←back to thread

127 points Anon84 | 2 comments | | HN request time: 0.407s | source
Show context
ufmace ◴[] No.38509082[source]
The article title is clickbaity, but the actual point is the proposal of using LLMs to translate large amounts of legacy COBOL systems to more modern languages like Java. Doesn't seem terribly useful to me. I expect you could get a 90% solution faster, but the whole challenge with these projects is how to get that last bit of correctness, and how to be confident enough in the correctness of it to actually use it in Production.

But then all of this has been known for decades. There are plenty of well-known techniques for how to do all that. If they haven't actually done it by now, it's a management problem, and no AI tech is going to fix that.

replies(11): >>38509198 #>>38509418 #>>38509802 #>>38509995 #>>38510231 #>>38510273 #>>38510431 #>>38511157 #>>38511186 #>>38512486 #>>38512716 #
victor106 ◴[] No.38510431[source]
> Skyla Loomis, IBM’s Vice President of IBM Z Software adds, “But you have to remember that this is a developer assistant tool. It's AI assisted, but it still requires the developer. So yes, the developer is involved with the tooling and helping the customers select the services.” Once the partnership between man and machine is established, the AI steps in and says, ‘Okay, I want to transform this portion of code. The developer may still need to perform some minor editing of the code that the AI provides, Loomis explains. “It might be 80 or 90 percent of what they need, but it still requires a couple of changes. It’s a productivity enhancement—not a developer replacement type of activity.”
replies(2): >>38510757 #>>38510835 #
1. koenigdavidmj ◴[] No.38510757[source]
That seems to be the spot humans are weakest at—reviewing something where we think the computer did a good job 90% of the time, but quickly noticing when something goes wrong. Similar to level 3 self-driving—requiring full attention, able to instantly snap into full unassisted driving.
replies(1): >>38510941 #
2. ◴[] No.38510941[source]