OpenAI's GPT-4: Eccentric Genius AI!

Dear Fellow Scholars, this is Two Minute  Papers with Dr. Károly Zsolnai-Fehér. ChatGPT has just been supercharged with  browsing support, I tried it myself too,   and I think this changes everything. Well, almost  everything, as you will see in a moment. And this   is going to be an incredible journey, so  hold on to your papers Fellow Scholars. You see, normally, ChatGPT with GPT-3. 5  has read a great deal of the huge body   of text in the internet, and therefore  it knows a lot. It can help us with so   much. But there is a problem. What it the  problem? Well, it can help us so much,   if whatever we are looking for happened before  September 2021. Anything after that is a no-go. Until today, because now we have live  browsing support. Let’s see together   what this is good for! Now, perhaps the  world’s best two investment experts,   Warren Buffett and Charlie Munger just held their  annual meeting where they discuss investing,   the economy and more. As always, it is a  treasure trove of information, however,   look. It is more than 7 hours long. So, it  is time! It is time for what exactly? Well,   for asking the new ChatGPT to summarize it for  us and see if it is worth the time investment! First, it starts browsing browsing,   and then. Ouch. Unfortunately, it cannot  watch videos, which is understandable.    Second try, oh my goodness. It is  hallucinating. Calm down, little AI,   this is not a video of a startup called Mindgrasp  AI. It’s all right. Now I am wondering, does this   startup exist or did it make it up? It can tell  us what it is, and I checked, it indeed exists. Third try, sigh. It also fails. But wait a  minute! This time it fails a little differently.    Look! A ray of hope! It can look up transcripts  instead and summarize those. Yes please! And,   let’s see…great! They talked a lot  about human nature, that is my favorite,   so I’ll definitely watch this. We could probably  even ask for the appropriate time stamps if   something specific caught our eye. So, you see, it  works, but sometimes it is a bit eccentric. Yes,   we have an eccentric, genius AI at  our disposal. What a time to be alive! But wait, if it can browse, it can also  help us with other things. For instance,   it can visit our coding projects, have a look  at our work, and even suggest improvements.    You see a ton of new code here, but it is  a bit unclear what it has changed and what   it is trying to do. No matter, look! It  summarizes the nature of the changes and   also tells us why it has done what it  has done. That is absolutely amazing. And if the database cutoff is not in September  2021 anymore, we can ask about things that   take place today. So, market research time!   How did the US interest rates change in the   last 3 months? That is a great and concise  summary, with citations. Loving it. Now,   tell me the story of interest rates in  the last 10 years. We get a lot of data,   I like it, and even better, lots of  additional blinking cursors for free.    Now let’s do some actual market research and  ask what the economy looked like last time   the US had interest rates like this. And  then, our whimsical genius is at it again. I tried again, and then it says that interest  rates looked like this last time around 2006   and 2007. And it ran out of time before it  could make a table with the appropriate data,   but it kindly asks if it can  try again. Of course, and,   good job little AI! Bravo! We now have some  numbers to compare these two situations. Now,   I hope it goes without saying, please do  not invest on this data, I just ran this   toy experiment to see a little clearer what is  going on and where we are in the debt cycle. So that was market research, and now, it  is time for research research. Namely,

I would love to know if there are real  scientific experiments about guitars. Now,   what experiment am I looking for? I am looking  for something about tonewoods. This is something   that has been a subject of intense debate among  artists who use electric guitars. You see,   manufacturers love to sell you all kinds of  woods they use for the guitar body and neck   and they say it makes a great deal of difference  in the sound they make. Perhaps it does. However,   we are Fellow Scholars over here, so we would  like to see hard evidence of that. But for me,   no luck. I don’t quite understand why, but I never  found a real paper on this topic. Until today,

because ChatGPT was able to browse its way  into a proper paper that is about exactly that. Oh my. Here is the setup of the paper, where  they actually done precise measurements on   what sounds these guitars make if they  are made of different kinds of woods.    And the result is. My goodness, can that  really be? This seems to suggest that they   indeed make a difference at least  in this particular setup. However,   we can ask our resident genius, ChatGPT  to read the paper and summarize it for us. And then, ouch. It couldn’t read the paper. But  it’s almost all text and it is right over there.    Still, nothing. Alright, so let’s give it the  abstract, results and discussion parts of the   paper as plain text, and ask the same question  again. And, bam! There we go! The body and the   neck both seem to make a difference. But it  still uses a great deal of technical jargon,   so let’s ask it to summarize it for a 10-year  old. And I think the result is excellent. It   compares the body of the guitar to the body of  a car and how it goes vroom. I love it. Clearly,   the comparison has its limits,  but still, I am very impressed. Also, every study has its own limitations.   Let’s ask it what those are - I am very excited   to see this, so let’s see. Oh come on! Let me  cultivate my addiction. So, quite a bit later,   we get this. And I have to say that this  is an excellent assessment of the paper,   especially that it only read parts of it. In  short, these are just two types of wood and in   laboratory conditions, and a bunch more  great insights. I absolutely love this. So this is the story of our little eccentric  genius. It can summarize videos that are many   hours long, be a coding teacher, do market  research, research research, and a ton more.    And I bet that just one or two more papers down  the line, and it will be way faster and more   reliable. And for me - this paper refers to dozens  of previous works, so I can have a glorious time   going through them and learning a ton. Thanks  ChatGPT, I can’t wait! What a time to be alive!

Thanks for watching and for your generous  support, and I'll see you next time!