On Tuesday this week, OpenAI released the much-anticipated upgrade of the current AI language model (GPT-3.5) which their eloquent human-like chatbot, ChatGPT, runs on: GPT-4.
Although OpenAI has explained that, in casual conversation, the difference between ChatGPT’s GPT-3.5 and the new GPT-4 language model, can be subtle, there are in fact marked differences between the two.
OpenAI describes GPT-4 as “the latest milestone in OpenAI’s effort in scaling up deep learning.”
“While less capable than humans in many real-world scenarios, [GPT-4] exhibits human-level performance on various professional and academic benchmarks,” states OpenAI.
What’s more, OpenAI says they spent six months “making GPT-4 safer and more aligned,” stating:
“GPT-4 is 82% less likely to respond to requests for disallowed content and 40% more likely to produce factual responses than GPT-3.5 on our internal evaluations.”
When reading through GPT-4’s many sophisticated “state-of-the-art” capabilities, I couldn’t help but be reminded of the plots of many blockbuster movies. Now, I’m not saying there is any tangible link at all between any of these movies and the reality of GPT-4’s capabilties, but these parallels may help paint a metaphorical picture.
So, what can GPT-4 do?
— OpenAI (@OpenAI) March 14, 2023
To clarify, throughout this article, when writing “ChatGPT-3.5,” I am referring to the current open access version of ChatGPT which runs on OpenAI’s GPT-3.5 language model; the predecessor to GPT-4.
Her: GPT-4 is an even better creative writing companion
ChatGPT-3.5’s ability to generate eloquent, human-like prose, poems, short stories and articles – in mere seconds – triggered some concerns amongst members of the writing community.
But rather than as a challenger, perhaps we should view GPT-4 and its writing abilities which exceed that of GPT-3.5, as an advanced writing companion rather than a challenger.
Could we perhaps think of GPT-4 like Joaquin Phoenix’s pocket AI writing buddy in “Her”?
Learning a user’s writing style, composing songs and writing screenplays are just a soupçon of GPT-4’s advanced writing capabilities.
“GPT-4 is more creative and collaborative than ever before,” states OpenAI, “it can generate, edit, and iterate with users on creative and technical writing tasks.”
6/ Creative writing prompts
Looking to break through writer's block?
GPT-4 can suggest ideas, character development, or plot twists.
I find it's a great way to unleash your creativity and allow the story to unfold—take this example of extending the Harry Potter saga. pic.twitter.com/PMMHVWNXWb
— Alex Banks (@thealexbanks) March 16, 2023
Memento: GPT-4 has an expanded contextual memory
ChatGPT-3.5’s has the ability to hold up a dynamic conversation; not only can it remember what a user has previously said, but it can draw upon this and expand on it.
But similar “Memento’s” main character, Leonard (who has a 15-minute memory span) there is a limit to how far back in any given conversation’s history that ChatGPT-3.5 is able to remember what a user has said. This window is only around 3000 words back.
Not only will GPT-4 therefore be able to have “extended conversations” and generate longer responses, but it will also be able to navigate searching through and analysis of large quantities of text in documents.
The Sixth Sense: GPT-4 sees things in images you might not
You see half a carton of eggs, flour, butter and milk and think – ingredients? Maybe pancakes?
GPT-4 “sees” the same ingredients and “thinks” frittata, french toast or quiche.
Until now, ChatGPT has been limited to text input only. But now, with the multimodal capabilities of GPT-4, you can, for example, take a picture of the contents of your fridge, input it as a visual prompt and the fourth generation language model can make handy recipe suggestions.
OK, it’s not “I see dead people” (hopefully), but GPT-4’s ability to see, understand and analyse visual inputs is impressive.
GPT-4 is even capable of taking a rough sketch of a website and turning it into the real thing!
I just watched GPT-4 turn a hand-drawn sketch into a functional website.
This is insane. pic.twitter.com/P5nSjrk7Wn
— Rowan Cheung (@rowancheung) March 14, 2023
Good Will Hunting: GPT-4 beats most people at graduate-level exams
ChatGPT-3.5 scored in the 10th and 31st percentiles of the Uniform Bar Exam and Biology Olympiad, respectively.
GPT-4, on the other hand, outperformed ChatGPT-3.5 as well as the majority of test-takers for both of these famously gruelling examinations, scoring higher than 90% of lawyers in the bar and higher than 99% of those taking the Olympiad.
The upgraded AI model is now able to solve graduate-level problems without a formal education, similar to Will Hunting in the movie.
Full set of GPT-4 benchmarks.
Maybe it should start writing the tests, too. pic.twitter.com/fJvrmI04br
— AI Breakfast (@AiBreakfast) March 15, 2023
Lost in Translation: GPT-4 has extended language skills
Rather than the plot of “Lost in Translation,” the parallel in this case relates to the title.
It’s true that computational coding languages are mostly based on English, but wouldn’t it be better if a conversational AI model could speak more than one language? Monolingualism possibly limits a chatbot’s potential across a large proportion of the world.
GPT-4 was able to outperform ChatGPT-3.5’s English-language performance in 24 out of the 26 languages tested.
The Icelandic government is also making use of GPT-4’s linguistic capabilities.
Split: GPT-4’s personality is steerable
By priming GPT-4 with a “system” message – for example, “You are a XXX who always responds in the XXX style,” – users are able to adjust the style of the AI within reason, in comparison to ChatGPT-3.5’s personality which OpenAI say has “fixed verbosity, tone, and style.”
Similar to the main character in the psychological thriller, “Split,” different “personalities” could be possible with GPT-4. However, along with positive personalities, there could possibly be the potential for negative ones if misused.
OpenAI states, “system messages are the easiest way to ‘jailbreak’ the current model,” but they are making a concerted effort to eradicate and circumvent these kinds of issues as much as possible, amping up the guardrails significantly.
A Beautiful Mind: Though brilliant, GPT-4 still has limitations
Similar to the main character in “A Beautiful Mind,” GPT-4 is brilliant but still has its limitations.
OpenAI has stated that GPT-4 “surpasses ChatGPT in its advanced reasoning capabilities” and that it “can solve difficult problems with greater accuracy, thanks to its broader general knowledge and problem solving abilities.”
Related Articles: ‘Journalists Beware: ChatGPT Is Coming for Your Job,’ Warns Chatbot | ChatGPT Poised for Upgrade: It Could Become as Complex as the Human Brain | Google vs. ChatGPT: Who Really Knows Best? | ‘I am a Machine, With no Soul or Heart’: An Interview With Artificial Intelligence | Microsoft vs. Google: The Tortoise and the Hare of the Tech World? | The ChatGPT ‘Cookbook’: Simple Recipes for Personal and Business Success
However, along with these sophisticated advancements, OpenAI has also mentioned that GPT-4 still has limitations such as “social biases, hallucinations, and adversarial prompts” which the company is working to address.
Everything Everywhere All at Once: GPT-4 is already everywhere
One answer to the question: “What is GPT-4 capable of?” figuratively speaking, could be just like the movie title: Everything, everywhere, all at once.
Obviously this is an exaggeration; there are of course still some limitations to GPT-4’s abilities and there are only a finite amount of use cases for the model at present, but OpenAI is already proving GPT-4’s value in integrating the model into the platforms of a diverse panel of companies and organizations.
In the finance sector, both Stripe and Morgan Stanley have deployed GPT-4. The former uses GPT-4 to make its platform more user-friendly and combat fraud by flagging suspicious account activity, with the latter using the model to better navigate its extensive expert knowledge bases.
Khan Academy is using GPT-4 to power a “virtual tutor,” as is Duolingo, incorporating GPT-4’s vast language abilities into its platform to provide an AI conversational partner for language practice, as well as a feature called “Explain My Answer,” providing learners with guidance on where they went wrong when making a mistake.
Be My Eyes is a Danish startup which is utilizing GPT-4’s image analysis capabilities to develop a “Virtual Volunteer” and “transform visual accessibility,” helping blind or low-vision individuals navigate daily life easier.
“This is a fantastic development for humanity,” says Michael Buckley, Be My Eyes CEO.
We are thrilled to present Virtual Volunteer™, a digital visual assistant powered by @OpenAI’s GPT-4 language model. Virtual Volunteer will answer any question about an image and provide instantaneous visual assistance in real-time within the app. #Accessibility #Inclusion #CSUN pic.twitter.com/IxDCVfriGX
— Be My Eyes (@BeMyEyes) March 14, 2023
“Shaping the future of technology”
As a company, OpenAI’s mission is to develop artificial general intelligence (AGI) which benefits humanity as a whole, and they’re dedicated to ensuring their models are as aligned with human intentions and values as possible.
“We’re excited to see how people use GPT-4 as we work towards developing technologies that empower everyone,” states OpenAI, with an employee also stating in GPT-4’s promo video:
“When we release a model, we know things are not done. We know we have to learn. We know we have to update. We know we have to keep improving all the systems to make it suitable for society.”
So, what is GPT-4?
Perhaps it could be said that it’s a tool, a chameleon and a companion, one which, along with all their other groundbreaking research, OpenAI hopes will help shape the future of technology.
Use GPT-4 with care, but use it to explore and progress we must!
excited 4 today pic.twitter.com/wKMRZID2Qg
— Sam Altman (@sama) March 14, 2023
Correction: This article has been updated since publication to amend the subtitle to clarify the differences between GPT-3.5 and GPT-4; rework the introduction including the removal a book reference and addition of quote on GPT-4’s safety; amend the “Lost in Translation” section to clarify the parallel drawn, GPT-4’s linguistic capabilities, and the specifics of the collaboration with the Icelandic Government; amend the “Split” section to clarify GPT-4’s steerabiltiy to avoid misunderstanding; amend the “A Beautiful Mind” section to clarify the parallel drawn; remove the “Terminator 2,” “2001: A Space Odyssey” and “Ex Machina” sections to avoid misunderstanding and unfair assumption.
Editor’s Note: The opinions expressed here by the authors are their own, not those of Impakter.com — In the Featured Photo: Artwork by Tim West depicting a large language model. Featured Photo Credit: DeepMind / Tim West