It pains me, deeply, that I have to see AI in this form. From very early age I was facinated by programming, my childhood dream was to create a chat robot. I studied machine learning and linguistics, and coded a lot. I used the local library, they had books but internet connection was too expensive for my family.
In my teenage years I was the token nerd, I liked technology, I liked new and shiny things (still too poor to afford any), and the progess we the humanity achieved together.
Fast forward today, I work for a multi, as a software developer. I did not chose this carrier, It was the easy path. I already love problem solving and coding, so why not, it even pays well. Now it is threatened, by AI. And by AI I mean LLMs, unfortunately research did not get founded by the artificial hype, it is the opposite actually. Some already betting that AI can do the research in the next X months/years. (claiming the same number no matter when you ask them)
It pains me that a project, which aimed to help cancer screening went under. It was developed by a bioengeneering/developer team. I helped by coding a lean AI which tracked cells moving in 4D in scans. (3 spatial 1 time dimension)
In my work, there was a greate engineer, we talked a lot, and shared ideas for hobby projects (yeah, true nerds 😅) unfortunately he had to go to work elsewhere. But some days ago, he got the opportunity to get back to our company, but joined another team. We were the only two person in the office, as it was pretty late. I striked a conversation, because I saw he is just watching an agent writing his code. Or so I thought.
(Sorry its get a bit techincal) Someone created a change in the code, and forgot to change the unit test with it, so it is failed on the pipeline. He prompted an agentic AI to resolve the issue. I checked the changes, only some reworded variables, nothing that needs to modify the logic of the tests. I watched the running Agent: AI failed to run the environment, so modified the the main cmake, read at least 3 totaly unrelated source files, opened 36 terminal. Then 20 minutes later "solved" the problem. I did not watched him instructing the AI to commit the changes. The AI did. With the modified environment. Safe to say it broke the whole pipiline. The solution? Yeah ask AI what happend and fix it, make no mistake... Another half an our later, he had 5 variable names renamed.
Same person. I sent him an article of a common intrest, the link to the pdf, and watched as he opened, copy pasted the title to google, and read the AI summary. For a fking article which has an excellent abstratct already!
I lost hope. Send me hugs.
Uff, thats really tough. Hang in there! I was fortunate enough to land a job, mostly because ppl I knew from uni recommended me.
Jeah, I am pretty disapointed that AI learned to code on my projects. Sometimes I wonder if it poisoned the model a bit at least... I had everything published I ever written, should have kept them on the floppy.