# OpenAI’s GPT-4 Just Got Supercharged!

## Метаданные

- **Канал:** Two Minute Papers
- **YouTube:** https://www.youtube.com/watch?v=Fjh1kwOzr7c
- **Дата:** 07.04.2023
- **Длительность:** 7:10
- **Просмотры:** 368,418

## Описание

❤️ Check out Lambda here and sign up for their GPU Cloud: https://lambdalabs.com/papers

📝 The paper "GPT-4 Technical Report" is available here:
https://cdn.openai.com/papers/gpt-4.pdf

More here:
https://openai.com/product/gpt-4

Try it out (note: the free version has the older GPT-3 as of now)
https://chat.openai.com/chat

My latest paper on simulations that look almost like reality is available for free here:
https://rdcu.be/cWPfD 

Or this is the orig. Nature Physics link with clickable citations:
https://www.nature.com/articles/s41567-022-01788-5

🙏 We would like to thank our generous Patreon supporters who make Two Minute Papers possible:
Aleksandr Mashrabov, Alex Balfanz, Alex Haro, Andrew Melnychuk, Benji Rabhan, Bryan Learn, B Shang, Christian Ahlin, Eric Martel, Geronimo Moralez, Gordon Child, Jace O'Brien, Jack Lukic, John Le, Jonathan, Kenneth Davis, Klaus Busse, Kyle Davis, Lorin Atzberger, Lukas Biewald, Martin, Matthew Valle, Michael Albrecht, Michael Tedder, Nikhil Velpanur, Owen Campbell-Moore, Owen Skarpness, Rajarshi Nigam, Ramsey Elbasheer, Richard Sundvall, Steef, Taras Bobrovytsky, Ted Johnson, Thomas Krcmar, Timothy Sum Hon Mun, Torsten Reil, Tybie Fitzhugh, Ueli Gallizzi.
If you wish to appear here or pick up other perks, click here: https://www.patreon.com/TwoMinutePapers

Chapters:
0:00 Intro
0:32 Meal Planning
1:42 Browsing
2:33 Image editing assistant
3:33 Research assistant
3:58 Mathematician and coder
4:52 It improves itself!
5:37 History in the making

Thumbnail background design: Felícia Zsolnai-Fehér - http://felicia.hu

Károly Zsolnai-Fehér's links:
Twitter: https://twitter.com/twominutepapers
Web: https://cg.tuwien.ac.at/~zsolnai/

#gpt4 #chatgpt

## Содержание

### [0:00](https://www.youtube.com/watch?v=Fjh1kwOzr7c) Intro

Dear Fellow Scholars, this is Two Minute  Papers with Dr. Károly Zsolnai-Fehér. My goodness. GPT-4 barely came out just  a few days ago, and plugin support has   already arrived. What does that mean? It  means supercharging GPT-4 by letting it   build on other applications. That is going to  be a game changer. Here are a few incredible   examples that showcase the power of this  system so you can also get more out of it. One, meal planning. Now, you may be wondering  why we are using a powerful mathematical system,

### [0:32](https://www.youtube.com/watch?v=Fjh1kwOzr7c&t=32s) Meal Planning

Wolfram Alpha for a meal, but you’ll see,  just wait a moment. Now we are asking for one   restaurant recommendation and a simple recipe for  the next day. And now, watch the magic of plugins   happen. ChatGPT recommends us a restaurant, and,  here comes the first plugin - this allows us to   just click here and reserve a table. Then, we  get our recipe recommendation, chickpea salad,   and now comes the magic: Wolfram Alpha  immediately helps us calculate the amount   of calories in the meal. Wow. Finally, if we like  what we are seeing here, with the third plugin,   ordering the ingredients is just one click away.   Now that is what I call an amazing assistant. And don’t forget, this is a chat assistant, so if  we are looking for something with fewer calories,   we can just ask again and it will calculate  and arrange everything for us. I love it.

### [1:42](https://www.youtube.com/watch?v=Fjh1kwOzr7c&t=102s) Browsing

And while we are talking about having  an assistant, check this out. Two. Here,   we are interested in knowing a little  more about the Oscar winner movies and   their box office performance compared  to others. And, hold on to your papers,   and look at that! It is browsing the internet,  tells us what is searched for, where it clicked,   it now knows about the awards, looks up box  office performance, and then, the most important:   synthesizes all this information and boils it  down to what we asked for. That is incredible.    And it cites its sources too, so if we  are asking about something critical,   like the good Fellow Scholars we are, we  can also double-check ourselves. So good!

### [2:33](https://www.youtube.com/watch?v=Fjh1kwOzr7c&t=153s) Image editing assistant

And if we give it an image, it tells us what  it depicts. That is already quite cool in   and of itself, but now comes the assistant  part. Without touching any software tool,   and requiring any sort of computer expertise, we  can just write a line and ask it to resize it,   then, make it grayscale. You know what,  little AI, give it a green tint. Finally,   we can ask for the foreground only, and  a download link. Now, don’t get me wrong,   none of these tasks are difficult to someone who  can code or use some other piece of software to   accomplish this. But, you don’t need any of  that now. Just imagine how many millions of   people will be able to use the power of today’s  computers without needing to be able to code   or buy a bunch of software and figure out how  they work. This is going to change the world.

### [3:33](https://www.youtube.com/watch?v=Fjh1kwOzr7c&t=213s) Research assistant

We can also upload a file with a bunch  of data and ask GPT-4 to try to make some   sense of it. And it does that beautifully.   This is low-key amazing: what it has been   given is just a bunch of numbers, and it had  to find out what part of this is interesting   and why, and it also visualized it in a  way that is easy to consume for a human.

### [3:58](https://www.youtube.com/watch?v=Fjh1kwOzr7c&t=238s) Mathematician and coder

Now let’s ask for something a little  different. This is not a hard problem at all,   but it requires a bit of mathematics. I  wonder what is going to happen here? Yes,   first it writes up two variables, and  with those, a system of equations,   and of course, now we need to solve  it. So, does it? Yes! Okay, but how?    Whoa! Look! It even writes a simple computer  program for it, for which it shows us the full   source code, so we can actually redo or check the  calculations ourselves. Or even extend it to more   complex examples. Here, not just problem solving  is happening. Teaching is also happening! So cool! And finally, here comes the bombshell. Writing  a plugin for ChatGPT is not that difficult,

### [4:52](https://www.youtube.com/watch?v=Fjh1kwOzr7c&t=292s) It improves itself!

here OpenAI creates a toy example in  minutes. But wait. Are you seeing what   I am seeing? They are using ChatGPT to  write a plugin for ChatGPT. My goodness!    We can use ChatGPT to improve itself.   This requires a little human help,   but it is possible. And all this will open  up the world of computers, and creativity   for millions and millions of people, all around  the world. And quite possibly all this for free,   or close to free. What a time to be alive! I feel  that we are witnessing history in the making. Wow.

### [5:37](https://www.youtube.com/watch?v=Fjh1kwOzr7c&t=337s) History in the making

And all this happened just in the last  few days! How cool is that? The rate of   progress in AI research is nothing short of  amazing. Remember, GPT-2 could barely finish   our sentences and was not that useful, and just  two more papers down the line, here we are. So,   what do you think? What will GPT-6 be capable  of? Let me know in the comments below! Thanks for watching and for your generous  support, and I'll see you next time!

---
*Источник: https://ekstraktznaniy.ru/video/13223*