On AI
These days, it seems like wherever you go The Middle Way is nonexistent. If you're not supporting AI, you're decrying AI. It's not even fashionable as a topic, as the points have been beaten to death by both sides. I wish reality is as easy, as black-and-white.
As someone who loves linguistics and technology, NLP (Natural Language Processing) was fascinating. Computer Vision was fascinating. Neural network was fascinating. In fact, I am where I am now because the whole field seemed like a vibrant world more than a decade ago. Today, confessing that I'm not always against AI feels like a dirty laundry, and the term has been tainted so badly it becomes almost impossible to explain to layman that AI doesn't always mean GenAI and LLMs. I'd had seen people who steadfastly refuse to engage with anyone who has hinted even an ounce of positive feeling about AI, because in their mind it's all GenAIs we might as well be supporting genocides. Nevermind how chaotic it has become in sites like itch.io which mandates disclosure but use the term "AI" when it means "GenAI" and every other game jam spurs confused devs asking if autonomous enemy units and pathfinding algorithms are exempt.
I'll never forgive the techbros.
Quanta Magazine published a story about the death of NLP and it was a harrowing, heartrending read.
Much ado has been said about LLM. Teachers seem to have it worst. I don't envy them. Teaching has always been hard and it just becomes harder. In other working situation, I don't buy that tools like ChatGPT helps the lowest performers, because without someone to coach them and help point out where GPT gets it wrong, you risk having someone making mistakes and not knowing it because they turn in something seemingly passable at a glance. Keyword seemingly. GPT will get something wrong. I say this as someone who has to work with LLMs 8 hours a day. If you spend time coaching them anyway, you could do away with the GPT outputs and just ask them to try. Those who are able to learn and synthesize their thinking on their own with GPT are those who will be able to cobble together their own education GPT or otherwise.
What I hate from the current situation is how it makes management and the higher-ups (because it's always them with heads in the clouds while everyone else have their heads buried under the sand) think they can assign more tasks to smaller teams (or even individuals), basically forcing everyone to embrace GenAI to deal with volumes and deadlines impossible to achieve otherwise, and then raise hell when GenAI inevitably does it poorly and the workers have to...what else? Fix it. Which means they're now doing twice the workload.
I know what you're probably thinking. Then they just have to learn better prompting!
You know, personally, I think machines should aid the way we work instead of us fitting the way we work around them. Ever heard of the term ergonomics?
Not to mention, I think a lot of people can attest that in many workplaces the routine works are... not exactly routine. Yeah, that factory client in Ankh-Morpork has gone through so much with us, we give them 20% wholesale discount for the first order per quarter and every blue moon we'll give 25% and if Wednesday falls on the 3th or it's their president's birthday month we're giving them 30% and all of these are going to be manually adjusted by Kevin in Accounting because our ERP cannot handle such granular controls. Sometimes, depending on where you work at, you deal with more exceptions than rules because doing business Humanly is messy because Human is messy.
I agree that places like that should probably get an intervention sooner or later, but it doesn't change the fact that reality is chaotic, clean data is a myth, someone still has to get the job done in the meantime and GenAI is horrible at this.
Besides, I'd rather train someone who'd eventually get it done better.
But I have no control over my workplace, so what I can do is to steadfastly refuse to use it outside. I'm happy to do all my hobby projects the old school way, and I'm glad I like niche quirky engines so GPT and the ilks won't be of use anyway. Honestly, I hesitate admitting that I deal with LLMs in my dayjob in the space I show off my personal projects, because the level of witch hunt is making me paranoid.
Speaking of that, please stop the witch hunt.
It does no one any service to fuel distrust and imposing impossible standards. It's a battle one can't win, because perfect result is now AI, bad result is now AI, so to maintain the face of your humanity you have to pretend to be a mediocre everyman, the mathematical average of everything?
Screw it.
I'll never let them take my em dashes.