Jen-Hsun “Jensen” Huang, who said something quite similar, is still the CEO of Nvidia and has been the CEO since 1993.
He is a marketing man. He is selling AI devices for billions of dollars. He needs to spice things up.
AI is nowhere near making programmers out of work.
You know what you have, when you explain your requirements in small details in english? You have CODE.
AI is basically a great tool for developers. Yes, god awful programmers could pass as decent ones for a while but AI often makes code that is useless, gobblety gook basically. Referencing functions, classes and methods that don’t exist. AI can’t tell you how any times the letter S appears in STRAWBERRY. OpenAI added specific code to answer this type of question because, plain and simple, AI can not answer that question.
AI simply tokenizes words that are likely to fit together. It has data to build up more likely answers and frankly I am often amazed how it found bugs that I stumbled on. I would have gotten it eventually. It is quite useful. It also is quite good at writing code that is just a lot of typing, like changing one dataset into another. But it has no understanding of what it is outputting. It is kind of remarkable that way.
If you take a CPU like Motorola 68000 from 1979, 46 years ago. You take a very modern CPU like 14700f, a beast of a CPU. The new CPU is many millions of times more powerful, it not only has many other CPUs built in (cores), it has hyper threading so an app can create more threads so the CPU is less likely to idle and can get more work done. It is like comparing a mud hut to a skyscraper in engineering, material physics and just plain magic of reality.
But it is still doing the same things, just bazillion times faster. But we also work with pictures and data that is bazillion times larger. CPU hasn’t somehow started to do magic.
Neither has AI. AI will just get faster and more accurate. It still will not know how many B are in the word CAR. AI is by all accounts reached it maximum capacity. We can train it better, we can try to omit fairytale answers. We can run its answers through filters to make sure some of the data it is referencing actually exists. But there really is no intelligence there. It is a process that through advanced statistics, can be trained to recognize people in photos as an example. Then we train it some more through private photos and point out that this is me, this is my wife, this is my kid. Those are two different AI processes that are needed. First it needs to find if a person is in that photo. Then based on number of statistical analysis of measurements of the face area, like it can find eyes, how far are they apart given average eye size, how large is the nose, how far is that from the eyes. Is this picture more to the left or to the right, is the person looking up or down. Given enough data, it can give you some information, how likely it is. Like this one
All the information there, age, if it is a male or female, if they are happy or scared is based on AI seeing and measuring tons of images where a human has written down, female, 28, happy, male, 33, angry …. on millions of photos.
AI is not going to remove your job. It will cause fewer programmers to be hired, for sure as AI is kind of like having a partner to ask and it is helpful if you know what information you are receiving.
You can even train specifically in AI computer science.
Who knows. Perhaps there will be even more opportunities out there after a while.
AI will one day be able to drive a car completely without any issues, until it has some issues like imagining there was a line on the road.
If you put two images together, the same picture, just draw a couple of red dots in one of them, it will not be able to see that it is the same image. Just 90 some percent probability that they are similar.
