Editor’s Note: Jeff is almost ready to launch his first brand-new research service in three years…
In just a couple of days, he’s unveiling his new Deep Access advisory. It’s named for the proprietary, real-time dataset the Deep Access team has exclusive access to.
They’ve used this dataset to develop a deep-learning technology focused on spotting market swings to help predict – and potentially profit from – everyday volatility in equities.
This kind of access is typically reserved for well-funded hedge funds and investment banks.
You can learn more about Jeff’s new Deep Access advisory this Wednesday, November 20, at 8 p.m. ET, clicking here to automatically sign up to attend.
One of the most underappreciated applications of artificial intelligence (AI) is the impact it will soon have on education.
Many would argue it is already having an impact.
Most 12-year-olds with access to a computer already know that ChatGPT and other chatbots are a great shortcut to finding a clear explanation to a question, drafting an essay, or providing a step-by-step solution to a problem.
But that’s not the future of education.
Querying a chatbot as we would Google search may be a great timesaver, a productivity hack, and oftentimes useful for learning a subject material quickly…
But the real potential is so much greater.
Irrespective of where we grew up around the world, our construct for education – kindergarten through university – is largely the same. It’s quite amazing how universal it is. One teacher – per subject – teaching a room full of students.
And the largest point of differentiation with regard to a quality education tends to be the student-to-teacher ratio. The general rule of thumb is that the larger the student-to-teacher ratio, the less effective the learning of the students.
This is also universal.
What we tend to see around the world is that most developed countries have lower student-to-teacher ratios compared to developing countries. It’s not always the case, but it is logical as lower student-to-teacher ratios always require a corresponding increase in budget to support a larger number of teachers.
Across all Organisation for Economic Co-operation and Development (OECD) countries, the average school class has 21 students in primary education. That’s a lot of students for a single teacher.
And in university, especially in freshman-year classes, it’s not unusual to have a hundred or more students in some classes. That was certainly my experience in many of my “101” classes.
Average Class Size by Country (2019) | Source: OECD
The issue is that the higher the student-to-teacher ratio, the more generalized the teaching becomes. The problem is that no two students learn in exactly the same way. And that means on a student-population level, the learning outcomes are far from optimized.
Generally speaking, this is the promise of private schools, which, on average, have much lower student-to-teacher ratios. We can see this in the OECD chart above, with the private schools represented by the light green circles.
And the better the private schools, the lower that ratio gets. For example, in the U.S., private high schools tend to have a student-to-teacher ratio between 12:1 and 6:1. And the very best private schools can be as low as 5:1.
The impact of these lower ratios is well known. That’s why private schools are so desirable:
There are, of course, two major problems with the above.
The first problem is very obvious. Access to private schools is very limited.
There are only so many “seats” at the best private schools, and they are extremely expensive. Access is limited to those who can afford private schools for their children and those lucky enough to receive financial aid or scholarships to attend.
It’s not scalable, it’s only accessible to a small portion of the population, and it can’t be democratized.
The other major problem at a global scale is there simply aren’t enough teachers. Earlier this year, the United Nations Educational, Scientific, and Cultural Organization (UNESCO) published a report identifying that the world needs 44 million teachers by 2030 to meet its Sustainable Development Goals.
We can interpret the report simply as an example of the magnitude of the teacher shortage problem on a global level. The magnitude is so large that it simply can’t be solved by just “hiring more teachers.” There aren’t enough to hire.
This trend was greatly exacerbated by the pandemic here in the U.S. and other parts of the developed world. Fewer and fewer qualified teachers have been willing to work under current conditions and pay for what is the norm in the U.S.
And no matter where we live, this issue touches us all. After all, if we can better educate our children on a global basis, economic outcomes, global GDP, quality of life, and human longevity will all increase.
But back to my early point, the traditional way of teaching just doesn’t scale well. And it’s generalized education as opposed to personalized education.
Of course, this is where artificial intelligence is such a profound technology.
As the industry gets closer and closer to artificial general intelligence (AGI), the process of learning will be wildly enhanced. So much so that the student-to-teacher ratio drops to 1:1.
Not only will students be empowered to receive one-on-one instruction for any subject matter at all, but they can also receive the education in a way that is optimized for their individual learning styles.
Each student’s learning can be customized to best suit learning style, but also based on subject matter.
This is such a powerful idea, and it will democratize education around the world. It will be equivalent to gaining a private school education at a public school education cost. No longer will the best education be limited to those who can afford private schools. And the world will benefit greatly from this change.
Now, I know what some may be thinking… “Doing this, even with AI, will be too expensive won’t it?” This is where I think you’ll be surprised.
The reality is that the cost of AI inference – that’s the running of an AI application – is dropped by a magnitude every year. By next year, the cost of having a personalized AI, or a personalized AI tutor, will be less than one dollar… a month.
To be clear, there will be different companies with different AI tutor product offerings and different price points, but the raw cost of running the AI per student will be less than a dollar. Said another way, this technology will be accessible to just about anyone on the planet that wants to learn.
And to get there, there are a few major things that need to happen to make this future of education a reality…
And all of these developments are underway.
The first development that must happen is that these large language models (LLMs) need to be trained on highly verified, fact-based data sets, to be effective tutors. After all, we’re in big trouble if our AI tutor teaches us that 2+2 equals something other than 4. Or if it ignores, for example, a biological truth of homo sapiens that we are either born with a prostate or a uterus – there are only two genders.
xAI is – not surprisingly – going to great lengths to develop its “maximum truth-seeking AGI.” xAI is widely discounted by the industry as a latecomer, which is a mistake. xAI is in the process of raising $6 billion at a $50 billion valuation. The majority of those funds will be used to acquire another 100,000 NVIDIA GPUs… and double the capacity of its AI supercomputer.
xAI has also been aggressively building out what it calls its Reverse University. This is an initiative to hire humans as AI tutors – you read that correctly, it’s humans who tutor/teach the AIs. These humans “need to be great writers, great fact-checkers, have a high attention to detail, and a great intuition for what makes for a good AI response.”
At the moment, xAI is looking for:
If you or anyone you know is interested, you can start by going here to check it out.
The idea is to hire full-time AI tutors to help curate information, annotate information, and teach AIs how to best teach humans. That’s why xAI calls it “reverse university.”
Most companies have outsourced this kind of work, which can lead to lower-quality data. The data inputs and what we teach AIs are critically important. When humans code AIs with bias or factually untrue information, it can, and will, lead to horrible outcomes.
That’s why efforts to curate the highest quality data set are so critical. As I’ve written before, garbage in garbage out. The better our inputs, the better the outcomes.
The second technology that needs to happen is the continued development of multi-modal artificial intelligence.
Students of all ages need to be able to show an AI objects or images in the real world. Our AI tutors need to be able to ingest voice, video, text, audio, images, code, and the real world in real time. This will enable the AI tutors to interact with students in the same way that a human teacher can do.
As the AIs’ knowledge base grows and becomes more verified, student outcomes will continue to improve, which leads us to our third development.
The third development that AI tutors will need to learn is various teaching methods that they can use to personalize education for each student. This is where the magic will happen. It’s where students of any age will be empowered to learn in a way that is best suited to their learning style. And think about it…
If an AI attempts to teach a student a math problem, for example, and notices the student struggling to adopt the method, the AI can simply pivot its approach — trying another method of many.
Perhaps even adopting its own method, which it “learned” by coming to understand how the student learns best.
What teacher do you know who has a dozen ways to teach a concept and can modify it on the fly to the student?
Ironically, this is not just useful in terms of improving learning outcomes. It will also be less expensive to deliver education in this way.
Because it is a far more efficient way to teach and learn, it also means less time spent in the process. And when it comes to running an AI, time is money. Longer computational requirements are equivalent to more expense. More efficient instruction results in less computation.
For those of us old enough, remember how it felt when we were first able to search the world’s knowledge by using an internet search engine. Those trips to the local library which often had limited hours suddenly were no longer necessary.
What’s coming will feel similar. We won’t have to sift through multiple pages from a Google search trying to find the information that we need. Since Google incorporated its Gemini large language model in its search results, the process is dramatically easier; but what comes next will feel like an even bigger improvement.
We’re on the cusp of an entirely new future in education. One that is democratized and accessible to anyone with access to a computer.
Yes, many will still go to private high schools and universities for the brand name, access to the network that it provides, and the social atmosphere.
But many won’t. And yet they’ll be just as educated (and have a fraction of the debt)…
Most will use AI tutors to fill in the gaps. Everyone will have their own “private” tutor to help with any subject imaginable, something that historically has only been available to those who can afford it.
It’s time for democratization. It’s time for the future of education.
Regards,
Jeff
The Bleeding Edge is the only free newsletter that delivers daily insights and information from the high-tech world as well as topics and trends relevant to investments.
The Bleeding Edge is the only free newsletter that delivers daily insights and information from the high-tech world as well as topics and trends relevant to investments.
