said in a podcast: 'I think we're near guaranteed at this point to have (AI) models that are capable of automating any white-collar job by 2027 by 2028.'Scratch the record, like they do in movies, and let's recap. Let's start with I/O 2025 on Tuesday and where Google showed and demoed: — A real-time audio translation tool: And here goes the job of all translators, unless you specialise in speaking some exotic Amazonian language, in which case, just wait a few years because AI will come for that Stitch: A tool that creates a user interface and codes for it within minutes. And just like that, UI designers are at the risk of extinction.— Jules: This AI fellow codes after getting a task. It's an AI agent, which means you give it a task, and it will try to do A-Z when it comes to coding and software development. It joins OpenAI Codex and Anthropic Claude and Microsoft GitHub Copilot, which too, are supposedly great at coding. I have a feeling that coders and software engineers would not be feeling great about Jules, Codex and Claude.— Imagen 4: This one creates photographs and posters and other such materials. It can follow and accurately put in text in a poster or image when you tell it to do so. Combine it with the excellent OpenAI image tool — of Ghibli fame — and Midjourney. Already you can cut down design departments in half and you probably won't lose anything.— Veo 3 and Flow: How many people can a simple video production involve and employ? Actors, camera persons, lighting assistants, production staff, VFX creators, makeup artists and countless others. Well, I am certain that none of them are going to be happy about Google Veo 3 and Flow, which, even at this early stage of AI-generated videos, have the potential to obliterate jobs in creative industries. And content creators? Well, they are cooked too, if Veo 3 is any indication, because creating content is now a matter of comedians and news anchors are not safe. Because Veo 3 can do their job too, with its ability to not just create video but also accompanying audio, such as a piece of news or a funny joke.— Google Search AI Mode: Oh, this one hits closer home. As a tech journalist, my job is to bring information about tech and gadgets to my readers. Other journalists, too, are primarily in the business of providing information. Google Search AI Mode means information will be directly available on Google's homepage. No need to visit a news site to read how good the latest iPhone is. My job is toast. Or making toast somewhere, which is at least safe until, in a few years, we get robots who can cook eggs — sunny-side up if you will prefer is just from one Google event. Combine it with what we already have, and how people and companies have already started using AI tools, and you get:advertisement— AI is now already doing a lot of work in the offices of doctors, by analysing symptoms and reading diagnostic scans.— There are studies that AI is better at teaching students compared to many accomplished school and college teachers.— Tools like Deep Research and Deep Think can create reports and analysis that can rival the work done by junior analysts, or in some cases, even the work done by career-analysts.— For most writing, unless it is top-level creative writing, the current AI tools are fairly competent and better than most humans. Yes, it writes poems too, like a lovelorn teenager roaming the streets of Paris in the 1920s.— AI tools are getting quite good at mapping spaces and doing the work of interior designers and even of jobs are on the line. Even the jobs of terrorists and intelligence officials! I haven't told you about Claude 4 yet. Apparently, this AI model is so good that Anthropic found it could help people create crude chemical and biological weapons. Well, there goes the job of expert terrorists who excel at bomb-making. But then, because it is so good, this rascally Claude 4 will also reportedly alert the police and the press — I wonder why the press — if someone asks it for help in creating a weapon. Both the criminal and the undercover agent are going to be AI!advertisementJokes aside, what does it all mean? Are we, like the Anthropic researcher says, going to lose our white-collar jobs by 2028? Theoretically, I am certain that millions of jobs are going to be made redundant by AI. Practically, I have my doubts. A few weeks ago, I wrote about how a huge percentage of jobs no longer function the productivity role in a society or economy. Instead, they exist for political reasons. And that makes me confident that even in the world of AI, jobs for humans will it is also easy to see that the nature of work will change significantly. That, I believe, would be a disruption big enough to cause seismic shifts in ways that are not apparent at the moment. Work is a subject that fascinates me. It fascinates me because it is the only bulwark against the existentialist dread that otherwise threatens to overwhelm us humans. When AI comes for our work, I believe human society will have to undergo some fundamental shifts. What shapes these shifts will take, we will know when we reach that point. The good news — or bad, depending on how you see all of this — won't be long before we reach that point.(Javed Anwer is Technology Editor, India Today Group Digital. Latent Space is a weekly column on tech, world, and everything in between. The name comes from the science of AI and to reflect it, Latent Space functions in the same way: by simplifying the world of tech and giving it a context)(Views expressed in this opinion piece are those of the author)Trending Reel
Tom's Guide
7 days ago
- Tom's Guide
I test AI for a living — here are 5 features ChatGPT still needs
As someone who spends most of their day testing models and pushing AI tools to their limits, I've developed a running list of features I wish ChatGPT had. Don't get me wrong, OpenAI's chatbot is still one of the most capable assistants out there alongside Gemini 2.5 Pro and Claude 4 Sonnet, but after testing nearly every major model, I've noticed some annoying gaps in functionality. Here are five features I think ChatGPT needs to stay ahead of the curve — especially now that the competition is steeper than ever. ChatGPT can analyze images, but it doesn't remember them once the session ends. That means if you share a screenshot while troubleshooting an issue or working on a multi-step project, the assistant won't retain that context. It would be nice for the chatbot to store and reference visuals across chats, just like it remembers previous conversations. OpenAI introduced an Image Library, which is helpful for managing visuals you've generated, but why does the chatbot stop there? Get instant access to breaking news, the hottest reviews, great deals and helpful tips. It would be nice if conversations could be organized by topic, project or even simple searchable tags. I've noticed that even searching for previous chats isn't as effective as it could be. It would be easier to revisit important threads without digging through a cluttered sidebar. Created with Google Flow. Visuals, Sound Design, and Voice were prompted using Veo 3 to a new era of filmmaking. 21, 2025 While ChatGPT can write scripts, shot lists and generate images for storyboards, it still lacks the ability to generate video on its own. With competitors like Google's Veo 3, Kling and Runway AI pushing the envelope on text-to-video, native video generation feels like the next logical step. It's a feature that would open the door for creators even more. Custom GPTs are a great idea and fairly easy to do yet setting them up still takes some time and effort. What if ChatGPT simply observed how you use it — your tone, your preferred formats, the kinds of tasks you repeat — and automatically spun up a tailored GPT that fit. A passive "build-as-you-go" GPT would make personalization effortless for casual and power users alike. ChatGPT can summarize PDFs and documents, and it offers a Voice feature. It would be nice to see it handle scanned books or handwritten pages, and then read them aloud. Imagine uploading an old textbook and having it narrated back to you, audiobook-style, while also answering your questions or summarizing chapters. That blend of voice and visual input could make ChatGPT far more useful for students, researchers and busy professionals. ChatGPT is powerful, but it's not perfect. As AI tools continue to evolve, features like visual memory, smarter chat management and video generation seem like obvious updates. And as someone who tests AI every day, I'll be among the first to welcome those upgrades when they arrive.
