i have a funny story about how i came in to the ai research space. it was through getting kicked out of apl for using ai (it is more nuanced...i swear). the result of leaving apl gave me time i could never have gone back for and it was during this time where i found my passion for the underlying systems of artificial intelligence. what that is will be discussed another time. let's focus on what happened.
i was a returning intern at johns hopkins apl. the summer before, i’d worked on ai reliability and presented at princeton university.
on my first day back, my mentor gave me a presentation about the project. the slides were full of jargon. the explanation was bad. i couldn’t understand what we were actually supposed to do.
so i put the presentation into chatgpt and had it connect the dots for me.
i didn’t know or think it was confidential. obviously, if i had, i wouldn’t have put it in there.
i was in the Air & Missile Defense Sector, group A5E, focused on Human Centered Design. the project was called sentinel. every morning, part of the work was going through u.s. military news, including emails from sources like usstratcom. they wanted to automate the sorting so they could get to the things worth reading and the keywords they cared about.
once i understood the problem, i started building.
i used the antigravity ide to build a dashboard that brought together news and updates from multiple sources, including x and usstratcom-related material. it showed graphs of the biggest keywords, when things were published, and what had been flagged as critical. there was also a human-in-the-loop component for people who wanted to review what the system found. i called that part agent triage.
there were other features too. i’d have to pull up the old project to get those details right.
this was a problem they’d been dealing with for years. i built a tool that solved the workflow problem they’d shown me. it worked. i was excited about it.
the following week, i got kicked out.
the reason they gave me was uploading the presentation to chatgpt. first, they told me it was confidential. then, when they kicked me out, they told me it actually wasn’t confidential.
that makes no sense to me. i’m still pissed about it. i was trying to understand a badly explained project, and then i went and built something useful.
what i took from this is that some places aren’t ready for innovation or willing to adapt, no matter what they say. i want to work with people who are actually interested in what new tools can do and willing to try things.
i’m glad i’m not there anymore. i still want to build things and solve difficult problems. good thing that hasn’t changed.
apl as a whole, i respect their mission. some of the people there do not know what they are doing.