It seems as though large language models (LLMs) exploded into public awareness almost overnight. Relatively few people had heard of GPT-2, but I would venture to guess relatively few people haven’t heard of ChatGPT.

But like most things, language models have a history. And, in addition to being outrageously interesting, that history can help us reason about the progress in LLMs, as well as their likely future impacts.

Let’s get started!

A Brief History of Artificial Intelligence Development

The human fascination with building artificial beings capable of thought and action goes back a long way. Writing in roughly the 8th century BCE, Homer recounts tales of the god Hephaestus outsourcing repetitive manual tasks to automated bellows and working alongside robot-like “attendants” that were “…golden, and in appearance like living young women.”

No mere adornments, these handmaidens were described as having “intelligence in their hearts” and stirring “nimbly in support of their master” because “from the immortal gods they have learned how to do things.”

Some 500 years later, mathematicians in Alexandria would produce treatises on creating mechanical servants and various kinds of automata. Heron wrote a technical manual for producing a mechanical shrine and an automated theatre whose figurines could be activated to stage a full tragic play through an intricate system of cords and axles.

Nor is it only ancient Greece that tells similar tales. Jewish legends speak of the Golem, a being made of clay and imbued with life and agency through the use of language. The word “abracadabra”, in fact, comes from the Aramaic phrase avra k’davra, which translates to “I create as I speak.”

Through the ages, these old ideas have found new expression in stories such as “The Sorcerer’s Apprentice”, Mary Shelley’s “Frankenstein”, and Karel Čapek’s “R.U.R.”, a science fiction play that features the first recorded use of the word “robot”.

From Science Fiction to Science Fact

But they remained purely fiction until the early 20th Century, when advances in the theory of computation, as well as the development of primitive computers, began to offer a path toward actually building intelligent systems.

Arguably, the field of artificial intelligence really began in earnest with the 1950 publication of Alan Turing’s “Computing Machinery and Intelligence” – in which he proposed the famous “Turing test” – and with the 1956 Dartmouth conference on AI, organized by luminaries John McCarthy and Marvin Minsky.

People began taking AI seriously. Over the next ~50 years, there were numerous periods of hype and exuberance in which major advances were made, as well as long stretches, known as “AI winters”, in which funding dried up and little was accomplished.

Neural networks and the deep learning revolution are two advances that are particularly important for understanding how large language models have evolved over time, so it’s to these that we now turn.

Neural Networks And The Deep Learning Revolution

The groundwork for future LLM systems was laid by Walter Pitts and Warren McCulloch in the early 1940s. Inspired by the burgeoning study of the human brain, they wondered if it would be possible to build an artificial neuron that had the same basic properties as a biological one, i.e. it would activate and fire once a certain critical threshold had been crossed.

They were successful, though several other breakthroughs would be required before artificial neurons could be arranged into systems that were capable of doing useful work. One such breakthrough was backpropagation, the basic algorithm that is still used to train deep learning systems. Backpropagation was developed in 1960, and it uses the errors in a model’s outputs to iteratively adjust its internal parameters.

It wasn’t until 1985, however, that David Rumelhart, Ronald Williams, and Geoff Hinton used backpropagation in neural networks, and in 1989, this allowed Yann LeCun to train a convolutional neural network to recognize handwritten digits.

This was not the only architectural improvement that came out of this period. Especially noteworthy were the long short-term memory (LSTM) networks that were introduced in 1997 by Sepp Hochreiter and Jürgen Schmidhuber, which made it possible to learn more complex functions.

With these advances, it was clear that neural networks could be trained to do useful work, and that they were poised to do so. All that was left was to gather the missing piece: data.

The Big Data Era

Neural networks and deep-learning applications tend to be extremely data-hungry, and access to quality training data has always been a major bottleneck. In 2009 Stanford’s Fei-Fei Li sought to change this by releasing Imagenet, a database of over 14 million labeled images that could be used for free by researchers. The increase in available data, together with substantial improvements in computer hardware like graphical processing units (GPUs), meant that at long last the promise of deep learning could begin to be fulfilled.

And it was. In 2011 a convolutional neural network called “AlexNet” won multiple international competitions for image recognition, IBM’s Watson system beat several Jeopardy! all-stars in a real game, and Apple launched Siri. Amazon’s Alexa followed in 2014, and from 2015 to 2017 DeepMind’s AlphaGo shocked the world by utterly dominating the best human Go players.

Substantial strides were made in language models. In 2018 Google introduced its Bidirectional Encoder Representations from Transformers (BERT), a pre-trained model capable of a wide array of tasks, like text summarization, translation, and sentiment analysis.

One Model To Rule Them All

It would be easy to miss the significance of AlexNet’s performance on the ImageNet competition or BERT’s usefulness across multiple tasks. For a long time, it was anyone’s guess as to whether it would be possible to train a single large model on a dataset and use it for a range of purposes, or whether it would be necessary to train a multitude of models for each application.

From 2011 onwards, it has become clear that large, general-purpose models are often the best way to go. This point has only become more reinforced, with the success of GPT-4 in everything from brainstorming scientific hypotheses to handling customer service tasks.

Contact Us

How Has Large Language Model Performance Improved?

Now that we’ve discussed this history, we’re well-placed to understand why LLMs and generative AI have ignited so much controversy. People have been mulling over the promise (and peril) of thinking machines for literally thousands of years. After all that time it looks like they might be here, at long last.

But what, exactly, has people so excited? What is it that advanced AI tools are doing that has captured the popular imagination? In the following sections, we’ll talk about the astonishing (and astonishingly rapid) improvements that have been seen in language models in just a few short years.

Getting To Human-Level

One of the more surprising things about LLMs such as ChatGPT is just how good they are at so many different things. LLMs are trained with a technique known as “self-supervised learning”. They take random samples of the text data they’re given, and they try to predict what words come next given the words that came before.

Suppose the model sees the famous opening lines of Leo Tolstoy’s Ann Karenina: “Happy families are all alike; unhappy families are all unhappy in their own way.” What the model is trying to do is learn a function that will allow it to predict “in their own way” from “Happy families are all alike; unhappy families are all unhappy ___”.

The modern crop of LLMs can do this incredibly well, but what is remarkable is just how far this gets you. People are using generative AI to help them write poems, business plans, and code, create recipes based on the ingredients in their fridges, and answer customer questions.

Emergence in Language Models

Perhaps even more interesting, however, is the phenomenon of “emergence” in language models. When researchers tested LLMs on a wide variety of tasks meant to be especially challenging to these models – things like identifying a movie given a string of emojis or finding legal chess moves – they found that in about 5% of tasks, there is a sudden, sharp increase in ability on a given task once a model reaches a certain size.

At present, it’s not really clear how we should think about emergence. One hypothesis for emergence is that a big enough model is able to learn some general piece of knowledge not attainable by a smaller cousin, while another, more prosaic one is that it’s a relatively straightforward consequence of the model’s internal statistical machinery.

What’s more, it’s difficult to pin down the conditions required for emergence in language models. Though it generally appears to be a function of model size, there are cases in which the same abilities can be achieved with smaller models, or with models trained on very high-quality data, and emergence shows up at different scales for different models and tasks.

Whatever ends up being the case, it’s clear that this is a promising direction for future research. Much more work needs to be done to understand how precisely LLMs accomplish what they accomplish. This will not only redound upon the question of emergence, it will also inform the ongoing efforts to make language models safer and less biased.

The GPT Series

The big recent news in AI has, of course, been ChatGPT. ChatGPT has proven useful in an astonishingly-wide variety of use cases and is among the first powerful systems to have been made widely available to the public.

ChatGPT is part of a broader series of GPT models built by OpenAI. “GPT” stands for “generative pre-trained transformer”, and the first of its kind was developed back in 2018. New models and major updates have been released at a rapid clip ever since, culminating with GPT-4 coming out in March of 2023.

At present, OpenAI’s CEO Sam Altman has claimed that there are no current plans to train a successor GPT-5 model, but there are other companies, like DeepMind, who could plausibly build a competitor.

What’s Next For Large Language Models?

Given their flexibility and power, LLMs are finding use across a wide variety of industries, from software engineering to medicine to customer service.

If your interest has been piqued and you’d like to talk to an expert at Quiq about incorporating it into your business, reach out to us to schedule a demo!

Request A Demo

The release of ChatGPT was one of the first times an extremely powerful AI system was broadly available, and it has ignited a firestorm of controversy and conversation.

Proponents believe current and future AI tools will revolutionize productivity in almost every domain.

Skeptics wonder whether advanced systems like GPT-4 will even end up being all that useful.

And a third group believes they’re the first sparks of artificial general intelligence and could be as transformative for life on Earth as the emergence of homo sapiens.

Frankly, it’s enough to make a person’s head spin. One of the difficulties in making sense of this rapidly-evolving space is the fact that many terms, like “generative AI” and “large language models” (LLMs), are thrown around very casually.

In this piece, our goal is to disambiguate these two terms by discussing ​​the differences between generative AI vs. large language models. Whether you’re pondering deep questions about the nature of machine intelligence, or just trying to decide whether the time is right to use conversational AI in customer-facing applications, this context will help.

Let’s get going!

What Is Generative AI?

Of the two terms, “generative AI” is broader, referring to any machine learning model capable of dynamically creating output after it has been trained.

This ability to generate complex forms of output, like sonnets or code, is what distinguishes generative AI from linear regression, k-means clustering, or other types of machine learning.

Besides being much simpler, these models can only “generate” output in the sense that they can make a prediction on a new data point.

Once a linear regression model has been trained to predict test scores based on number of hours studied, for example, it can generate a new prediction when you feed it the hours a new student spent studying.

But you couldn’t use prompt engineering to have it help you brainstorm the way these two values are connected, which you can do with ChatGPT.

There are many types of generative AI, so let’s spend a few minutes discussing the major categories: image generation, music generation, code generation, and a few others.

How Is Generative AI Used To Make Images?

One of the first “wow” moments in generative AI came fairly recently when it was discovered that tools like Midjourney, DALL-E, and Stable Diffusion could create absolutely stunning images based on simple prompts like:

“Old man in a book store, ambient dappled sunlight, sedate, calm, close-up portrait.”

Depending on the wording you use, these images might be whimsical and futuristic, they might look like paintings from world-class artists, or they might look so photo-realistic you’d be convinced they’re about to start talking.

Each of these tools is suited to specific applications. Midjourney seems to be best at capturing different artistic approaches and generating images that accurately capture an aesthetic. DALL-E tends to do better at depicting human figures, including faces and eyes. Stable Diffusion seems to do well at generating highly-detailed outputs, capturing subtleties like the way light reflects on a rain-soaked street.

(Note: these are all general impressions, it’s difficult to know how the tools will compare on any specific prompt.)

Broadly, this is known as “image synthesis”. And since we’re talking specifically about making images from text, this sub-domain is known as “text-to-image.”

A variant of this technique is text-to-video (alternatively: “text-to-4d”), which produces short clips or scenes based on text prompts. While text-to-video is still much more primitive than text-to-image, it will get better very quickly if recent progress in AI is any guide.

One interesting wrinkle in this story is that generative algorithms have generated something else along with images and animations: legal battles.

Earlier this year, Getty Images filed a lawsuit against the creators of Stable Diffusion, alleging that they trained their algorithm on millions of images from the Getty collection without getting permission first or compensating Getty in any way.

This has raised many profound questions about data rights, privacy, and how (or whether) people should be paid when their work is used to train a model that might eventually automate them out of a job.

We’re still in the early days of grappling with these issues, but they’re sure to make for fascinating case law in the years ahead.

How Is Generative AI Used To Make Music?

Given how successful advanced models have been in generating text (more on that shortly), it’s only natural to wonder whether similar models could also prove useful in generating music.

This is especially true because, on the surface, text and music share many obvious similarities (both are sequential, for example.) It would make sense, therefore, that the technical advances that have allowed coherent text production might also allow for coherent music production.

And they have! There are now a number of different tools, such as MusicLM, which are able to generate fairly high-quality audio tracks from prompts like:

“The main soundtrack of an arcade game. It is fast-paced and upbeat, with a catchy electric guitar riff. The music is repetitive and easy to remember, but with unexpected sounds, like cymbal crashes or drum rolls.”

As with using generative AI in images, creating artificial musical tracks in the style of popular artists has already sparked legal controversies. A particularly memorable example occurred just recently when a TikTok user supposedly created an AI-generated collaboration between Drake and The Weeknd, which then promptly went viral.

The track was removed from all major streaming services in response to backlash from artists and record labels, but it’s clear that ai music generators are going to change the way art is created in a major way.

How Is Generative AI Used For Coding?

It’s long been the dream of both programmers and non-programmers to simply be able to provide a computer with natural-language instructions (“build me a cool website”) and have the machine handle the rest. It would be hard to overstate the explosion in creativity and productivity this would initiate.

With the advent of code-generation models such as Replit’s Ghostwriter and GitHub Copilot, we’ve taken one more step towards that halcyon world.

As is the case with other generative models, code-generation tools are usually trained on massive amounts of data, after which point they’re able to take simple prompts and produce code from them.

You might ask it to write a function that converts between several different coordinate systems, create a web app that measures BMI, or translate from Python to Javascript.

As things stand now, the code is often incomplete in small ways. It might produce a function that takes an argument as input that is never used, for example, or which lacks a return function. Still, it is remarkable what has already been accomplished.

There are now software developers who are using models like ChatGPT all day long to automate substantial portions of their work, to understand new codebases with which they’re unfamiliar, or to write comments and unit tests.

Contact Us

What Are Large Language Models?

Now that we’ve covered generative AI, let’s turn our attention to large language models (LLMs).

LLMs are a particular type of generative AI.

Unlike with MusicLM or DALL-E, LLMs are trained on textual data and then used to output new text, whether that be a sales email or an ongoing dialogue with a customer.

(A technical note: though people are mostly using GPT-4 for text generation, it is an example of a “multimodal” LLM because it has also been trained on images. According to OpenAI’s documentation, image input functionality is currently being tested, and is expected to roll out to the broader public soon.)

What Are Examples of Large Language Models?

By far the most well-known example of an LLM is OpenAI’s “GPT” series, the latest of which is GPT-4. The acronym “GPT” stands for “Generative Pre-Trained Transformer”, and it hints at many underlying details about the model.

GPT models are based on the transformer architecture, for example, and they are pre-trained on a huge corpus of textual data taken predominately from the internet.

GPT, however, is not the only example of an LLM.

The BigScience Large Open-science Open-access Multilingual Language Model – known more commonly by its mercifully-short nickname, “BLOOM” – was built by more than 1,000 AI researchers as an open-source alternative to GPT.

BLOOM is capable of generating text in almost 50 natural languages, and more than a dozen programming languages. Being open-sourced means that its code is freely available, and no doubt there will be many who experiment with it in the future.

In March, Google announced Bard, a generative language model built atop its Language Model for Dialogue Applications (LaMDA) transformer technology.

As with ChatGPT, Bard is able to work across a wide variety of different domains, offering help with planning baby showers, explaining scientific concepts to children, or helping you make lunch based on what you already have in your fridge.

How Are Large Language Models Trained?

A full discussion of how large language models are trained is beyond the scope of this piece, but it’s easy enough to get a high-level view of the process. In essence, an LLM like GPT-4 is fed a huge amount of textual data from the internet. It then samples this dataset and learns to predict what words will follow given what words it has already seen.

At first, its performance will be terrible, but over time it will learn that a sentence like “I sat down on the _____” probably ends with a word like “floor” or “chair”, and probably not a word like “cactus” (at least, we hope you’re not sitting down on a cactus!)

When a model has been trained for long enough on a large enough dataset, you get the remarkable performance seen with tools like ChatGPT.

Is ChatGPT A Large Language Model?

Speaking of ChatGPT, you might be wondering whether it’s a large language model. ChatGPT is a special-purpose application built on top of GPT-3, which is a large language model. GPT-3 was fine-tuned to be especially good at conversational dialogue, and the result is ChatGPT.

Are All Large Language Models Generative AI?

Yes. To the best of our knowledge, all existing large language models are generative AI. “Generative AI” is an umbrella term for algorithms that generate novel output, and the current set of models is built for that purpose.

Utilizing Generative AI In Your Business

Though truly powerful generative AI language models are less than a year old, they’re already being integrated into numerous business applications. Quiq Compose, for example, is able to study past interactions with customers to better tailor its future conversations to their particular needs.

From generating fake viral rap songs to generating photos that are hard to distinguish from real life, these powerful tools have already proven that they can dramatically speed up marketing, software development, and many other crucial business functions.

If you’re an enterprise wondering how you can use advanced AI technologies such as generative AI language models for applications like customer service, schedule a demo to see what the Quiq platform can offer you!