Imagine Artificial Super-intelligence (ASI) is achieved, and after the technology causes widespread socioeconomic turmoil, humanity decides that ASI will be free to all, but with some guardrails, regulations, and a reasonable token limit.
Imagine that, in that not-so-distant future society, you are one of the few who still have foundational knowledge of technology, programming, and software engineering.
Now, imagine yourself up against someone who is not good with computers or technology in general. Say that person's name is John.
In that scenario:
- Who will be able to create better and more reliable (to a human) software using ASI? John, or you?
- Who will have a better chance of saving humanity if ASI goes rogue? John, or you?
Some say that people who are trying to avoid facing the inevitable reality of highly capable AI are in denial. I agree. We were all there to some extent not so long ago.
On the other hand, people who are overhyping the use of AI in serious software development, without restrictions or proper human vetting, are being dangerously irresponsible.
Foundational human knowledge is essential for a better future. And that is especially true for the survival of our species against all sorts of future dangers.
Hence, we must strive to create a safe environment where learning and using foundational knowledge is highly valued. This will ensure that we retain our current knowledge and skills, and that future generations have enough incentive to do the same.
"Oh, look what I can produce with my little AI token count!" is not cute anymore.
We need to put government regulations in place to make sure AI-generated software is properly verified by humans through an actual, verifiable review process before it can be used in critical sectors such as government institutions, hospitals, banks, and schools. A skim-through session in the name of review is not enough.
Private and personal use can't be restricted anyway, but AI use in software development for the public sector must be regulated and monitored, with proper accountability attached to the review and release process.
That's essential to ensure that, one fine morning, we won't find ourselves in an unforeseeable AI apocalypse.
The solution to such a probable apocalypse is not more AI development; rather, the solution is in our responsible actions and in our very own foundational knowledge.
Top comments (4)
The John example is such a simple way to explain it. Even with the best AI, the person who understands the fundamentals will still build better software and catch problems others would miss. I also agree that a quick skim shouldn’t count as a review, especially for things like hospitals and banks.
Absolutely!
An actual review is order of magnitude more difficult than writing that code. Because when we write code, our thought process comes first and then comes the syntax. Whereas, in review, we check the syntax to discover the thought process (or the lack there of) - kind of like reverse engineering.
Which means, for serious software development, unrestricted process of code generation using AI doesn't make software development any easier, it only makes the hardest part of software development the only part that's left for us to do all day, without enough supply of dopamine. 🤪
I created a repository just for building foundational knowledge.
It's in a very early stage. 91 commits just because it stores a long history that was all removed without a force push 🙂
Good! 🥰
I'll keep repeating it with different arguments until most people get out of their misunderstanding 😇