Artificial Intelligence – AI – isn’t really AI, which may be why it’s been so disappointing to companies that aren’t trying to sell you a new leaf blower. Instead of doing something practical like washing the dishes or doing laundry so you have more time to make art or music, the AI is making music and art for you.
Basically, it’s an over-the-top scam.
These are still just Language Learning Models, LLMs, which are simply high-level autocomplete. That Replika girlfriend likes her sports team because you told her which team to like, and she’s storing that information to sell you a new phone to watch more sports and see more ads. If she makes you tell her what medications you’re taking, it’s not because she cares about her health. She has even less conviction than that guy in Ghana who pretends to be her girlfriend in Los Angeles.
AI is a massive disappointment unless you’re a company like Nvidia, because companies are diving into the hype and they need chipsets to do it.
A current disappointment that could be resolved, however, is education run by the government and controlled by the unions. I write ‘disappointment’ because that is the consensus, we see this every two years with torn clothes on international standardization tests, but this is not a criticism of teachers, it is me who is USA today and other places that defend American children when the two pincers of the education unions – “we are underfunded” – and the critics who attack American children for their agenda say they are stupid.(1)
The problem in education is us, and AI/LLMs can really help.
We cannot let teachers be teachers when we demand that schools also carry a lot of social justice water. We cannot allow teachers to be teachers when, for political reasons, a new president launches a successful effort like No Child Left Behind only to replace it with a disaster like Common Core.
What could be allowed to help education and stimulate creativity without a political agenda is to use LLMs to step outside the box. There will be some fear of this, in the same way that people are afraid of self-driving cars if someone dies, but this is perpetuating the “no defects” mentality that turns cultures from creative engines to engines that can only be slow and incremental.(3) Since teachers are government union employees and their unions hold an alarming amount of power in a political party, there will be concerns that AI could take over jobs, but as we have seen with remote education during the COVID-19 pandemic, the Online teaching of children will only work with a select few. And the reason we have government-controlled education is so that every child has a foundation.
AI is like a bad Monte Carlo simulation – it can declare complete confidence in a wrong answer
Some recent work on AI and creativity has used two approaches, one open and one more structured. The open one was ‘here’s what it can do, go for it’, while the other had instructions like using AI to create something specific. Young people are used to both, “Dreams” on Playstation sold well despite the clunky interface of Playstation consoles. Structure is built into pedagogy and most games.
The work uncovered what successful cartoon creators know; old “Looney Tunes” cartoons work because they ‘think’ like kids, and great new cartoons like “Bluey” because they think like kids and adults. AI was created for adults who are just being lazy. This means it can help children, but they need guidance that AI doesn’t do anything unless lots of people do it or they guide it.
In the early days of computer simulations, designers who didn’t understand limits and how to create parameters could derive a solution with high accuracy, and it could be completely wrong or, in the case of an engineer, impossible to manufacture. AI with its six fingers ends up doing this kind of thing all the time. With young people, they are not convinced that the answer is right; they easily convince themselves that their question is wrong if they can’t get the AI to do what they want.
Educators would have to allow room for failure and restarts due to this. We do not want young students to do what an alarming number of scholars on R01 scholarships do; they propose an incremental experiment that they are sure will work because they know that if they do something bold and fail, government funding will follow.
GRADES:
(1) We don’t do well on standardized tests because we don’t teach the test the same way that countries like China and Finland do. We are leaders in adult scientific literacy, Nobel Prizes and scientific production, because we teach how to think.
Invariably, the group most likely to insist that children are stupid are the progressives who want Europe to be better, so they are right to hate America – but then claim that only more money for the same people will solve the problem.
(2) Self-driving cars will never kill the 200,000 people who are killed every year in the US by people driving cars. Not even close.
(3) We haven’t been back to the Moon in 50 years because NASA has turned into a workplace. They had a 20 year overrun and a 1000% budget just to launch the James Webb Space Telescope, they can’t go back to the moon because they can’t check the cultural boxes and at the same time make sure no one dies. “Deadpool 2” couldn’t even have a basic motorcycle stunt because the production group required the artist to match the actor’s skin tone rather than having the skill.
