Bakz T. Future
Why GPT-3 and “Prompt Engineering” are both really bad names/labels we’ve given to very important ideas.
—
► Remember to Like, Comment, and Subscribe!
—
Connect with me:
Podcast – https://bakztfuture.podbean.com/
Substack Newsletter – https://bakztfuture.substack.com
Twitter – https://www.twitter.com/bakztfuture
Instagram – https://www.instagram.com/bakztfuture
Github – https://www.github.com/bakztfuture
Feel free to send me an email or just say hello:
bakztfuture@gmail.com
1
7:30 is the moment when he stops bitching about the name gpt-3 it saves you some time probably 🙂
Another broken video
-GPT-3 isn't what the consumer uses, they use an app that's made using GPT-3.
-The engineering in Prompt Engineering implies the engineering of a prompt to get the wanted outcome, not that's it's a field of research as wide as other sciences. ("the action of working artfully to bring something about" is a definition I just found on google but I think it fits nicely)
As a 23 y/o novice in the programming world thought when I heard these names:
GPT-3: Another 3 letter acronym that contains words I've never heard before, just like all thous other well respected tech tools. And the 3 implies that it's the third iteration, so it's proven to be a thing worth improving.
Prompt Engineering: The prompt part sounds horrible, abstract, time-consuming and frighteningly vague but engineering implies that it's more then trail and error, that there's something to learn and get good at.
Why is it such a concern that these terms sound like all other programming terms, it's not like someone who struggles with excel would ever need to interface directly with GPT-3, there will allways be layers of abstraction the end user doesn't have to worry about.
The person Googling doesn't need to be comforted by the proper IT name of Googles search ai to know they can use normal language prompts.
We need this sort of critique, not many programmers care about these things. I would consider myself to fall under that group of not being considerate. It's good seeing conversation about topics I never even considered, that's how true progress is made. That being said, the arguments above are my input on these names. I like them and I don't shear you're concerns.
I agree with the points you are making, this is generally a problem in higher-level circles.
It really just feels like an evolution of a phrase like "social engineering"– which has a lot of the same problems but I don't think people really obsess about whether the terminology is precise. It works fine. Describes an "art" as much as anything else.
Prompt Response Priming would be fine. But I think honestly prompt engineering is more like "search engine optimization" or query optimization. "NLP Completion Optimization" or something
We are not stuck at all with the name GPT3. Only people who think the AI field will never advance any further thinks that. The next 10T parameter model will be the only thing that matters and GPT3 will be thrown away as just another stepping stone. GPT2 is "irrelevant" now even though it was impressive in the past. BERT is another big model with another name overshadowed by GPT3. The models themselves will be swapped out when the next cool thing comes out, and with it the name. It might be called GPT4, but then we just have to wait a little longer. Just ignore it and say natural language ai model or something which describes the category instead of the specific model when talking to non-techies.
Prompt Engineering is easily understood by software engineers what it means, it's a good name for that. Call it just prompting if you need to speak to casuals…