Latest news with #ChatGPTVision
Tom's Guide
3 days ago
- Tom's Guide
This ChatGPT ‘memory hack' changes everything — use these prompts to make it remember you
If you've ever found yourself reintroducing yourself, your tone preferences or even something as basic as your name to ChatGPT, you're not alone. While OpenAI's chatbot is famous for being smart, helpful and shockingly conversational, it's not always great at remembering what matters to you, which is why you might want to teach it. The good news is that ChatGPT actually has a memory feature that's smarter than most people realize, and there are a few tricks you can use to make it remember exactly what you want. Here's how to unlock ChatGPT's memory, plus a few sneaky hacks to get it acting like a real personal assistant. ChatGPT's memory is designed to remember helpful facts about you including your name, your job, your writing style and even your goals. Knowing these things can help the AI tailor responses over time. Think of it like your AI assistant building a mental file cabinet with your preferences inside. OpenAI first rolled out memory to GPT-4o users in early 2024, and now it's automatically turned on for most ChatGPT Plus users. But unless you're actively using it, or customizing it, you might not get the full benefit. To check if memory is on, go to Settings → Personalization → Memory. From there, you can view, edit or wipe everything ChatGPT has remembered about you. One of the simplest ways to store a fact in ChatGPT's memory is to literally prompt it to remember something. For example: Get instant access to breaking news, the hottest reviews, great deals and helpful tips. If memory is enabled, ChatGPT will usually respond with: 'Got it. I'll remember that for future chats.' If not, it may ask for permission to store that information. I've noticed that when I've asked ChatGPT to remember things, it doesn't always remember the first time. Sometimes, not even the second time. If you run into this problem, stay persistent and keep reminding ChatGPT to remember something until it actually does example, I once used ChatGPT Vision to help my mom match fabric for a project. From then on, ChatGPT thought I was a quilter. I had to tell the chatbot to forget that (much to my mom's dismay, I'm sure).Pro tip: You can also say 'Forget that' or 'Update my memory' if something changes — like your job or preferred tone. If you want to be sure it does not remember something, you can also use the temporary chat feature. Even if memory is off or you aren't a Plus subscriber, you can still simulate long-term memory using what I call the context chaining trick. Here's how: Start your prompt with: 'For the rest of this conversation, assume I'm a second grade teacher working on an end-of-year project for my students.' This doesn't persist across sessions, but it works surprisingly well for one-off tasks or multi-step projects. OpenAI makes it easy to see what ChatGPT has remembered — and yes, you should check it occasionally. Just type: 'What do you remember about me?' It'll respond with a summary of the info it has on file, like: 'You're a mom of three who juggles work, parenting and writing with a good sense of humor. You're no stranger to trampoline parks, fourth grade homework chaos or PTA drama. You're based in New Jersey, drive a Jeep and sometimes test AI tools for personal life (like IEP meetings or canceling gym memberships).' Here are a few fast rules to get the most out of ChatGPT's memory: ChatGPT won't magically know your preferences unless you teach it, but with memory, it can get surprisingly close. Whether you're writing a novel, planning a trip or just tired of repeating yourself, these memory hacks can turn ChatGPT into a genuinely helpful sidekick.
Tatler Asia
15-05-2025
- Tatler Asia
From generative AI and multimodal AI to digital twins, AI explains itself
2. Multimodal AI Above Gemini, developed by Google Deepmind, is a multimodal AI that can interpret various inputs such as images, text, code and video () Multimodal AI refers to AI systems that can understand and process multiple types of information or 'modes' at once, such as text, images and audio. This allows them to combine different information sources to understand situations more comprehensively. Think of it like how humans use multiple senses: we hear words, see facial expressions and notice body language simultaneously to understand a conversation. Similarly, multimodal AI might analyse a video by recognising spoken dialogue, understanding visual actions and processing any text that appears on screen. This creates a richer understanding than any single mode could provide alone. You might experience multimodal AI in tools like ChatGPT Vision or Google Gemini, where you can upload an image and get detailed text-based analysis or ask questions about it, seamlessly blending visuals and language. Read more: AI leaders are unlocking the future with technology 3. AI companions Above A user customises an avatar for their personal AI chatbot Replika through a smartphone app () AI companions are virtual assistants or digital friends powered by AI. These applications are designed to interact with you in a human-like manner; they can chat, help with tasks, answer questions or even provide entertainment. Unlike simple tools, they aim to offer a sense of relationship and personalisation. They learn your preferences over time, adapting to your needs and habits. A virtual assistant on your phone might notice you typically set morning alarms and eventually begin suggesting them at your usual times. Some more advanced companions can even engage in casual conversation or remember details from past interactions to create a more continuous relationship. Yet, there is some controversy surrounding the impact of the AI companion bots on their users. In 2024, a mother in the US filed a lawsuit against Character Technologies, Inc, claiming that its bot contributed to her son's suicide. She alleged that the teenager openly discussed suicidal thoughts and engaged in highly sexual conversations with it. Read more: Meet Moflin, Casio's emotional support robot, and other AI companions 4. Computer vision Above A display showing the image processed by Quantum AI security camera at a booth during the Mobile World Congress 2024 in Spain () Computer vision enables machines to 'see' and make sense of visual information from the world around them. Just as our eyes capture images and our brains interpret them, computer vision systems use cameras to capture visual data and algorithms to understand what they are looking at. This field encompasses a range of capabilities: image classification, object detection, image segmentation and facial recognition. It has transformed numerous industries, from autonomous vehicles that rely on computer vision to navigate safely to medical imaging systems that help doctors spot diseases in X-rays and MRIs with greater accuracy. Read more: How AI Guided's Florence Chan helps the visually impaired navigate the world better with a smart belt 5. Digital twins Above A tablet displays the buildings at The Bund in Shanghai using a digital twin app () Digital twins are virtual replicas of real-world objects, systems or processes. Imagine having a digital model of your car, your house or even an entire factory that mirrors how the real thing operates and behaves. They have become increasingly valuable across industries to improve efficiency and predict potential issues. For example, in luxury real estate, digital twins can preview smart home configurations or model interior lighting throughout the day before the property is ever built. Read more: The AI book scraping issue explained 6. Synthetic media Synthetic media refers to content generated or altered using AI and other digital technologies. This includes images, videos, audio and text that are either entirely created by AI or manipulated to create new, realistic experiences. You might encounter synthetic media in social media filters that transform your appearance in real time or in music, where AI can compose songs that sound remarkably like popular artists performed them. Above A CBC news video explores the potential misuse of AI-generated videos in elections (Video: CBC News) While this technology presents creative opportunities, it also raises significant concerns regarding authenticity and trust. In some instances, individuals have been deceived by synthetic media featuring the voices of their loved ones in distress, leading them to transfer thousands of dollars before realising they had been scammed. Read more: 7 most expensive AI art pieces ever sold 7. Prompt engineering Above A user typed a prompt to inquire about OpenAI's ChatGPT () Prompt engineering is the skill of crafting questions or instructions for AI systems in ways that produce the most valuable and accurate responses. Think of it as learning how to communicate effectively with AI. Just as you might phrase a question differently depending on whom you're asking, prompt engineering involves understanding how to frame your requests to AI systems for optimal results. For instance, asking a vague question like 'What do you think?' might yield a generic response. However, asking something specific, such as 'What are three critically acclaimed British films released in the past two years?' is more likely to generate a precise and helpful answer. As AI becomes more integrated into our daily lives, understanding how to communicate our needs to these systems effectively becomes an increasingly valuable skill. Meet the Gen.T Leaders of Tomorrow from the Technology sector. NOW READ The hidden wisdom of investing legends: 5 powerful lessons for a prosperous life Here's what to know about 'soft power' and Joseph Nye, the man who conceptualised it From the world's best action camera to the worst drone: The ups and downs of GoPro Credits This article was created with the assistance of AI tools
