Aaron Jack
An overview of the gpt-3 machine learning model, why everyone should understand it, and why some (including its creator, open AI) think it’s dangerous.
Like if you learned something && subscribe for more machine learning (we can learn together)
====
Links
====
gpt-3 paper
https://arxiv.org/abs/2005.14165
the open AI API
https://openai.com/blog/openai-api/
gpt-2 open source repository
https://github.com/openai/gpt-2
==========
Social Media
==========
https://twitter.com/_aaronjack
https://instagram.com/aaronjack
#io #ai #ml
my bad on the audio on this one guys, i switched microphones. it sounds better if you play it at 1.5x
https://www.readerworld.co/2020/07/gpt3.html
If anyone wants to know about gpt3
5:19. This is impossible to distinguish from a human-written article. The people that said it was AI just guessed. Period.
7:40. This sort of concern makes me laugh. It is worthless virtue-signalling at best. This tech will be used, just like everything else, for both good and bad reasons. It is inevitable. Reminds me of the silliness related to "holding back" AI in general so that it doesn't cause harm. Right. Good luck.
How do we now differentiate text generated from GPT-3 to the original text? 😅
@8:29 TIL 'Cool' is an Atheist term lol
What do gpt 3 stand for? Do you only code in Linux in your videos? I wish that you can explain a little more and more detailed when showing things like how to make aI-comments on youtube, i am trying to learn programming but it is either beginning videos or high level that i do not understand..but i want to learn so i guess i have to gind 🙂
it looks your glasses are a foot thick, man
Where to find that to use it?
I think beyond language processing it could easily autoconvert between programming languages or platforms. As centralized technologies become obsolete they will need rapid replacement by new decentralized systems and gpt3 should help do it a lot faster.
making it "politicaly correct" is wrong.
Correction: as the models increased in complexity, people were less likely to tell the difference; hence the roughly 50% score (which is optimal in this case, i.e., a coin toss). The 80% score is for the smaller models, meaning people could tell the difference between the model's generate output.
NLP is also neurolinguistic programming:) lol really great video. Thanks
This API will be public soon. Imagine what they actually have in active development. Now imagine what china might have
Nice u neck
GPT-3 is insane!
When did Casey Affleck get into machine learning?
We're getting closer to the star trek "computer" and 300 years early. Nice
Great intro! It must be hard to make people watch the entire video, but you really got me excited to watch till the end!
Was the frame rate of this video really low, or is something wrong on my end?
Hi, I didn't understand the negative side/effects of gp3… Could someone kindly summarise it to me? Many thanks
This is just the beginning. Pretty soon it will be integrated into everything we do. We need to give it a voice.
If the YT algorithm be listening what you say, why not read what's on your screen as well 😹
"it's dangerous" is BS marketing. It can't do any actions.
haha great video! 20K likes!
Gpt 3, a reflection of the toxic web lol, the ultimate echo chamber, imagine a child growing up spending every minute on the internet and that’s what you end up with
If you pay close attention you'll see that during the video Aaron will turn into Wide Aaron
AI it's becoming corporate machine xD This is the end for open AI and start of corporate AI, somehow analogous how facebook was the end for small personal webpages and forum etc. They still exist but gigant are at the center.
Next generation, text generation
at 3:01 it's "cost" not "costed"
I can't stand that noise in background, music is alright but that ugly disgusting noise
I'm sure there is more than "guys" watching this video
Closer and closer to our savior , A.I.
When AI takes off and starts to fix and improve itself we will be able to unlock all the mysteries of the universe, we will be set free
Just wait until AGI! SingularityNET
Just wait until v4. its too bad we cant do this at home and have to use their API. They control the machine. ( so to speak )
Wait the output of 52% was the best result not the worst, it means humans only had a 50/50 chance of guessing correctly – that's amazing for an AI model!
Imagine: In a game asking an NPC anything and he or she will answer like a normal human. WOW!
I think this could become dangerous very quickly. Anyone agree?
I hope someone makes a loli chatbot
And then there's the applications of this tech in phishing and identity theft.
Great video – thank you for it. Yet, this part: https://youtu.be/lQnLwUfwgyA?t=505 – is scary. You literally self-censor yourself. That's just crazy.
How to enable this in excel workbook (o365)?
8:23 You know, it's pretty ironic that in order to simply convey information about an AI model, you have to take steps to try to fool a different AI to prevent your video from being flagged for review.