Not an AI maven, but working in recruiting/hiring software I'm happy that (for now) company leadership seems to be very cautious about LLMs, especially since they tend to embed many of the same biases we're trying to avoid. (For example, preferring prestigious university names in resumes.)
> It constantly suggested plausible but completely non-functional code, scaffolded the project in an outdated format, and autogenerated CSS classes that looked like they could be Bootstrap classes, but weren't.
Yeah, it's almost like some kind of P/NP discussion where it's now easier to generate a proposed solution... but verifying the solution was already the functional bottleneck.
I believe that LLMs--due to their core algorithmic nature--can only supplant an Enthusiastic Intern. No matter what they generate, it still needs to be carefully reviewed by a human developer to detect if it is wrong/insane and to fix/reject it.
__________
> And as they became trendy, people often used [blockchains] for purposes where their characteristics weren't necessary — or were sometimes even unwanted — and so they got all of the flaws with none of the benefits.
Absolutely, and one of my favorite examples of this involves how it is NOT a good fit for voting. (Digression follows)
I believe the main test for "is a blockchain worth it" is to ask: "Does this need to allow anyone anywhere to create any number of new nodes?" That is the core feature which differentiates "blockchain" from regular old distributed databases, and that single requirement explodes into a tree of complicated supporting code-features, performance tradeoffs, and other limitations.
So... Is that true for voting? No: We don't want to allow a foreign country to create a billion nodes and spoil the result just because we didn't budget for having two billion nodes running on that one voting-day, nor do we don't want to have to wait at the polls for hours for our vote to be truly confirmed, etc.
In contrast, it would be a thousand times faster and simpler to create a network with a predefined fixed membership, where dozens of nodes are run by different state/province/federal authorities, and trust that nobody could corrupt all of them simultaneously.
> It constantly suggested plausible but completely non-functional code, scaffolded the project in an outdated format, and autogenerated CSS classes that looked like they could be Bootstrap classes, but weren't.
Yeah, it's almost like some kind of P/NP discussion where it's now easier to generate a proposed solution... but verifying the solution was already the functional bottleneck.
I believe that LLMs--due to their core algorithmic nature--can only supplant an Enthusiastic Intern. No matter what they generate, it still needs to be carefully reviewed by a human developer to detect if it is wrong/insane and to fix/reject it.
__________
> And as they became trendy, people often used [blockchains] for purposes where their characteristics weren't necessary — or were sometimes even unwanted — and so they got all of the flaws with none of the benefits.
Absolutely, and one of my favorite examples of this involves how it is NOT a good fit for voting. (Digression follows)
I believe the main test for "is a blockchain worth it" is to ask: "Does this need to allow anyone anywhere to create any number of new nodes?" That is the core feature which differentiates "blockchain" from regular old distributed databases, and that single requirement explodes into a tree of complicated supporting code-features, performance tradeoffs, and other limitations.
So... Is that true for voting? No: We don't want to allow a foreign country to create a billion nodes and spoil the result just because we didn't budget for having two billion nodes running on that one voting-day, nor do we don't want to have to wait at the polls for hours for our vote to be truly confirmed, etc.
In contrast, it would be a thousand times faster and simpler to create a network with a predefined fixed membership, where dozens of nodes are run by different state/province/federal authorities, and trust that nobody could corrupt all of them simultaneously.