No, I don’t understand the subject in depth to have arguments against it.
When I don’t understand a subject in depth, I look for the source credibility to decide if it’s read worthy before going down the rabbit hole of making sense out of it.
It's alright to not read it but let the criticism to someone who knows if there is something wrong. Highschoolers can be extremely smart even before getting their diploma and formal education is not necessarily a good indicator of knowledge.
Actually - when you see someone with no degree in an area speaking with this level of authority you should be skeptical. High schoolers are known for copy paste work as well. I've interviewed some for internships and some of them have lots of great "papers", but when you interview them you realize they don't even actually have any knowledge AT ALL of even ultra fundamental things (in my field).
I'm reminded of Nikola's director of hydrogen tech - guy had no degree in anything close to hydrogen and had worked doing driveways. I am skeptical in those cases, as I am here!
So I'd seriously looking at something like journalistic or academic fraud or plagiarism with this guys background.
I'm not saying blindly accept whatever. If you see plagiarism or fraud, or if the article is wrong, say it! But don't just claim that if the guy has highschool diploma, it must be wrong - come with arguments, or let someone else come with them.
Non-experts usually have much larger unknown unknowns than experts, and much smaller knowns than experts.
The apriori reason for not trusting a non-expert's argument, when you yourself are not an expert, is that your unknown unknowns are also large, so you can't find all mistakes in the argument.
I had provided a link, but I checked the website and found that I cannot copy their work for uses other than personal use. So the post is unpublished now. Thanks for the reminder.
So we have options and this is all ten years out before we actually hit this wall.
However, according to this, our current solutions to the issues faced at sub-20nm lithography were developed by IBM in the 90's. So that is the question.
What technique exists today that will be used for the next 30 years?
I am envisioning a true death of "Moore's Law" but we will have time before we encounter the limits of Amdahl's law[0] in the use of chiplets and other dedicated hardware. Apple has their T2 and security enclaves, NVIDIA has tensor cores and other fixed function silicon, AMD is proving chiplets with seperate compute dies and I/O dies works well, Intel is now working on the same but for GPU's[1] ontop of compute and I/O, and the new Playstation 5 has a dedicated ARM CPU for the SSD.
Having dedicated silicon for specific tasks (such as encryption, ray-tracing, storage, networking, and so on), or even an FPGA of some sort will likely be the future.
It will be like the 90's again with northbridge and southbridge chips and 5 or 6 expansion cards.
But I am excited to see the solutions that will drive computers for the next 30 years.
Many of these new alternative technologies improve on some parameters (switching speeds or thermal efficiency), but in terms of miniaturization (which also leads to better switching speeds and thermal efficiency) the established tech (silicon) will rule for quite a while, simply because of the gigantic head start it has. At the end, none of these technologies really provide for a way to have things smaller than an atom, and silicon tech is already approaching the scale of tens of atoms. It would take a while to get other tech to that regime.
raspi has already come close enough with the raspi zero, the bottleneck is not hardware price/availability but rather people's ability to write efficient programs and solve human problems
28 comments
[ 1.8 ms ] story [ 75.7 ms ] thread> High school diploma
Sounds like a credible scientific journal
I'm reminded of Nikola's director of hydrogen tech - guy had no degree in anything close to hydrogen and had worked doing driveways. I am skeptical in those cases, as I am here!
So I'd seriously looking at something like journalistic or academic fraud or plagiarism with this guys background.
The apriori reason for not trusting a non-expert's argument, when you yourself are not an expert, is that your unknown unknowns are also large, so you can't find all mistakes in the argument.
E-beam lithography can get to real 1nm in practice quite easily, but you can spend a week on a CPU sized IC for a single exposure.
However, according to this, our current solutions to the issues faced at sub-20nm lithography were developed by IBM in the 90's. So that is the question.
What technique exists today that will be used for the next 30 years?
I am envisioning a true death of "Moore's Law" but we will have time before we encounter the limits of Amdahl's law[0] in the use of chiplets and other dedicated hardware. Apple has their T2 and security enclaves, NVIDIA has tensor cores and other fixed function silicon, AMD is proving chiplets with seperate compute dies and I/O dies works well, Intel is now working on the same but for GPU's[1] ontop of compute and I/O, and the new Playstation 5 has a dedicated ARM CPU for the SSD.
Having dedicated silicon for specific tasks (such as encryption, ray-tracing, storage, networking, and so on), or even an FPGA of some sort will likely be the future.
It will be like the 90's again with northbridge and southbridge chips and 5 or 6 expansion cards.
But I am excited to see the solutions that will drive computers for the next 30 years.
[0]:https://en.wikipedia.org/wiki/Amdahl%27s_law?wprov=sfla1
[1]:https://www.anandtech.com/show/16021/intel-moving-to-chiplet...
You can't really democratize compute, if there is a global oligopoly on manufacturing processor chips.