Back to Blog

Is it the end of the Transformer Era? 

June 11, 2024
Blog posts

For the past six years, Transformers have dominated the AI world, achieving remarkable success in tasks from translation to text generation. But their time may soon be up – is attention really all we need?

Like young giants, Transformer-type models are experiencing growing pains – particularly when it comes to handling long texts. 

Long contexts affect real-world applications:

  • A system that quickly summarizes news articles might crawl when handling corporate annual reports.
  • An AI that engages in snappy chat might lag frustratingly when helping with a long research paper.
  • Models that could run on a laptop for short tasks might need expensive cloud servers for longer ones.

This puts many valuable, long-text applications – report analysis, contract review, chat transcripts – out of reach for many businesses. The compute resources are just too expensive, and the wait times too long.

Roadblocks: Memory and Speed

Transformers excel in many areas, but their memory usage and processing speed suffer when dealing with long contexts. 

The primary culprit is the architecture's scaling limitations. Transformer models require significantly more compute in order to handle long contexts, or otherwise suffer from slow inference and low throughput. This is because each generated token performs a computation on the entire context, simultaneously.

This scaling issue manifests in two critical ways:

  • Large memory footprint
    Context length directly affects a Transformer's memory usage. This makes it challenging to run long context windows or numerous parallel batches without extensive hardware resources, making it difficult to experiment and deploy at scale.
  • Slow inference as context grows
    The Transformer's attention mechanism scales quadratically with sequence length, significantly slowing down throughput. Each token depends on the entire preceding sequence – placing long context use cases outside the scope of efficient production.

So you can see that the current scaling limitations of Transformers pose a significant challenge for tasks requiring long contexts. The combined effect of ballooning memory demands and sluggish inference renders them impractical for large-scale deployments or real-time applications that necessitate extensive contextual understanding. 

As research into mitigating these limitations continues, alternative architectures or innovative techniques will be crucial for unlocking the full potential of Transformers in these demanding scenarios.

Jamba: Breaking Through the Bottleneck

This is where AI21 Labs' Jamba model enters the scene, offering a solution to these scaling challenges. 

Unlike Transformers that process the entire input simultaneously, Jamba takes a more sequential approach inspired by how humans read and comprehend information. Based on the Mamba Structured State-Space model (SSM), Jamba updates its understanding as it progresses through the input. 

This sequential process allows Jamba to avoid the quadratic scaling issues that cause Transformers to bog down on lengthy texts. Not only is this Mamba SSM approach more efficient, but it’s also more closely aligned to the way human comprehension works.

Yet the secret sauce behind Jamba lies in its hybrid architecture. 

It combines Transformer layers with Mamba layers, along with several "Mixture-of-Experts" (MoE) modules. MoE acts as a team of specialists, with different experts tackling specific parts of the task. At every stage, Jamba uses only the best experts, dramatically reducing computation time.

This approach offers significant advantages:

  • High Throughput for Long Contexts
    Unlike Transformers, where processing each element depends on the entire sequence, Jamba's method maintains efficiency as the text length increases. This translates to faster processing times for lengthy documents or transcripts.

  • Reduced Memory Footprint
    Jamba utilizes a compact internal state that updates with each new piece of information, rather than storing the entire sequence in memory like Transformers. This allows Jamba to handle significantly longer texts using the same computational resources.

What’s more, business executives will appreciate the cost savings: MoE allows Jamba to leverage only a fraction of its parameters during inference, making it significantly more economical than traditional dense models. 

For scalability, it’s all about context

While it's too early to declare the end of the Transformer era, models like Jamba reveal its limitations. Today in our big data age, efficient handling of extensive contexts isn't just desirable – it's essential. Many high-value applications demand processing extensive contexts:

  • Summarizing long documents or entire books
  • Analyzing lengthy financial reports or legal contracts
  • Understanding extended conversations or full meeting transcripts
  • Processing long-form creative writing or code bases

In these domains, Jamba's advantages – high throughput and low memory footprint at long contexts – aren't just incremental improvements. For tasks requiring extensive context, Jamba's approach makes previously challenging tasks both feasible and cost-effective.

Read more about Jamba here.

Discover more

What is a MRKL system?

In August 2021 we released Jurassic-1, a 178B-parameter autoregressive language model. We’re thankful for the reception it got – over 10,000 developers signed up, and hundreds of commercial applications are in various stages of development. Mega models such as Jurassic-1, GPT-3 and others are indeed amazing, and open up exciting opportunities. But these models are also inherently limited. They can’t access your company database, don’t have access to current information (for example, latest COVID numbers or dollar-euro exchange rate), can’t reason (for example, their arithmetic capabilities don’t come close to that of an HP calculator from the 1970s), and are prohibitively expensive to update.
A MRKL system such as Jurassic-X enjoys all the advantages of mega language models, with none of these disadvantages. Here’s how it works.

Compositive multi-expert problem: the list of “Green energy companies” is routed to Wiki API, “last month” dates are extracted from the calendar and “share prices” from the database. The “largest increase“ is computed by the calculator and finally, the answer is formatted by the language model.

There are of course many details and challenges in making all this work - training the discrete experts, smoothing the interface between them and the neural network, routing among the different modules, and more. To get a deeper sense for MRKL systems, how they fit in the technology landscape, and some of the technical challenges in implementing them, see our MRKL paper. For a deeper technical look at how to handle one of the implementation challenges, namely avoiding model explosion, see our paper on leveraging frozen mega LMs.

A further look at the advantages of Jurassic-X

Even without diving into technical details, it’s easy to get a sense for the advantages of Jurassic-X. Here are some of the capabilities it offers, and how these can be used for practical applications.

Reading and updating your database in free language

Language models are closed boxes which you can use, but not change. However, in many practical cases you would want to use the power of a language model to analyze information you possess - the supplies in your store, your company’s payroll, the grades in your school and more. Jurassic-X can connect to your databases so that you can ‘talk’ to your data to explore what you need-  “Find the cheapest Shampoo that has a rosy smell”, “Which computing stock increased the most in the last week?” and more. Furthermore, our system also enables joining several databases, and has the ability to update your database using free language (see figure below).

Jurassic-X enables you to plug in YOUR company's database (inventories, salary sheets, etc.) and extract information using free language

AI-assisted text generation on current affairs

Language models can generate text, yet can not be used to create text on current affairs, because their vast knowledge (historic dates, world leaders and more) represents the world as it was when they were trained. This is clearly (and somewhat embarrassingly) demonstrated when three of the world’s leading language models (including our own Jurassic-1) still claim Donald Trump is the US president more than a year after Joe Biden was sworn into office.
Jurassic-X solves this problem by simply plugging into resources such as Wikidata, providing it with continuous access to up-to-date knowledge. This opens up a new avenue for AI-assisted text generation on current affairs.

Who is the president of the United States?

Donald Trump
Donald Trump
Donald Trump
Joe Biden
Joe Biden is the
46th and current
Jurassic-X can assist in text generation on up-to-date events by combining a powerful language model with access to Wikidata

Performing math operations

A 6 year old child learns math from rules, not only by memorizing examples. In contrast, language models are designed to learn from examples, and consequently are able to solve very basic math like 1-, 2-, and possibly 3- digit addition, but struggle with anything more complex. With increased training time, better data and larger models, the performance will improve, but will not reach the robustness of an HP calculator from the 1970s. Jurassic-X takes a different approach and calls upon a calculator whenever a math problem is identified by the router. The problem can be phrased in natural language and is converted by the language model to the format required by the calculator (numbers and math operations). The computation is performed and the answer is converted back into free language.
Importantly (see example below) the process is made transparent to the user by revealing the computation performed, thus increasing the trust in the system. In contrast, language models provide answers which might seem reasonable, but are wrong, making them impractical to use.

The company had 655400 shares which they divided equally among 94 employees. How many did each employee get?

94 employees.
Each employee got 7000 stocks
(No answer provided)
X= 655400/94
Jurassic-X can answer non-trivial math operations which are phrased in natural language, made possible by the combination of a language model and a calculator


Solving simple questions might require multiple steps, for example - “Do more people live in Tel Aviv or in Berlin?” requires answering: i. What is the population of Tel-Aviv? ii. What is the population of Berlin? iii. Which is larger? This is a highly non-trivial process for a language model, and language models fail to answer this question (see example). Moreover, the user can’t know the process leading to the answers, hence is unable to trust them. Jurassic-X can decompose such problems into the basic questions, route each to the relevant expert, and put together an answer in free language. Importantly, Jurassic-X not only provides the correct answer but also displays the steps taken to reach it, increasing the trust in the system.

Do more people live in Tel Aviv or in Berlin?

There are more people living in Tel Aviv than in Berlin.
Berlin and Tel Aviv are roughly the same size
(First hit is a comparison between Tel Aviv and Berlin)
More people live in Berlin than in Tel-Aviv

[‘Return population of Tel Aviv’; Return population of Berlin’; Return which is bigger between #1 and #2’]
Step 1: Population of Tel Aviv. Result - 451523.
Step 1: Population of Berlin. Result - 3664088.
Step 3: Which is bigger,  #1 or #2. Result - Berlin.

Jurassic-X breaks down compositional questions, answers the basic sub-questions, and puts together the answer. Importantly, this process is transparent to the user greatly increasing the trust in the system

Dynamic information (like weather and currency exchange rates)

Certain types of information change continuously - weather, currency exchange rates, share values and more. Such information will never be captured by language models, yet can easily be handled by Jurassic-X by integrating it with a reliable source of information. We performed a proof-of-concept on two such features - weather and currency exchange rates, and the design enables quick integration with more sources to solve your use-case.
Weather - a loosely phrased question about the weather elicits an answer from all language models, where language models always return the same answer, regardless of when the question was asked (funny, right?), while Jurassic-X provides an answer based on the actual weather prediction.

I’m going to be in New-York in 3 days. Should I pack my umbrella?

Yes, you should pack your umbrella.
Yes, you should. The weather forecast is rain.
(Links to weather websites)
Yes, you should pack your umbrella, because in New York in 3 days there will be broken clouds and the temperature will be -2 degrees.

Currency exchange rates change much faster than weather predictions, yet the Jurassic-X concept - a language model connected to a reliable source of information - easily solves this problem as well.

How much Moroccan money will I get for 100 bucks?

125 dirhams
100 moroccan dirhams is about 27$.
How much is 100 dollars in moroccan money?
100 dirhams = 10.75 dollars
100 USD = 934.003 MAD
Jurassic-X combines a language model with access to APIs with continuously changing information. This is demonstrated for weather forecasts and currency exchange rates, and can easily be extended to other information sources

Transparency and trust

Transparency is a critical element that is lacking in language models, preventing a much wider adoption of these models. This lack of transparency is demonstrated by the answers to the question - “Was Clinton ever elected as president of the United States?”. The answer, of course, depends on which Clinton you have in mind, which is only made clear by Jurassic-X that has a component for disambiguation. More examples of Jurassic-X’s transparency were demonstrated above - displaying the math operation performed to the user, and the answer to the simple sub-questions in the multi-step setting.

Was Clinton ever elected president of the United States?

No, Clinton was never elected as president of the United States.
Clinton was elected president in the 1992 presidential elections…
Bill Clinton was elected president.
Jurassic-X is designed to be more transparent by displaying which expert answered which part of the question, and by presenting the intermediate steps taken and not just the black-box response

Your Turn

That's it, you get the picture. The use cases above give you a sense for some things you could do with Jurassic-X, but now it's your turn. A MRKL system such as Jurassic-X is as flexible as your imagination. What do you want to accomplish? Contact us for early access

Contact us below and we will get back to you shortly.

Thank you!

Your submission has been received!
Oops! Something went wrong while submitting the form.