write a 4chan greentext ai

7 Greentext Stories That Will Make You Feel Better About Yourself

And this is still the lighter side of 4chan. 

• Advertisement

Stay entertained and up to date by following us on Facebook and Instagram!

Next on memebase.

Follow The Laughs

Welcome to the age of ChatGPT: an overview of 15 generative AI writing tools

By Nikki Michaels — Contributing Writer

No, this wasn’t written by an AI, but you’d be forgiven for wondering. When OpenAI released ChatGPT to much fanfare in November 2022, they changed the technological landscape forever — and brought generative artificial intelligence (AI) to the forefront of our collective consciousness.

The rapid adoption of ChatGPT has led to (or at least unearthed) a sea of generative AI tools designed to help us with everything from manual research to SEO optimisation to bonafide knowledge work .

But if you’ve attempted to look into any one of these tools, you may’ve felt overwhelmed and given up entirely (understandable; I’ll confess to a certain level of overwhelm in my research for this piece). So to make things easier, we thought we’d run through 15 of the major players in the AI writing tools and content creation space.

Here’s a quick outline of what exactly these tools are, their major use cases, and a few general advantages and limitations — as well as an overview of some tech you might want to consider, from Canva’s new Magic Write tool to ChatGPT itself.

What do the BBC, Tripadvisor, and Penguin have in common? They craft stunning, interactive web content with Shorthand. And so can you! Publish your first story for free — no code or web design skills required. Sign up now.

What is an AI writing tool?

AI writing and content creation tools automatically produce content based on prompts. They do this using a combination of AI technology (such as natural language processing) and proprietary algorithms.

They’re trained on neural networks (like OpenAI’s GPT-3, DeepMind’s Gopher, or Bloom), systems designed to mimic the human brain and perform complex tasks such as pattern recognition, problem-solving, or storytelling .

Basically, these tools take these massive language models and pair them with custom algorithms, creating novel technology specifically designed for writing and content.

So a tool trained on GPT-3, for instance, will use the network’s 175 billion parameters of data as its foundation, and then its proprietary algorithms interact with those parameters to produce a desired result — in this case, natural-language writing. 

More and more businesses are turning to AI writing software and content creation tools for help with short- and long-form content writing. Popular use cases include social media copy, landing page copywriting, sales and marketing emails, and SEO’d blog posts. Content teams are also experimenting with many other types of content.

Although there’s huge potential in this field, there are also significant drawbacks you should be aware of before deciding to implement an AI writer in your work.

Advantages and limitations of generative AI

As a rule, every generative AI product touts time savings, cost savings, and automation as key benefits. And it’s true — we're simply no match for the speed of these new tools.

AI writing and content creation tools can also help businesses improve efficiency and outputs, optimise existing content for SEO with keyword research, and scale their content marketing operations.

But these tools have a few considerable limitations as well:

• Misinformation . When an AI tool writes content for you, it’s pulling from the massive dataset that its neural network has been trained on — and we all know the internet is rife with misinformation. Basically, this means it’s imperative that you fact-check any content before you publish it to ensure vital information is correct.
• Plagiarism . Similarly, it’s possible that any given AI tool may produce plagiarised content for you and call it original. Not ideal.
• Lack of “humanity.” Don’t get me wrong; a few of the options out there are genuinely impressive when it comes to writing and grammar checking skills, and they often score well on baseline readability metrics. They can write a compelling subject line and email, and they can produce quirky LinkedIn and Meta content that’ll make you stop scrolling. But they don’t have a “voice” in the same way that a human writer does, because humans don’t consistently choose the statistically most likely word at every turn. So if you rely on them for all your content, you run the risk of it being boring, unoriginal, and uncreative.
• Devaluation by Google algorithms . Of all the limitations of AI writing and content creation tools, this may prove the biggest in terms of business impact. Google itself says its goal is to “help people see more original, helpful content written by people, for people, in search results.” Read: use AI to pump out content for search engines to your detriment. Ironically, Google also uses machine learning algorithms to detect AI-written content, so it knows what to rank and what to devalue.

With all this in mind, you can probably understand why many of these tools are marketed as “writing assistants” or “members of your team” rather than as replacements for your in-house or freelance writers, editors, and content creators.

But if you’re in the market for an AI writing or content creation tool, here’s a rundown of 15 to consider.

15 AI writing and content creation tools

A chatbot designed specifically for conversation, OpenAI’s ChatGPT is most useful for the research phase of the writing process, although you can use it to write content like emails if you wish. It can converse with you and answer follow-up questions, and it produces remarkably sophisticated natural-language responses that require little grammar or spelling oversight. It’s fast and accurate, and it’s constantly improving as OpenAI refines the GPT-3 model it’s built on.

But it’s also limited by GPT-3; it may produce inaccurate or biased responses or responses that don’t fit within the context of your question, and its output is only as good as your prompts.

ChatGPT has a free version, but you may not be able to use it during peak hours; $20 per month on the Plus plan will get you access even during those peaks, and you’ll also enjoy faster responses and first use of new features and improvements.

2. Jasper AI

Formerly known as Jarvis AI until Marvel got wind of the name, Jasper AI is one of the most popular alternatives to ChatGPT. Billed as a generative AI platform that can help you create tailored and high-converting brand content 10 times faster than you’re used to, it’s commonly used to produce copy for social media, ads, blog articles, emails, and website pages. It can also make you AI-generated art. Jasper allows you to input minimal information on a topic and then mines its data (GPT-3 at the time of writing) to create relevant content in your chosen format; it also lets you choose a particular tone so it’s closer to your brand voice.

It’s fast, able to translate copy into more than 25 languages, offers more than 50 content templates, has a google doc integration, chrome extension, and claims to produce plagiarism-free content — but its output requires thorough fact-checking, editing, and proofreading, which may mean you don’t save as much time and money as promised.

Jasper’s pricing models are based on the number of words generated, not the number of words you end up with — so if the AI misinterprets your request or produces an irrelevant response, you’ll be billed for it. For this reason, you’re perhaps better off using Jasper for short-form copy. Boss Mode, which they say works best for individuals and small teams, works on a sliding scale; 100,000 words is estimated to cost $82 per month. Custom plans are on offer for larger businesses.

Used by the likes of Microsoft and Nestlé, Copy.AI claims the same 10-times-faster statistic as Jasper. Built on GPT-3, their technology is designed to produce human-like content marketing copy, and they offer more than 100 templates for both short- and long-form content. This AI copywriting tool is particularly useful for ads, product descriptions, and social copy. The platform itself is intuitive and easy to navigate, and the quality of its output is high.

Quality notwithstanding, you still need to proof and edit your Copy.ai content. It’s also worth noting that Copy’s output fully depends on the quality of your input. That said, this tool offers a free plan, which nets you 2,000 words per month and unlimited projects. Copy.ai ’s Pro plan is $36 a month for unlimited words.

As you might guess from the name, SEO.ai specialises in producing content optimised for SEO. You tell the AI about your topic and your target audience, and it’ll identify keywords that it then uses to write intros, headings, full blog posts, and more. You’ll see your SEO score actively rising as you instruct the AI on what to write, and you can even benchmark your content against your competition.

SEO.ai also allows you to customise your content by asking it to do things like “write a metaphor”, or “rewrite this section as if I’m a PhD student.” It’s able to write in more than 50 languages and can also suggest SEO-friendly topics for your business. But its SEO prowess is also its weakness; this isn’t the most creative or compelling writing, and it’s likely to sound generic and repetitive.

You also have a limited number of AI-generated words per month; after your seven-day trial, the Standard plan is $49 a month for 40,000 words and 40 AI keyword reports. Enterprise pricing is available on request.

5. Sudowrite

The only creative writing tool on this list, Sudowrite is billed as Earth’s most advanced long-form writing app. It’s designed to help you “have more fun” and to be a brainstorming partner; using GPT-3 and proprietary narrative model-based algorithms, it claims to offer AI writing features unavailable anywhere else. Meant for the likes of novels, short stories, and screenplays, Sudowrite knows the common tropes of literary genres and can help you with prose and descriptions or give you ideas for where your plots can go.

It’s much better at replicating human emotion than other AI writing tools because it’s designed for creativity — but there’s an argument to be made that overuse of or reliance on this tool limits the writer’s originality. A Hobby & Student plan will cost you $10 a month for 30,000 words; Professional is $20 for 90,000 words; and Max is $100 for 300,000.

Rytr counts clients like Adidas, Dell, Ford, and Pfizer among its roster and claims to “create high-quality content, in just a few seconds, at a fraction of the cost.” Choose your use case — there are more than 40 templates on offer, so less than some competitors — add your own input for context, and Rytr will produce original content in more than 30 languages and more than 20 tones of voice. Rytr enables users to write a 1,000-word piece in 15 minutes using its rich-text editor and markets itself as an “all-in-one writing platform,” with features for SEO, grammar, and more.

The platform is easy to use, but the output quality requires thorough review. Use Rytr’s free plan to generate 5,000 characters a month; upgrade to Savings for $9 a month (50,000 characters) or Unlimited for $29 (infinite characters).

7. Chibi AI

Chibi’s point of difference hinges on “context”; the tool allows you to add as much context as you’d like, unlike other tools that restrict this input. Designed for bloggers and marketers, it offers pre-written and tested prompts that you can pair with your own custom options to create quality content at speed. You can use Chibi to rephrase pieces you’ve already written or ask it to be your content writer; its templates are all fully customisable.

The platform isn’t quite as intuitive as many of its competitors, and there’s less information available on its capabilities — but you can use it as much as you want for $42 a month or $420 a year.

Frase is designed for SEO specialists, content marketers, and marketing agencies. Like SEO.ai, it’s first and foremost an SEO tool, capable of helping you research, write, and optimise high-quality SEO content at speed. Use it to generate content briefs or complete copy, and benchmark everything against your competitors with the Optimization feature. Integrate Google Search Console to use Analytics, which can identify your content gaps, prevent content decay, and automatically identify your most advantageous content opportunities.

Again, the writing quality here leaves much to be desired; as an SEO tool, Frase produces content that’s generic and often filled with errors. It’s also expensive compared to competitors; though you can try it for free at $1 for five days, Solo will run you $14.99 a month for one user, four articles a month, and 4,000 AI-generated words. Basic rises to $44.99 for one user, 30 articles, and 4,000 words, and Team will net you three users, unlimited articles, and 4,000 words for $114.99 a month.

Anyword is focused on sales and conversions, and their clients include names like Red Bull and NBC . Built for performance marketers, content marketers, and full marketing teams, Anyword is capable of writing copy for the likes of social media, landing pages, blogs, webpages, and emails, and it offers predictive analytics that will estimate engagement before you publish, allowing you to save time and instantly identify top-performing copy.

Anyword touts more than 85 “data-driven ways to create copy that converts,” as well as ad account integration, the option to blacklist words, and a Continuous Optimization feature. You can use it to create, improve, and score copy based on your business goals, and its built-in plagiarism checker claims to ensure original content.

Anyword is best suited for short-form copy, and it only uses APIs — no plug-and-play. It’s also expensive if you want to use features like Predictive Performance Score. Those are only available on the Data-Driven plan, which begins at $83 a month and works on a sliding scale based on how many words you need. Starter employs the same sliding scale and begins at $24 a month.

10. INK for All

INK’s key differentiator is its ability (or so it claims) to detect AI content and protect your brand by creating copy that won’t get devalued by search engines. Billed as an all-in-one solution for content planning, writing, paraphrasing, editing, and SEO, it offers patented SEO and user engagement optimization that increases reach by 450%. Use it to rewrite and optimise existing copy, or start from scratch with its template library. You can also integrate it with WordPress.

INK only works in English, and you can only use it on desktop. It also occasionally produces nonsensical outputs. Professional offers unlimited writing, SEO articles, and AI images for $49 a month; Enterprise gets you three users, advanced team management, team performance tracking, and premium support for $119 a month. Get discounts if you pay yearly.

11. Copysmith

Copysmith is designed for manufacturers, retailers, and agencies and best used for product descriptions, marketplace content, and social media posts. It produces SEO-optimised content that outranks competitors, and it can write descriptions tailored to the requirements of major retail platforms like Amazon, eBay, and Etsy. Use it to refresh old content, collaborate with your team in real-time, A/B test Facebook ads, and write blogs at speed.

CopySmith is ideal for short-form content; longer-form content comes with errors and requires thorough review.

The Starter plan is $19 a month for 20,000 words and 20 plagiarism checks; Professional is $59 for 80,000 words and 100 plagiarism checks; and Start-up runs on a sliding scale beginning at $299 a month for unlimited words and plagiarism checks.

12. Canva Magic Write

Canva’s newly released Magic Write allows you to input a simple prompt to produce outlines, lists, captions, and content ideas. It’s designed for marketers, small business owners, and anyone who needs to produce quality content at speed and on a limited budget. It’s billed primarily as a way to push past writer’s block, brainstorm ideas, and optimise your workflow, and it can help you go from keywords to full draft fast.

It’s also designed, of course, to work intuitively with Canva, allowing you to build beautiful designs that complement your text. Magic Write works on all devices and offers real-time collaboration, but its pricing options are complicated. Check out Canva’s website for more details.

13. Notion AI

Notion is marketed as your “teammate” before, during, and after the writing process. They encourage you to think of it as a partner — an AI writing assistant that can help you brainstorm, summarise, write, and edit. Its primary applications include writing first drafts, generating lists about virtually any topic, editing grammar and syntax, and summarising long notes or documents.

Notion’s limitations are those of all generative AIs; it’s prone to producing copy with errors and inaccuracies, so anything it creates needs thorough review. Try it for free with any Notion plan, and add it to a paid plan for $8 per member per month.

14. Content at Scale

Content at Scale is built for companies wanting to scale their content marketing: agencies looking to streamline processes, start-ups that need to quickly grow their content operations, multi-site owners who need content for numerous outlets, and the like. This platform claims to produce content “so human-like it bypasses AI detection.”

Unlike many AI writing tools that are better suited for short-form content, Content at Scale works well for long-form content like blog posts, producing high-quality writing that’s optimized for SEO. The content isn’t perfect, though — your plan comes with two free reruns for that purpose — nor is the platform’s AI detection prevention infallible. It’s also got expensive starter pricing; plans begin at $150 a month.

15. WriteSonic

WriteSonic claims the same 10-times-faster statistic as Jasper AI and Copyai . Designed for ecommerce stores, marketing teams, and entrepreneurs, it’s an AI content generator that writes blogs, ad copy, emails, and website copy that are SEO-optimised, plagiarism-free, and designed to increase conversions.

WriteSonic offers AI-powered editing tools that competitors don’t, and its ecommerce tools can help you improve your Google ranking. The platform also touts a tool called Chatsonic, which they say is “like ChatGPT but with 4x superpowers.” Like other tools of its kind, it produces copy that needs proofing and fact-checking.

WriteSonic has an interesting pricing model; the number of words you get for a certain price changes based on the “quality” you choose: economy, average, good, or premium. On the long-form plan, $12.67 a month will get you 19,000 premium words or 190,000 economy words. Try WriteSonic for free for a month, and then check out their website for more pricing details.

Publish your first story free with Shorthand

Craft sumptuous content at speed. No code required.

Tired Of Web Scraping? Make The AI Do It

[James Turk] has a novel approach to the problem of scraping web content in a structured way without needing to write the kind of page-specific code web scrapers usually have to deal with. How? Just enlist the help of a natural language AI. Scrapeghost relies on OpenAI’s GPT API to parse a web page’s content, pull out and classify any salient bits, and format it in a useful way.

What makes Scrapeghost different is how data gets organized. For example, when instantiating scrapeghost one defines the data one wishes to extract. For example:

The kicker is that this format is entirely up to you! The GPT models are very, very good at processing natural language , and scrapeghost uses GPT to process the scraped data and find (using the example above) whatever looks like a name, district, party, photo, and office address and format it exactly as requested.

It’s an experimental tool and you’ll need an API key from OpenAI to use it, but it has useful features and is certainly a novel approach. There’s a tutorial and even a command-line interface, so check it out.

19 thoughts on “ Tired Of Web Scraping? Make The AI Do It ”

I’m tired of web scraping existing in the first place. The internet is for humans.

Humans use web scraping to make life easier for themselves and others.

That’s always how it starts.

What in the world does that mean?

Don’t get him started.

Only about 5% of the Internet is for humans. Only 0.03 – 0.04% is available without a login.

96% of statistics are made up on the spot……

Plenty of sources. Here is one https://www.spiceworks.com/it-security/security-general/articles/dark-web-vs-deep-web/

> The internet is for humans.

When was that ever true / the case? It was developed as a nuclear Armageddon safe communications network for military use (wasn’t it?). And since then it was fur businesses, advertisement, porn, cats, lies, propaganda, connecting all kinds of idiots to stay in their Q believes (religions=sects, conspiracy stories, esoteric BS just to name a few).

I mean yes, it helped connect scientists, paved the way for more open source ides in several areas (and many other things) but still…

Wars are being fought “for humans”, the developed world is destroying nature for >100 years for (their) humans. -> What exactly do you mean with “for humans”?

Likely to be used more and more by different types of bots as time goes on. Humans maybe relegated to 2nd class users. Or 3rd class if you include cats.

Indeed. We have electronic mail before the Internet. Systems used to dial up to each other on a regular basis to exchange messages.

… really makes you think !

The nuclear war thing is a myth, some of the Arpanet crowd began to speculate about survivability of communication networks later on, but it was created as a boring old way to network a bunch of mainframes operated by the military and academia. I believe the origin myth came from InfoWorld in the 1990s, and it doesn’t really make sense given the dependence on a fairly small number of leased lines and point to point microwave links.

The point about it not being for humans still stands though.

Surely “scrapegoat” would be a better name?

“you’ll need an API key from OpenAI to use it,”

This means it’s going to cost money to scrape pages… slowly. I wouldn’t really trust it to return perfect results either. I’ll stick to writing bots.

Right… What happens when it decides that the scraping results are too sensitive and it leaves out information? You’d have to run a normal web scraper in parallel to verify that it isn’t leaving information out or hallucinating extra information.

Exactly this ,the limitations of chatgpt doesnt make it worth it

Pretty cool project. I’m working on something similar that uses GPT to generate web scrapers -> https://www.kadoa.com

Leave a Reply Cancel reply

Please be kind and respectful to help make the comments section excellent. ( Comment Policy )

This site uses Akismet to reduce spam. Learn how your comment data is processed .

Never miss a hack

If you missed it.

You’ve Probably Never Considered Taking An Airship To Orbit

The Great Green Wall: Africa’s Ambitious Attempt To Fight Desertification

Your Open-Source Client Options In The Non-Mastodon Fediverse

Supercon 2023: MakeItHackin Automates The Tindie Workflow

The Computers Of Voyager

Our columns.

The Art Of Hackaday Hack Chat

Hackaday Links: May 12, 2024

Institutional Memory, On Paper

Hackaday Podcast Episode 270: A Cluster Of Microcontrollers, A Rocket Engine From Scratch, And A Look Inside Voyager

This Week In Security: TunnelVision, Scarecrows, And Poutine

By using our website and services, you expressly agree to the placement of our performance, functionality and advertising cookies. Learn more

Anonymous group uses AI to add 'modest' clothing to women in photos, igniting debate over policing women's bodies online

I n the wake of explicit AI-generated photos of Taylor Swift spreading on social media, and conversations surrounding the lack of resources for deepfake victims, another group of people are using artificial intelligence technology to manipulate women’s bodies differently — in this case to cover them up.

The hashtag movement called #DignifAI was born on 4chan, an anonymous online message board with little moderation or regulation on what users can post. The AI Swift photos allegedly also originated on 4chan .

A post from Jan. 31 describing #DignifiAI has since been taken down but has been copied and pasted elsewhere. Titled “Operation/TradAI,” an anonymous 4chan user asked participants to use “the power of AI” to “clothe” women in photos, “purify them of their tattoos” and “lengthen their skirts.” Yahoo News will not be linking out to any of the posts.

By “TradAI,” the 4chan user is referring to “trad” or “ tradwife ," a shortened slang of “traditional” that was also coined on 4chan. The “trad” ideology is about wanting to go back to a stereotypical archetype of a woman; it’s a romanticization of the woman being the homemaker and doting on her husband and kids.

“Give them modest+stylish clothing such as loose sweaters, dresses that cover shoulders/chest/back/knees, blue jeans/flannel long sleeve for country girl aesthetic, angelic/righteous theme,” an anonymous 4chan user detailed. “Put them in a feminine and wholesome role, such as gardening, cleaning, being a caregiver, etc.”

Nic Coppage, a social worker at the National Alliance for Eating Disorders who studies the correlation between mental health and body acceptance, told Yahoo News that regardless of whether someone is adding or taking away clothes on photos of women, it “reduces women to sexual objects.”

“[It] brings up the question of bodily autonomy and takes power away from the woman in the image to decide who can see her body and how much of her body she wants [to be] shown,” Coppage explained.

While #DignifAI isn’t about revealing more when it comes to women’s bodies, the behavior is still relevant to the overarching concern that AI is helping strangers manipulate and police women’s bodies online.

4chan strongly emphasizes anonymity; unlike other online message boards such as Reddit, 4chan users don’t need to make an official account or pick a username. They can’t be directly messaged and, unless they use the same name across several posts or reveal something personal about themselves, it’s difficult to trace them.

That anonymity is what makes pinpointing the origins of #DignifAI to one user challenging. It started on 4chan’s /pol/ board, which stands for “politically incorrect” and is one of the platform’s most popular boards. An investigation presented at the 2017 International Association for the Advancement of Artificial Intelligence (AAAI) Conference on Web and Social Media found that “hate speech is predominant on /pol/,” with “many of its posters subscribing to the alt-right and exhibiting characteristics of xenophobia, social conservatism, racism and, generally speaking, hate.”

Is this part of a larger problem seen on 4chan? Or is #DignifAI a result of ‘trolling’ and trying to incite a reaction?

4chan is commonly associated with incels . The term “incel,” which refers to “involuntarily celibate,” is ascribed to people who share a belief in male supremacy and dominion over female bodies.

Andrew G. Thomas , an integrative psychotherapist and senior lecturer at the Swansea University School of Psychology who has studied the incel phenomenon, told Yahoo News that there is no clear-cut explanation for why 4chan users are participating in #DignifAI. He believes it can be traced to a desire to lash out at women who they perceive hurt them by posting provocative photos online.

“The women who want to show off their bodies don’t want to be covered up or have their tattoos removed, and [Swift] doesn’t want AI images generated,” Thomas said. “I see this as, ‘Well, you are hurting me, so I’m going to hurt you back.’”

But some 4chan users claim #DignifAI is one big joke to deliberately provoke criticism. Recent comments seem to revel in the current media coverage and share links to articles, calling the outrage “hilarious.”

Kat Tenbarge, a tech and culture reporter for NBC News Digital who has reported on #DignifAI, told Yahoo News that “4chan posits itself as a community that engages to troll” so it can appear as if participants don’t have any stake in what they’re doing.

“The action that they're taking still has meaning and subscribes to some form of ideology,” Tenbarge explained. “And oftentimes, it does reflect on the ideology that they truly have. And I think that's the case here.”

There’s an argument several 4chan commenters make that #DignifAI “is not harming” the women “in any way.” But, Tenbarge points out, creating “non-consensual images” of women is harmful.

“A lot of women who are now being attacked with this #DignifAI tool appear to be the ones who are in charge of their own image,” Tenbarge explained. “The message that's being sent is that [#DignifAI is] trying to humiliate you for your choices, whatever those choices may be.”

#DignifAI targets aren’t always celebrities or influencers. What can the average person do to protect themselves?

Emily Poler , a Brooklyn, N.Y.-based litigator specializing in intellectual property, told Yahoo News that the issue that lawmakers need to figure out quickly with regard to AI is what to do about non-commercial users.

Copyrighting personal photos is possible — and easier if they haven’t been posted online. But even copyrighting photos doesn’t guarantee people won’t manipulate them. The copyright holder has to find unauthorized copies or versions of the photo and then file a “takedown notice.”

Overall, Poler said, lawmakers need to move quickly to figure out how to protect people from non-consensual AI edits. Especially as 4chan users in the #DignifAI forum are warning the public that “they haven’t seen anything yet.”

Pocket-lint

With gpt-4o, chatgpt can generate art with text that’s actually readable.

OpenAI's new GPT-4o model for ChatGPT finally fixes one of AI's biggest flaws: Text on AI-generated images.

Key Takeaways

• ChatGPT has a new model, GPT-4o, and it can integrate text, visuals, and audio.
• More importantly, the new model allows correct text placement on generated images.
• GPT-4o's updated capabilities could help with graphic design. It also doesn't require a paid subscription, and there's a desktop app.

ChatGPT can spit out text and, with the DALL-E integration , churn out images, but ask the artificial intelligence platform to combine the two and the result is typically an unreadable, jumbled mess. That’s changing, however, with the move to ChatGPT’s GPT-4o or Omni. While OpenAI’s demonstration on May 13 focused on using the end-to-end text, vision, and audio capabilities to have a real-time conversation, the update could bring key graphic design capabilities to ChatGPT. Early demos show the AI not just generating images that have legible, correctly-spelled text, but using an existing image of a person to replicate that face in the new image.

8 ways ChatGPT Plus is better than Gemini Advanced and Copilot Pro

Gpt-4o's approach to text, visuals, and audio, everything is integrated into a single model.

The key change coming with the launch of GPT-4o is the ability to both input and generate any mix of text, audio, and images. That’s because OpenAI trained a new model end-to-end that works across text, vision, and audio. Previously, GPT-4 would use separate models for audio, text, and images. With everything integrated into a single model, OpenAI explains that ChatGPT doesn’t lose information between models, which opens up a number of new possibilities.

I tried ChatGPT Plus. Here's everything it can do

While the live demo on May 13 focused on how that single end-to-end model allows you to use video to solve homework problems or have a real-time audio conversation, it also helps correct something the AI model is notoriously bad at: Placing text on an image. GPT-4 can attempt to place text, but it typically results in misspellings, even when you tell the chatbot exactly how to spell it.

ChatGPT was able to generate images with legible, correctly spelled text taken from the prompt.

In several samples of the upcoming GPT-4o’s capabilities, the AI was able to place writing on an image of a typewriter, create a graphic with a poem, and create a movie poster. In the demonstrations, the wording was given to the AI, with misspellings in generated text not explicitly spelled out. But ChatGPT was able to generate images with legible, correctly spelled text taken from the prompt.

You can use real faces in generated images

Imagine making a movie poster with actors' faces.

In one demonstration, ChatGPT created a movie poster with the actors’ faces on it along with the correctly spelled text. This was made possible by uploading the photos of the actors and spelling out the text to include. While some AI platforms can create a new photo with a real person's face, ChatGPT wasn't previously able to create a photo that had much likeness to the original.

ChatGPT created a movie poster with the actors’ faces on it along with the correctly spelled text.

In another deomstration, the chatbot was able to place the OpenAI logo on an image. Another tasked the bot with creating a concrete poem where the word Omni appeared in the shape of the OpenAI logo.

The generated images in OpenAI’s demonstrations are not perfect -- when asked to take one correctly spelled poem image to dark mode, the software generates some misspellings. But the demonstration shows a much more legible, sensible result than the nonsensical way that GPT-4 generates text on images.

What you should know about ChatGPT Voice: How it works, what it can do and more

The software’s new capabilities in handling a mix of text-photos-speech also allow it to answer questions about a photo and extract text from images.

The demonstrations suggest ChatGPT could have more capabilities in graphic design with the launch of GPT-4o over the next few weeks. However, those capabilities could have some consequences. One of the easiest ways to tell if an image was generated by AI is to look at things like street signs or laptop screens where text appears jumbled. If AI learns to spell on images, that’s one less feature to signal the authenticity of an image floating around the web.

The end-to-end model integration text vision and audio also comes with faster speed, more features without a paid subscription, and a desktop app for Mac . OpenAI says that GPT-4o will roll out over the next few weeks.

Q: When will GPT-4o be available and how much does it cost?

OpenAI's GPT-4o will start rolling out its text and image capabilities on May 13. It is free for all users, with paid users benefiting from up to five times the capacity limits.

Teach with Generative AI

Resources for faculty

Bharat Anand, Vice Provost for Advances in Learning , and colleagues at VPAL, the Future of Media Lab, and the Harvard Teaching and Learning Consortium

When it comes to the future of education, virtually no recent technology has sparked as much debate as generative AI (GenAI) and large language models (LLMs). Some have seen this technology as destructive, with school districts from Los Angeles to New York initially banning its use, and others have touted its transformative impact and possibility of changing the game for educators and students alike.

Harvard has consistently tried to embrace new technology across our classrooms, residential and virtual. This past year was no different. In fall 2023, we introduced the Harvard AI Sandbox , enabling faculty to use LLMs in their classrooms while respecting copyright and data security policies.

Several faculty colleagues experimented with these tools in their classrooms during the semester. Here, we have surfaced the learnings from interviews with 30-odd faculty across Harvard, with short 3-7 minute videos, all answering three questions:

• What is the challenge you were trying to address?
• How did you use generative AI tools to tackle it?
• What did you learn?

Our small team of faculty, students, and staff extracted from these a non-comprehensive answer through debate, discussion, and our own learning. I encourage you to do the same, perhaps with your peers or teaching teams, to learn about how these might be relevant for your own courses. Here are some of our initial thoughts:

• GenAI tools have raised concerns about how they may compromise student assessments, promote academic dishonesty, and facilitate “lazy learning.” Our faculty colleagues who experimented with these tools were not oblivious to these concerns; indeed, many share them. At the same time, what you’ll see is how they were also looking to understand how GenAI tools can enhance the educational experience and build more vibrant classrooms.
• Beyond text: For visual aids and images , coding , analysis , games , simulations , and more.
• Prompt design: There’s an old saying: “garbage in, garbage out.” The output of LLMs is only as good as the input, and it’s essential to learn (and perhaps teach) how to write a prompt that works. This is highlighted through discussions on the critical role of deliberate prompt formulation, from having students iterate on their prompts through the course , to engaging students in debate on the ethics of AI use , to making advanced statistical concepts accessible to diverse learners . Our new System Prompt Library offers a range of effective prompts that can be used by educators.
• Interrogate hallucinations : Errors arise not just because of algorithmic or data limitations but, importantly, because LLMs are fundamentally probabilistic. Faculty have found that errors can be reduced through detailed prompt engineering and balanced AI-human partnership .
• Going beyond the simple question-and-answer interface: Sal Khan popularized the idea of using LLMs to ask questions of a student, not just answer them. Several faculty colleagues take this further, illustrating how LLMs can be used to simulate any persona you want, and to ask anything of them. Examples include simulating experts , peers , graders , course designers, and personal chatbots .
• More than the “first draft”: GenAI needn’t compromise student creativity; in fact, it can augment it. Some colleagues are using it to help students refine project prototypes and polish final drafts .
• Work alongside what you already have: Many faculty used LLMs to improve different (and sometimes mundane) aspects of existing teaching and learning “workflows,” such as producing course materials , personalizing feedback , generating assignments , summarizing real-time student responses , and tutoring students .
• Identify, and overcome, hidden or invisible barriers: Students and educators sometimes confront hidden prerequisites that present barriers for teaching and learning. GenAI can assist with overcoming these skill gaps: coding for a business class , foreign languages for research, art skills for building visual aids, and even building games for class engagement.
• Reimagining the classroom : While we’re still in the early days of GenAI, some of these examples already start to surface more profound questions: What does a class with GenAI at its core look like? Ultimately, what is the role of a teacher?
• Questions around genAI’s efficacy on learning arise: can we use GenAI tools—specifically tutor bots—to improve the way students learn? One faculty member created a tutor bot that answered questions like course staff . Beyond such research on students’ interactions with genAI tools, it might be helpful to imagine how it can help you and your students now, in other ways as well: by increasing task efficiency, improving student engagement, increasing their confidence, or even improving learning outcomes.
• The risks of LLMs present valid concerns. While popular debates often focus on “big picture” concerns like algorithmic biases, digital divides, and fake content, some faculty explore the risks at a micro scale , within our classrooms, such as hallucinations, failed reasoning , or superficial thinking, pushing students to understand these issues more deeply.

We hope this starts to give you a taste of the richness embedded in these videos. To start experimenting, request access to the Harvard AI Sandbox . As you think about what’s relevant for your course, support exists across Harvard to help you and your team experiment as well.

Looking ahead, it is clear we will continue to learn. We will experiment. We will refine. And we will advance our understanding. In the next months, we’ll continue to expand our catalog to include more innovations from Harvard faculty—please reach out to the Office of the Vice Provost for Advances in Learning (VPAL) via email if you would like to share your own experiences with colleagues across Harvard. We’re deeply grateful to our colleagues for sharing their experiments and experiences, featured here in the first Harvard GenAI Library for Teaching and Learning .

Frequently asked questions

The following information offers advice for educators interested in using generative AI tools in their teaching and course preparation. As this technology is constantly evolving, this page will be updated frequently with new resources and advice.

How can I use generative AI to help prepare my curriculum?

Creating materials for courses—syllabi, lesson plans, assignments—takes time. Generative AI tools aren’t just useful at broad, general prompts, but, as our colleagues have shown, they are useful in tackling the preparations before a student even arrives in the classroom:

• Preparing to teach: Starting a syllabus or a lesson plan from a blank page is daunting. AI can help you outline your course, create learning objectives, and suggest assignments or in-class activities, while making content that fits your course by feeding it specific reference materials .
• Assignments: Reusing the same assignments across multiple years can be time efficient, but it creates challenges for assessments. Some faculty have explored how AI tools can help write, modify, or create question sets. The more information you put in about the structure and concepts you want it to use, the better it will be . And it can even make a rubric for it.
• Course content: Classes often include large amounts of content for reading, from case studies to readings to slideshow text. Just a few sentences of a prompt using GenAI can generate summaries of relevant readings or videos you might want to include for students, brand new cases to discuss, and even what your slideshow for a lecture should include. Some faculty have gone a step further, inputting all the course materials in the materials that trained a teaching assistant chatbot. Hear what this faculty member learned from students’ interactions with this “faculty copilot.”

How can I use generative AI to help me teach in the classroom, day to day?

How to engage your students in the classroom is an age-old question. Some faculty have used GenAI’s real-time thinking and range of outputs to help offer contemporary solutions, such as a content-generator, a data analyst, or a personal tutorbot.

• Activity leader: Creating interactive classes is easy to advocate for, but hard to do. Some of our faculty have used GenAI to act like a peer student to stimulate critical thinking, perform real-time analysis of student responses to make them feel heard, or even help make games that align with class content ,no coding needed!
• Personal tutor: You (or your TA) can’t be everywhere at once, but GenAI might be able to. Feeding it your syllabus, lectures, and an in-depth prompt can help make a personal tutorbot, generate unlimited practice problems, or even remind students of course-specific information .
• Custom reviewer: LLMs can be used to provide initial personalized feedback to your students, so you can focus on the big picture. Some faculty used them to quickly summarize student responses before office hours, or even point out areas in student responses that need improvement .
• Skill leveler: Classes often have hidden prerequisites: familiarity with coding, texts barred in foreign languages, or even art skills. GenAI tools can be leveraged creatively to help students overcome such barriers, like teaching a business school class about data analysis without requiring every MBA student to learn code, or analyzing trends of thousands of photographs without having to do so manually.

How can I redesign assessments to reduce the misuse of generative AI?

Since LLMs can write essays, respond to readings, and finish problem sets, how can one not be concerned about misuse? For you and your students, addressing this concern means first making sure we know what GenAI can and can’t do—then creating assessments that emphasize skills where AI tools fall short.

Here are some strategies faculty have found useful:

• Human-based learning: Design tasks and assessments that require creativity, practical application of concepts, and critical thinking. For example, instead of asking your students to summarize perspectives on a given issue, you may ask them to critically analyze which perspective is most convincing to them and explain why; to relate their answers to class discussions; or to assess their peers’ performance during a live problem-solving session.
• Process-based assessments: Another approach is to test intermediate steps in the learning process, instead of just the final product. (It’s easy to fudge your report card to your parents; it’s harder to fudge not having gone to school for the past two months.) Testing evidence of original thinking, planning, peer-to-peer conversations, etc. can make relying on genAI less attractive.
• Establishing norms : Emphasize original work and academic honesty, and at a minimum provide clear guidelines about the use of AI-generated content in assignments and assessments.

What are the most important risks of using GenAI tools in my classroom?

Here are the most important ones to keep in mind:

• AI models can make mistakes : We’ve all had an incident (or two!) where a GenAI tool seems to have lost its mind, yielding garbled or entirely made-up answers. These are called AI hallucinations. It’s tempting to think these will get eliminated over time as technology improves. But since LLMs are fundamentally probabilistic rather than deterministic, this may not be the case.
• AI models can be biased: AI adopts the biases of the material and data it was trained on. Good AI use involves being aware of, checking for, and making efforts to correct such biases.
• AI models can violate privacy: AI is very good at doing what you want, but it is also very bad at knowing if what you want it to do is allowed. Personal data is not supposed to be fed to GenAI models. Make sure you are aware of Harvard’s data protection policies and FERPA .
• AI models can be misused : Of course, AI could be used to plagiarize assignments. Unless you are interested in grading robots, you should shift the kind of assignments you are providing students (see above) as well as enforce academic dhonesty policies.

How do I know whether GenAI is even effective for teaching?

GenAI’s ability to meet learners where they are, both in terms of prior knowledge and learning progress, can increase students’ understanding , and AI-powered educational games can increase student motivation and engagement , particularly in STEM courses. One fascinating study showed that when students tutored by AI are pitted against students taught in the traditional classroom setting, the AI-tutored students performed as well or better .

It is natural for instructors, particularly successful ones, to wonder: GenAI may be useful for the average educator. But my classes are great; why would I need it?

One way to think about this is in terms of the efficiency benefits of GenAI tools—they can save time, facilitate meaningful non-classroom learning experiences, and make classroom discussions more interactive. For example:

• Utilize highly thoughtful prompt engineering with GenAI to build a tutorbot that gives students an unlimited number of interactive statistical problems.
• Challenge students with DIY interactive simulation games created on short order—and without any coding prerequisite!
• Empower students to experiment with visualizations of interactions with just a few minutes prompt engineering with DALL-E.

How can I become a GenAI whiz?

• Where to start experimenting : If you’re ready to jump right in, then the best place is to start is the Harvard AI Sandbox . Like any sandbox, it’s a great place to play around with tools; unlike any sandbox, it has five large-language-models to choose from and is accessible by request here .
• Where to modify images: If you’re looking to use GenAI to manipulate and create images, then you can use download Adobe Firefly through the Harvard Adobe Creative Cloud license . For more information, see Getting started with prompts for image-based Generative AI tools .
• Where to learn more : If you want to level up your GenAI knowledge before you start creating, check out the FAS Division of Science Generative AI Resources or the Bok Center’s Artificial Intelligence page.
• Where to find higher level tools : API access to tools like Azure OpenAI, Google Vertex, and AWS services is available by request from HUIT. If you don’t know what these are, that’s okay; teachers have made remarkable tools just using ChatGPT and a well-crafted prompt!

What policies should I keep in mind when using LLMs in my classroom?

When using large language models (LLMs) in your classroom, it’s essential to be aware of the University’s guidelines designed to ensure responsible and effective use of these technologies and to refer to School-specific policies and resources. It is also important to remember that other existing policies, such as Harvard’s Information Security Policy , Digital Accessibility Policy , and Intellectual Property Policy , also apply to GenAI and the use of LLMs.

Across Harvard, there’s a strong emphasis on using LLMs ethically and in ways that uphold academic integrity . The use of generative AI must align with the principles of honesty, respect, and responsibility, ensuring that students’ work remains original and reflective of their understanding and skills. In crafting a response to the use of LLMs in the classroom, it’s crucial to strike a balance between leveraging these tools for educational enhancement and ensuring that they do not compromise educational objectives.

At the course level, Schools within Harvard allow for the customization of AI usage policies by individual instructors , provided these are clearly communicated . Each encourages innovative and thoughtful integration of AI in teaching and learning practices , including learning to use generative AI productively . Above all, students and faculty are encouraged to be transparent about the use of generative AI in academic work. This includes proper attribution when AI-generated content or assistance is utilized in the creation of academic materials. Consider co-creating course-specific norms around the use of generative AI with your students.

For any faculty who wish to simply explore the use of these tools, we encourage you to engage with peers and organizations at Harvard that are interested in these topics.

School-based resources

Visit your School’s website for the latest policies and guidance around using GenAI in the classroom. This list will be updated as further School-specific guidance becomes available.

• Harvard Business School
• Harvard College
• Harvard Division of Continuing Education
• Harvard Graduate School of Design
• Harvard Kenneth C. Griffin Graduate School of Arts and Sciences
• Harvard John A. Paulson School of Engineering and Applied Sciences
• Harvard Kennedy School
• Harvard Medical School