“Attention Is All You Need.”
That’s the title of a 2017 paper by eight artificial intelligence (“AI”) researchers.
It gave rise to generative AI models GPT-4, DALL-E, and Google’s Bard – the biggest breakthroughs in AI over the past year.
The authors – most of whom were working at Google at the time – showed a clever way to train AI to reason through a block of text.
They programmed the AI to find the relationship between words to figure out the overall meaning.
If you think about translating between languages, it’s like translating the meaning of a phrase or sentence… instead of word for word.
And it’s the entire reason why AI has taken a huge leap forward in recent years.
This framework is called the transformer model. And it’s the most widely used method for building ChatGPT and other generative AI models.
But while the transformer model is great, it has its limitations.
It struggles with long strings of data.
The best transformer models can’t reason through anything longer than your average book.
That’s why today’s AI models struggle with creating long clips of video… decoding the secrets hidden within our DNA… and even predicting the next move in the stock market.
These require millions… and sometimes billions… of inputs. It’s just too much for the current AI models to handle.
But last December, two researchers published a paper with a model that fixes these weaknesses.
Known as “Mamba,” it is the next leap forward in AI.
The mainstream media hasn’t picked up on the story. But leading AI developers have. And they’re already working on building new AI models with this new technology. So, I want to prepare you now for what’s coming.
That’s because this next leap forward in AI is going to happen much faster than previous ones. And if you’re not paying attention, you’re going to miss out on profiting from it.
Mamba’s Advantage
As I said above, Mamba is a new AI model that solves the issues plaguing transformer-based AIs.
Mamba can take in millions of data points without getting bogged down or confused. That makes it much more capable of tackling problems that have a lot of data.
Below is a chart of Mamba (brown line along the top) outperforming other AI models on large blocks of text data.
Mamba doesn’t show any drop in accuracy up to a sequence length of 1 million units. But as you can see, other models suffer plunges in accuracy around a sequence of just 1,000.
Mamba picks out the most important bits of info to reason through a problem.
It’s a bit like skim reading. You don’t need to read every word. Your eyes can dart to the keywords in each sentence and grasp the meaning of the text.
Think Bigger
Mamba holds the key to using AI to decode secrets locked in our DNA.
The human genome – our complete set of DNA – is made up of about 3 billion “base pairs.”
Each base pair forms a rung on the ladder of our DNA.
The base pairs (adenine, thymine, guanine, and cytosine) make up the genetic instructions of DNA. (Source: National Institute of Health)
Right now, there are thousands of diseases with no known causes.
Once we can turn an AI loose on our DNA, we may be able to find the genetic source of diseases including epilepsy, scoliosis (which causes the spine to curve), and Parkinson’s disease.
And once we know the cause, we can start working on a cure.
That’s far and away the most exciting application of this new kind of AI. But it’s not the only one…
Mamba will work great for creating films.
Today’s AI models are just starting to get decent at creating short clips of video.
But even the best AIs struggle to make anything beyond 10 seconds.
Every second of video contains at least 24 images… all rich with data. So, each additional second adds to the data the AI has to work through.
Mamba should be able to create sequences that are much longer.
It will tackle one of the most difficult challenges for AI today – the stock market.
The global stock market has more than 50,000 stocks trading on any given business day. That’s on top of hundreds of commodities… thousands of bonds… and the constant deluge of news that impacts these markets.
The volume of data is mind-blowing. And AI stock pickers still haven’t bested out their human counterparts.
With more development, Mamba may start to gain an edge.
This is an area I’m especially interested in. Over the coming weeks and months ahead, I’ll be exploring ways to use this new breed of AI model to profit from the stock market.
I’ll share what I find with you in these pages.
The big takeaway is that AI isn’t plateauing, as some folks claim. The biggest breakthroughs are still ahead of us.
It took roughly five years between the time “Attention Is All You Need” was published and when GPT3 went live.
We won’t have to wait that long for an AI model based on Mamba’s breakthroughs to launch to the public.
With even more interest, and money, we could see Mamba AI’s within the next year or two.
In the meantime, it’s not too late to profit from AI. In both my Near Future Report and Exponential Tech Investor advisories, I’ve recently raised the buy ranges on multiple positions in the portfolios.
Recent recommendations are up as much as 50% in less than three months.
And I’ll continue to be on the lookout for opportunities for you to profit from the rise in AIs based on Mamba’s framework.
Regards,
Colin Tedards
Editor, The Bleeding Edge
P.S. As AIs grow more advanced and new models like Mamba become more widely known… It will usher in a new wave of AI… And with it, a new wave of profits for those who know where to look.
That’s why I’ve put together a presentation on this “second wave” of AI that’s set to unfold. You can go here to access it. I’ll even share with you three stocks to buy to get ahead of the wave… And seven you should absolutely avoid.