Remember:
- The effort needed to babysit the LLM to ensure it is producing the correct output is always equal to the effort needed to do the work yourself, and,
- The computer is only ever as smart as the idiot who punched its deck in the first place.
this post was submitted on 29 May 2024
351 points (97.3% liked)
Facepalm
2618 readers
4 users here now
founded 1 year ago
MODERATORS
Not really. Just going to post a comment I post whenever this topic comes up, related to how our company utilities it as software developers / engineers.
Sure they make mistakes, but even when wrong it still gives a good start. Meaning in writing less syntax.
Particularly for boring stuff.
Example: My boss is a fan of useMemo in react, not bothered about the overhead, so I just write a comment for the repetitive stuff like sorting easier to write
// Sort members by last name ascending
And then pressing return a few times. Plus with integration in to Visual Studio Professional it will learn from your other files so if you have coding standards it’s great for that.
Is it perfect? No. Does it same time and allow us to actually solve complex problems? Yes.
Yep. I use it to scaffold, rubber duck, and comment. I check it after but it does a pretty good job at that. It gives slightly better feedback than the stuffed bear on my desk and the comments are usually pretty good.
In addition, I'll write up a big ol' presentation that needs to be summarized but I'm not great at self editing. So I'll feed it through and have it give me an elevator pitch, an executive summary, and a shorter presentation. It does a passable job with very few edits for that.
It's definitely not going to replace me. It can't do my job or even anything close. But it has helped me be better at my job the same way that VS Code and PyCharm make things faster and better than just typing in Notepad.
Right. I use it every day and I'm probably 600% more productive now. What's shocking to me is the apparent difference in quality between ChatGPT and Google's AI. ChatGPT is incredibly useful. Google's AI seems dangerous and ridiculous.
"punched its deck" i love this
I don't think its ALWAYS true that the effort is equal. I think that for important things that need to have very low error rates it is going to be equal or exceed. But if you are doing creative writing and have writers block. Generative AI can help give ideas that can be helpful even if wrong because you can identify why they are wrong which might point you in the right direction.
You know... people give Google's AI assistant a lot of shit, but this recipe was fire. 🔥
That's so stupid!
Yeah, didn't they put up a fence to keep people from jumping?
Smh my head
Not a fence, a net. So it now just really hurts when you jump.
Not only that, they have a flotilla of boats covered with mattresses to catch those that get through the fence. But they don't have enough boats yet, so the success rate of catching people of diminishing altitude is only about 20%.
Where #2?
There should be two images there!
But here's a 3rd just because:
That is 2.
The first is smoke 2-3 cigs a day while pregnant. The second is jump off a bridge if you are feeling depressed.
I'm just dumb and didn't see that it was 2 separate images
All good. It took me a minute to notice it was 2 separate images.
I wonder how long it will take for Google to pull the plug of that feature. I mean, this is hilarious of course so there's that, but I'm pretty sure that relying on the awareness of the user to fix the oopsie woopsies is not a good idea. This will end badly.
For everything deadly answer that is also hilarious and obviously wrong for most people, there might be tens or hundreds of deadly answers that are not as obvious for a significant percentage of the users.
I is just a question of when Googles artificial "intelligence" will kill someone.
Probably not until they see a lawsuit
How are more people not on board with this?
Its the perfect way to train a new generation that they shouldn't trust everything they read online
These have to be fake right... please tell me they are fake.... please
I thought they must be photoshopped....but then I messed around with it just now for, like, a minute:
Okay you can't post that and not tell us the best recipe for hockey pucks.
To be fair, anything is edible at least once.
They're very real. There's also one telling people to eat glue, which got a lot of attention recently. Lol
The actually scary thing is when the AI suggestions become less ridiculous. It seems very unlikely for anyone to try to sauté garlic in gasoline, but at a certain point these suggestions are going to sound more reasonable but be equally as dangerous. Say, what medication is safe. Or what to do in an emergency. Those are the things that are going to get people killed.
I'm extremely surprised we haven't seen the Ai telling people that mixing vinegar and bleach makes a killer cleaner
See, you say that, but...
That's pretty wild but it's not quite suggesting you mix them. And I run vinegar in my washing machine regularly and never had any problems on the rare occasions I use bleach instead.
🤦♂️🤦♂️🤦♂️🤦♂️
Well in all fairness it would be, just think how clean the person lungs would be.
I would guess that it is to busy suggesting ammonia and bleach.
Actually I looked and at least Google passes the don't mix poisonous gasses in you home test. Maybe not a high bar, but hell the floor has to be somewhere right?
Bleach is such a fun ingredient! Did you know if you mix bleach and rubbing alcohol you get chloroform?
Those are the things that are going to get people killed.
And Google will be held accountable right.... Right?
Is it known where the LLM got this one from?
Their ass, presumably Edit: Someone elses ass, presumably.
They're not creative enough to pull things out of their metaphorical ass. They can only mindlessly spew what they've seen somewhere before.
All that useless reddit data.
OP didn't precondition their query with the requirement of living through making and consuming the spaghetti.
What in the actual fuck? How is Google not facing massive fines every day that they allow this abomination to exist online?
I love how confident it is
"I am telling you for the last time, no you cant"
I'm thinking it will be a while before AI can take over the world - considering it currently has the intelligence of a newborn monkey.
Though before that happens, it might launch all the nukes because someone asks it to boost rocket power and forgets to specify just for a rocket design they are working on.