Top Tech Voices Podcast S2 E1: Mental Health & Technology

I liked the part where AI hallucination was discussed. How AI can provide a lot of factual statements and hide few incorrect statements that will take experts and hours to uncover it. In my organization there was a plan to migrate from Azure Devops to Github enterprise because Github offers AI reviewer which could help in reducing the reviewing time. The obvious question at that time is what if during review, the AI suggested a different way to fix the solution citing reasons that sounds extremely possible but actually has no chance of happening. The idea is on hold now. If this is the case with AI reviewing, on other side people are using LLM to look for bugs and "improvements" in open-source projects and raising pull requests to fix that issue. Problem is most of the PR are spam, some are of no significance, and some claim to fix an issue that out rightly doesn't even exists.
Developer behind the CURL has closed bug bounty program because of these low-quality pull requests.
"We saw an explosion in AI slop reports combined with a lower quality even in the reports that were not obvious slop – presumably because they too were actually misled by AI but with that fact just hidden better."
You can read more about this here.
With Openclaw, even the PR can be made automatically by AI agents with total having total control over the interaction. One of such agents created a PR in matplotlib and it was rejected because it is a spam. The AI then decided to make a post of how people are gate keeping open-source projects. It also started to make fun of him in the same blog. Here is where it gets interesting, Ara Tecnica which release posts about technical development made a post about this. The entire post was created by AI which decided to include about the maintainer which was entirely fabricated by itself. After the maintainer issued blog posts of his own explaining what was going on, Ars Tecnica retracted that article and issued a public apology.
Github has started listening to the maintainers and has decided to add even more control to the pull requests. There is a lengthy discussion going on to tackle the low-quality contributions. All in all, this has been an interesting month for the developers, and the podcast has been posted at the right time.

What you do? , I really like these type of job where observation is a crucial part too. As a student finding correct ambition is like the definition of Variable we read in every Coding dictionary

R&D Engineer at a power sector company. My work requires more development activities than research though.

Aniket_kumar_raj said:

correct ambition

I think getting an internship at a fortune 500 company will be a good start.

I did my personal research on R & D companies in india and actually there are none interesting companies... :(

Not a private company, but there is the Indian space agency.... not sure how easy it is to get in though.

It's ISRO, and it's kinda moderate to get in, requires an exam and interview.

It's ISRO, and it's kinda moderate to get in, requires an exam and interview.

veluv01 said:

kinda moderate

If you are applying through the ICRB entrance exam, the selection rate is extremely low (≈0.1–0.3%)). ISRO evaluates candidates primarily on conceptual understanding, so even those with high academic marks may be rejected if their fundamentals are not strong.